Pinduoduo and Vertically Integrated Social Commerce

How the son of factory workers grew Pinduoduo from Zero to $100 billion in five years.

In 2015, Colin Huang founded his third company, Pinduoduo (PDD). By June of 2020, it had become China’s second largest ecommerce company and was valued at over $100 billion in the public markets. How did a company that helped farmers sell fruit on the internet rise so fast in a market dominated by Alibaba and JD?

Pinduoduo, meaning “together, more savings, more fun”, eliminated layers of middlemen and flipped the retailing model from being supply-driven to demand-driven. The team used a mobile-first approach that gave it a fundamentally different product DNA than incumbents. It used fruit as a wedge to combine consumption with entertainment and created a vertically integrated gaming company. It took advantage of down payments from suppliers and used stretched payment terms to create float out of customer transactions. It used that float to fund customer acquisition, and then leveraged clever growth tricks on an emerging distribution channel (WeChat) to acquire hundreds of millions of overlooked customers for practically free.

Humble Beginnings

Colin grew up in Hangzhou, the home of Alibaba located in the Eastern Chinese province of Zhejiang. His father never finished middle school and worked in a factory with his mother. Colin excelled in math. At 12 he was invited to the Hangzhou Foreign Language School, attended by the children of the cities’ elites. He credits this to changing the trajectory of his life. He was among the top students at the school and received a scholarship to study Computer Science at Zhejiang University, one of China’s oldest and most prestigious schools.

He joined the Melton Foundation his first year and secured an internship at Microsoft China making $900 per month - more than his parents combined annual salaries. He then transferred to Microsoft’s US HQ, making over $6,000 per month.

In college, Colin met NetEase (gaming) founder William Ding after helping him with a coding question in an online forum. This serendipitous meeting changed Colin’s life. William introduced him to many other Chinese tech luminaries like Tencent (WeChat) founder Pony Ma, and SF Express (logistics) founder Wang Wei.

Colin then moved to the US in 2002 to pursue a Masters in Computer Science from the University of Wisconsin-Madison. By graduation in 2004 he had a full-time offer from Microsoft, and an impressed professor wrote letters of recommendations to the large US tech giants of the time (Oracle, Microsoft, IBM).

The summer before moving to the US to start at Wisconsin, William at NetEase had also introduced Colin to Duan Yungping, fellow Zhejiang University alum and founder of BBK Electronics. The two grew very close. Colin considers him a close friend, mentor, and he even helped Duan with his investing. Duan recommended he move to San Francisco to work at a promising young startup. Colin then turned down all of his other offers to join a pre-IPO Google.

Colin joined Google as a software engineer working on early ecommerce-related search algorithms. He quickly became a Product Manager. In 2006, Duan won the annual charity auction for lunch with Warren Buffett with a $620k bid. Colin joined alongside Duan’s wife and five other friends. It's said that this meeting with Buffett greatly influenced Colin’s crafting of the Pinduoduo business model. This included the power of simplicity, utilizing float, and redistributing wealth (as Buffett has famously pledged to donate 99% of his wealth after death).

Colin returned to China shortly after to work on a secret team launching Google China. He reportedly grew tired of constantly flying back and forth to the US pitching Google founders Larry and Sergey on trivial matters. The last straw was a trip to get in-person approval of a change in the color and size of Chinese characters shown in the search results. He left many of his unvested options behind and Google eventually shut down the division. Colin then followed many of his mentors into a journey of entrepreneurship.

The Birth of a Serial Entrepreneur

In 2007 Colin founded his first startup,, an ecommerce site selling mobile phones and other consumer electronics. His mentor Duan’s company was a large player in the Chinese electronics supply chain. Duan was an angel investor and likely helped in the early days. Colin built up Oaku to “several hundred millions of yuan” in revenue (~$20-40 million USD), but he stepped down and sold the company in 2010 after realizing JD’s scale would always grant it better terms from suppliers and he could never beat them.

Almost immediately, he brought members of the team to build his second company: Xunmeng. It was a gaming studio that built role playing games on Tencent’s WeChat. Some ex-Oaku and future Pinduoduo employees launched Leqi, which helped companies market their services on other ecommerce sites (like Alibaba’s Taobao and JD). Both companies took off. Then Colin got sick.

He had an acute form of Otitis media, which causes severe inflammation and pain behind the eardrum. This typically causes a loss of appetite and occasional fever, and Colin specifically struggled sleeping. He stopped going into the office, and briefly retired in 2013 at 33 years old. He spent over a year at home. He considered moving to the US to open a hedge fund. He also thought about starting a hospital after going through the painful process of treating his ear infection.

Over the next two years, Colin came up with the idea for what became Pinduoduo by observing China’s two largest internet giants: Alibaba (ecommerce) and Tencent (social, games). He’s quoted as saying "these two companies don't really understand how the other makes money." Both are massive, successful companies, however neither had figured out how to penetrate the others business.

Pinduoduo fell directly in the center of the two; social gamified ecommerce. It helped manufacturers cut out middlemen by selling discounted items directly to low income consumers, and monetized largely with advertising. It fell within the intersection of unique insights Colin gained growing up poor and every previous business he, his mentors, and his team had worked on.

Pinhaohuo: Selling Fruit in WeChat Group Chats

Pinduoduo was initially founded in early 2015 as, or Pinhaohuo (PHH, “piece together good goods”). PHH’s initial business model consisted of buying fruit in bulk from farmers and then selling it directly to consumers. China’s fresh fruit market was growing fast in 2015, but less than 3% was sold online. Colin raised an angel round from his mentors, and once again brought over the team from his prior companies. Many were lifelong friends, including current members of PDD’s management team like Sun Qin, Lei Chen (first CTO, now CEO), Zhenwei Zheng, and Junyun Xiao.

Pinhaohuo’s business grew entirely through group chats on Tencent’s popular WeChat (often called the Facebook of China). To start, they bought boxes of fruit from a local Hangzhou fruit market and separated them into smaller boxes. On April 10th of 2015, they spent a few hundred USD to run one ad on an official Hangzhou WeChat Account (similar to a Facebook Page) that showed up in users’ feeds. They had more than a thousand employees, relatives, and friends of the company share the post. By May 1st, they’d fulfilled a total of 5k orders. Daily order volume surpassed 10k soon after. They paid an average of $0.30 cents for each of these earliest users.

Pinhaohuo also relied heavily on WeChat Pay, WeChat’s in-app digital wallet that had launched in 2013. Most users carried a balance due to the popular Red Envelope feature, in which users sent small monetary gifts to family and friends during the holidays. Routing all payments through WeChat Pay provided extremely low payment fees, low friction for order placing, and PHH’s low order sizes enticed early customers to pay with their outstanding balances. Most importantly, Pinduoduo’s primary competitor today, Alibaba, had also banned its sellers from using both WeChat and WeChat Pay. Its biggest incumbent competitor was un-incentivized to react to this newfound distribution channel.

Over 100 employees were directly responsible for sourcing, inspecting, and purchasing fruit directly from local farms. Initially, this model worked in Pinhaohuo’s favor. Fruit has a short shelf-life and this model reduced the number of supply chain middle men. Wholesalers, logistics companies, fruit markets, and supermarkets taking a cut to support unloading, re-loading, and high fixed overhead costs were all removed. Farmers could charge higher prices while also cutting prices for consumers by 80%.

Logistics infrastructure that was built over the prior decade to support Alibaba and JD’s operations made it possible for the company to grow so efficiently, so fast. As described later, many purchases were entertainment-driven with low intent, which allowed for slower, cheaper shipping. They partnered with SF Express for deliveries, which was founded by one of Colin’s mentors (and an angel investor in the company), Wang Wei.

By September of 2015, Pinhaohuo had became China’s #1 free app and was receiving over 100k orders per day. Then came a pinnacle moment in 2015 when it sold its first batch of Lychee fruits (below). Its single fulfillment center, built to handle 70k orders per day, broke down when demand suddenly spiked to 200k. The front-end of the business wasn’t well-connected to the back-end. The order printer jammed up and deliveries bottlenecked at the warehouse as SF Express wasn’t equipped to handle the demand. Meanwhile, orders kept rolling in. Most of the inventory went rotten and the team missed a majority of its deliveries that week. Daily order volume quickly dropped 80% to 20k and customer retention the following month fell to 25%.

Colin held an emergency lunch meeting in the warehouse. He instructed the team to inform customers their orders would be fulfilled. When many were delivered rotten, customers were refunded. They hired 100 new operations employees and opened six new warehouses. Finally, Colin beefed up the executive team and everyone began to re-architect the entire supply chain.

Pinhaohuo quickly automated the warehouses to intake, sort, package, and distribute inventory onto trucks to be shipped out. Within a month, most fruit spent only a few hours in any of PHH’s six fulfillment centers. The time from farm-to-table was often no more than two or three days. They also productized the supplier sourcing process. This made the business extremely capital efficient and allowed PHH to hit escape velocity. The business ramped to over 300k fruit orders per day, and 600k soon after. In only a few months, they had surpassed the fruit businesses of both JD and Alibaba.

By the end of 2015, 10 million registered users were placing 1 million orders per day. Customer retention was 77%. Half the orders were still coming directly from WeChat. This all exceeded the local shipping capacity of SF Express and the team had to find more courier partners (some owned by JD and Aibaba) and automate their logistics business. This transformation proved crucial as many upstart competitors quickly entered and failed with Pinhaohuo’s initial model.

Pinduoduo and the Team Purchase Phenomenon

In 2016, Pinhaohuo merged with another company Colin had built, a game-like ecommerce platform called Pinduoduo. This combined PHH’s logistics expertise with PDD’s intimate understanding of the end consumer - which had roughly 70 million users playing 30 different social-based mobile games. By Q1 of 2017, the combined entity had completely wound down its operationally intensive direct sales model in lieu of an asset light marketplace. PDD took a 0.6% cut of all sales, or enough to cover payment processing costs. The majority of its revenue would eventually be generated from in-app ads purchased by merchants, as described later.

Pinduoduo was founded in the second half of 2015, around the time many of China’s first wave of group buying startups failed. Most launched in the early 2010’s to help brands unload excess inventory. Many were location-based services that deteriorated in quality with scale and single-player experiences that didn’t influence others. Popular items included discounts on consumer goods, meals at high-end restaurants, swimming lessons, yoga classes, and photography sessions. Brands were promised the discounts would bring new customers, but in reality no affinity was built towards any of these group buying products or their merchants: all consumers cared about was cheap prices. Of what were initially thousands of players, only a few were still in the market by the time PDD launched in 2015.

The largest was Meituan-Dianping. The company formed in September of 2015 when first-place Meituan (“Groupon of China”) and third-place Dianping (“Yelp of China”) merged, giving them an estimated 80% market share and near total control of what was an early and battered group buying market. Today, Meituan-Dianping’s main business consists of food delivery. It's begun layering on a very high gross margin (89%!) in-store, hotel, and travel business - indicating where Meituan’s strategic focus was in the years following the merger.

Another survivor was Juhuasuan (“extremely cheap”), which was the second player and incubated by Alibaba. It focused on connecting Chinese manufacturers with overseas customers and eventually found success as a part of Alibaba’s Tmall focused on high-end products. Vipshop was founded in 2008 as a site for in-season discount and off-season overstock clearance, and used special offers to grow into what is now a public company valued at $15 billion. When JD eventually entered the flash sale market, it tried to differentiate from Taobao by claiming to never sell fake or low-quality products.

This all left an opportunity for Pinduoduo in what looked like the unattractive, low-end of the market. The general consensus among experts was that China was going through a consumption upgrade. Consumers appeared to be moving from cheap goods to quality and overseas imports. The combined Pinduoduo and Pinhaohuo teams began quickly experimenting with new products, including meat and seafood. One of their biggest experiments was team buying.

Pinduoduo’s model was simple: buy everyday items and receive discounts of up to 90% by completing in-app actions or inviting friends to buy them as well. Prices were set by suppliers, and a user's very first purchase was almost always subsidized at a discount by PDD. Subsequent purchases could be made at the full-price, or discounted with a team or through in-app rewards. Users could join existing teams in the app, but prices were even lower if users started their own team. Early products were so cheap that there was very low friction to buy. The K-factor, or the average people a new user invites, was never less than 1. Any money spent to acquire one customer was ultimately acquiring multiple customers. As PDD grew, invites became less prominent. Purchases can now be batched together automatically (almost instantly), making every purchase a team purchase (though at slightly higher prices).

Where others failed with asynchronous purchases of “nice to have” high-end products, Pinduoduo created a synchronous mobile shopping experience equivalent to going to the mall with friends. Cheap essentials like toilet paper created loyal customers that invited their friends, worked together, and came back again and again. Failed competitors mostly had their own apps or websites; PDD was built entirely on top of WeChat. This loose network of friends allowed it to weave purchases made by friends as recommendations throughout the app - including “someone just bought this item” pop-ups that overlay products as users browse. While most ecommerce listings had reviews, few built-in these trustworthy word of mouth-like referrals from friends.

Socialized team buying also gave a psychologically different pricing structure than a product like Groupon. Instead of a listed price, Pinduoduo’s team buying approach meant everything was negotiable and most items could be earned for practically free with enough work. All team buys required an upfront payment that was refunded if the minimum team size was not met within 24 hours. This reduced the friction of initially committing to a purchase, while subconsciously committing buyers to work together to reach the target.

Team buying also flipped the traditional retailing model from being supply driven (“how do we sell what we’ve produced?”) to demand driven (“how much should we produce?”). This is a fascinating aspect of the business that I’ll touch on further, and it was only possible due to Pinduoduo’s mobile-first DNA.

Mobile-First Ecommerce

It's easy to claim team buying as the primary driver of Pinduoduo’s rapid growth. In reality, it was one of many subtle UI and business model changes that PDD used to become China’s second largest ecommerce company.

The biggest difference between Pinduoduo and incumbents was that the product was designed for mobile (web preview here). Instead of manually searching for products like on Amazon or Google, products were sourced to users in a feed similar to Facebook, Instagram, Twitter, or TikTok. In the early days, the app often had no more than 20 SKU’s at any one time. This allowed it to focus on a few core products while incumbents worried about customizing a long-tail of listings. It could then customize those products for consumers (as discussed later) after it hit scale, greatly reducing costs. It was positioned like a digital Costco while incumbents operated more like Walmart.

Pinduoduo (left) vs Alibaba’s Taobao (right)

Without an endless stream of search results, PDD’s users had a constrained choice. It also gave more accurate data on user behavior and interests that fed back into its algorithm to target users later. This set the stage for what eventually became a very lucrative advertising business very similar to the Facebook news feed. Being mobile-first allowed it to build a customer acquisition engine centered on messaging and social media, not email or SEO. It took desktop-first competitors years to react.

Like a traditional social company, Pinduoduo started predicting what users might buy. However unlike a social product that weaves ads into an unrelated experience, and ecommerce where there must be a related search prompt, PDD’s core product was designed around spurring purchases as part of the browsing experience. And its algorithm could push deals, not specific items. It became a virtual mall.

Pinduoduo’s product allowed new suppliers to quickly reach customers. Competitors’ interfaces were search-driven, required heavy investment in SEO, and often centered on multiple product displays. PDD’s feed-based interface placed more exposure on single products. This gave it full control over distribution to influence consumer purchase decisions in ways most favorable to PDD and its suppliers’. This was similar to how TikTok helped new creators quickly build social capital and earn fans:

TikTok could deliberately control the allocation of social capital to its most talented creators, both new or old, as new users poured into the app. Alex Zhu, founder and now Head of Product at TikTok, likens the process to creating a new country and giving a greenfield of opportunities to a new class of creators. Hyper fast user onboarding and no friend graph let it use the entirety of time spent in-app allocating social capital. Source

By February of 2016, Pinduoduo’s monthly Gross Merchandise Volume (GMV) exceeded ~$15 million USD. They raised a stealth Series A round from IDG and Lightspeed China in March, and a $110 million Series B from Baoyan Partners, New Horizon Capital, and Tencent (among others) in July. Shortly after, they surpassed 100 million Annual Active Buyers.

Tencent's involvement proved critical, as it allowed them to further invest in WeChat as a distribution channel. They could analyze the flow of social information and apply those insights to product recommendations, pricing, shipping optimization, and the product roadmap (like the clothing store below). Alibaba had also considered investing but moved too slow, allowing Tencent to start accumulating what became a 18% ownership stake at the time of IPO.

WeChat then launched its now popular Mini Programs in 2017. This allowed developers to build pared down apps (often less than 1 MB) that lived inside of WeChat. Many developers experimented with apps that complemented their existing products, but for PDD it offered a more sophisticated extension of the native distribution channel it built on top of WeChat.

By Q1 of 2017, Pinduoduo’s Annual Active Buyers had doubled again to 200 million. It then raised raised a $214 million Series C in February of 2017. GMV continued doubling every quarter and passed RMB100 billion, or ~$14.7 billion by the end of 2017, less than three years after it was founded. This was a historic milestone that took Taobao, VIP, and JD over five, eight, and ten years respectively to reach.

Pinduoduo continued growing at an incredible pace, primarily over WeChat. By July of 2017, it had served over 300 million users. In Q1 of 2018, it had over 230 million WeChat Mini MAU’s (Monthly Active Users). This represented 57% of all 400 million active Mini Program users, and more than the 166 million MAU’s that used PDD’s own app.

In April of 2018, Pinduoduo raised $3 billion at a $15 billion valuation. Tencent invested over $1.4 billion which further cemented PDD’s role as a favored member of the WeChat ecosystem. By the end of the year, over 1 million merchants were selling on its platform. 11.1 billion orders were booked in 2018, up 158% from 2017 and an estimated half of all incremental Chinese ecommerce orders added that year. Its most popular item? Tissue paper.

Fruit and agriculture remained a core pillar to Pinduoduo’s business as it retained customers and offered a launchpad for other product lines. Agricultural GMV surpassed RMB65 billion (14% of its total GMV) in 2018, making it China’s largest agricultural ecommerce platform. The agriculture business grew 114% in 2019 to reach RMB139.4 billion, with over 12 million direct and indirect agriculture suppliers reaching 240 million buyers (most of the active user base). In Q1 of 2020, orders of apples, cherries, kiwis, oranges, and strawberries all increased over 120% year-over-year. Rice, wheat, cooking oil, meat, dairy, and vegetables averaged a 140% increase.

While some of Pinduoduo’s incredible growth could be attributed to WeChat, one of the biggest shifts in Chinese consumer behavior has been from the “search, pay, and leave” model of traditional ecommerce to social commerce. That wasn’t just WeChat. Much of PDD’s recent success has been due to its game-like mechanics.

Half Ecommerce, Half Gaming Company

When Pinduoduo submitted its filings to go public in 2018, many were surprised it described itself as “Costco meets Disneyland”. It’s a strange comparison; and perhaps “Dollar Tree meets Candy Crush” or “a digital TJ Maxx” are more appropriate as the product stands today.

WeChat offered a tremendous arbitrage opportunity to initially acquire customers. In the midst of its meteoric rise, Pinduoduo was acquiring new customers for as low as $2 each, roughly 20x cheaper than the $40 paid by JD and Taobao. Over time, competitors figured out how to use WeChat for distribution, and PDD needed new levers to grow the business.

Citizens in China’s lower-tier cities where Pinduoduo amassed its initial user base have a lot of free time. Lots of that time is spent playing mobile games, and many of WeChat’s most popular early Mini Programs were games. It explains why the initial PDD concept took off so fast - it was really just a game.

The deals on Pinduoduo changed every day. The app had colorful photos and discounts were hidden everywhere. There was a wheel to spin for daily coupons, discounts for sharing invite links with friends, free products in exchange for reviews, and rewards for daily check-ins similar to Snapchat’s streaks. Some discounts lasted as short as two hours. This prompted quick participation and impulse buys. There was even a leaderboard showing who had saved the most money. One-tap payments and saved info after the first purchase made this entire process frictionless.

In May of 2018, Pinduoduo launched Duo Duo Orchard. It was a game to grow virtual trees inside the app. The more purchases, actions with friends, and time spent in-app, the faster the trees grew. When the trees were fully grown, the player was shipped a box of fruit. It was a virtual stamp card. One month after launch, 2 million trees were being planted per day. The game had 11 million DAU’s by June of 2019 and 60 million by December of 2019.

Unlike many social or gaming products that sell ads alongside non-related content, Pinduoduo’s entire experience is part of the monetization model. Every interaction builds up to a purchasing event. Many of the ads mobile games are monetized with eventually spurs some form of commerce. PDD built commerce directly into its games to capture more of the value chain. And much of that value went to its suppliers.

Consumer to Manufacturer: Eliminating Layers of Middlemen

Before the internet, most of manufacturing and retailing was “how do we sell what we’ve produced?” Even today, despite having direct relationships with consumers, many brands still order inventory from their suppliers up front before any sales are made.

In Pinduoduo’s early days, it helped farmers in small villages sell fruit directly to their neighbors over the community’s newly purchased smartphones. Combining fruit orders at the front-end helped PDD predict volume. It could guarantee sales upfront and reduce risk for suppliers; similar to government subsidies that support much of the world’s agriculture production. As consumers cultivated rewards in virtual games, it built in even more predictability. Farmers could optimize their harvests (for product quality, not shipment length), had more predictable income streams, and consumers paid lower prices for fresher fruit than they could buy in-person because so many layers of inefficiencies were removed.

As it amassed users, Pinduoduo began predicting, and then influencing consumer demand elsewhere. Its team buying model flipped the retailing model from supply-driven to demand-driven. It would be as if Facebook went a step further than partnering with direct to consumer brands dropshipping from overseas, and instead created a Shopify-like tool for those brands’ suppliers. In some cases, these were similar to a Pinduoduo private label. In others, manufacturers were building a consumer-facing brand for the first time. As Colin wrote in a now-deleted blog post: it eliminated economies of scale advantages, made small scale customized services viable, and allowed both small manufacturers and retailers to compete against larger competitors.

Taking learnings from Google and adopting the model used by Taobao (and pioneered by Pinduoduo investor and ex-Taobao CEO Sun Tongyu), Pinduoduo turned this into a marketplace. They called this “Consumer to Manufacturer”, or C2M. There was high fragmentation in the original supply of local farmers and the model scaled very fast. It guaranteed demand to suppliers, who would in-turn pay for that demand. Suppliers could eventually bid on in-app feed placement, banners, links, logos, and keywords in search results. They could control prices, group sizes, and how many orders they would ultimately fulfill. This allowed them to precisely predict their volume and input costs months ahead of time. It allowed small producers to reap some economy of scale benefits typically captured by only the largest players in an industry.

Pinduoduo also helped manufacturers customize their products. The direct relationship with consumers gave them insight into consumer behavior. They could cut costs by reducing demand mismatches, tweak clothing styles, redesign food packaging, accelerate product development timelines, launch entirely new products, and even integrate PDD’s data directly into their manufacturing lines. Similar to how marketers might use Google keywords to predict demand, Chinese suppliers could do the same. This process eliminated layers of middle men across a host of industries and let producers (many without their own brand) sell directly to consumers.

Suppliers paid Pinduoduo an upfront fee when first listing on the platform to guard against fraud (they were also fined 10x the value of any counterfeit goods they were caught selling!). They were required to pre-pay for advertising that was redeemed over time, and received payment from PDD an average of 15 days after the goods were shipped out of their factory. Since payments from customers were collected at the time of sale, this gave PDD a negative cash conversion cycle, or float, that collected cash before actually delivering any goods or services (as highlighted below).

For a company like PDD that might see its business grow 25-100% in any given month, receiving cash up front from consumers and paying its bills two weeks later had massive downstream implications on its capital needs. This negative working capital provided excess cash to subsidize user acquisition and allowed the company to mitigate shareholder dilution as it scaled.

This was all scaling at a time the Trade War with the US was evaporating foreign demand for Chinese manufacturers. Chinese exports slowed in 2018, and February of 2019 brought the lowest of Chinese export volume in three years. The COVID-19 pandemic to start 2020 also didn’t help foreign demand. The C2M model that was core to Pinduoduo’s business model began as a broader country-wide push to modernize the manufacturing industry, but increasingly became a tool to stave off a recession.

Many of the suppliers joining Pinduoduo’s marketplace were invisible-to-consumer factories that lost business selling products to overseas brands who then marked up prices 5-10x. They were familiar with online retail, had excess capacity, and some had even tried and failed to launch their own consumer-facing brands. PDDs ability to pool demand allowed them to sell similar volumes of products to consumers at similar pricing they were getting from big brands. These savings were then passed along to consumers and further reinforced PDD’s “more savings, more fun” flywheel.

Pinduoduo heavily recruited these local suppliers that were losing international revenue. It launched 106 of these “manufacturer owned brands” in 2019, and aimed to create 1,000 more in 2020. According to a Citi note “more than 1 million merchants operated on PDD’s platform to end 2018, but they include only 552 established brands. That compares with 150,000 established brands on Tmall.” Most of these manufacturers will gladly accept foreign demand when it returns, but will continue to benefit from PDD’s newfound domestic business.

Moving up the Consumption Stack

In 2018, Taobao had over 500 million users compared to WeChat’s 1 billion. The gap between the two represented mostly citizens in third-tier and below cities, many of them senior citizens.

This increasingly became a problem for China’s incumbent ecommerce players as Taobao’s penetration was as low as 1% in some of China’s smaller cities. Third-tier cities and below makeup over 50% of the population and are estimated to make up 66% of China’s consumption growth between 2020 and 2030. While upper-class consumers cared about international brands, this newer wave of consumers cared more about lower prices and quality products. Traditionally, over 55% of Singles Day GMV, or Chinese ecommerce’s annual discount day, came from third-tier cities and below.

Pinduoduo built its early business model around these young consumers in rural communities and China’s smaller cities. Meanwhile, incumbents tried serving them with drone delivery. PDD would first target rural customers around a city, get farmers and other merchants on board, and then move to the city center. In some cities, PDD saw penetration as high as 35%. Pinhaohuo’s initial user base was 80% female who likely drove most of their household purchasing decisions. For many of its earliest customers, PDD was their first experience using ecommerce.

Colin has said “People living in the five rings of Beijing wouldn’t understand our purpose. The new consumer economy isn’t about giving Shanghainese the life of Parisians. It’s about providing paper towels and good fruit to people in the Anhui province.” He uses Tian Ji’s Horse Racing Strategy as an analogy to Pinduoduo’s model:

Tian Ji and the king of the Qi Kingdom both like horse racing and race each other often. They frequently make bets. The king of Qi has better horses and Tian Ji loses every time. One day, Tian Ji’s friend says “take me to the race next time and I can help you win”. His friend learns that for every race, Tian Ji and the king both choose three horses classified as good, better, and best. There are three rounds, and the winner of the race is the one who wins at least two rounds. Both of them were using their “good” horse for the opponent’s “good” horse, “better” horse for the opponent’s “better” horse, and “best” for the “best”. The king had a slightly better horse in all three levels, and won every round. Tian Ji’s friend then brings up an idea: use Tian Ji’s “good” horse for racing the king’s “best” horse, then use the “best” horse against the king’s “better” horse, and the “better” horse against the “good” horse. As a result, Tian Ji loses the first round, but wins the second and third round, winning the race.

Similar to Tian Ji, Pinduoduo noticed that incumbents were overlooking lower income consumers. These consumers flocked to PDD’s low prices which allowed it to operate under the radar while building a model that allowed it to quickly scale up market, catching competitors off guard. In Q1 of 2019, 44% of new users came from first and second-tier cities. Over $1 billion in agriculture products were sold in the first two weeks of June (up 310% year-over-year), 70% of which were bought by consumers in these large cities. By November, 45% of PDD’s GMV came from Tier 1-2 cities.

The Ongoing Battle for the Chinese Consumer

By March of 2020, over 630 million active buyers had made a purchase on Pinduoduo in the past year. The app had 488 million MAU’s, with many still accessing its services solely through WeChat.

Pinduoduo’s growth did not go unnoticed. By 2018, JD launched its C2M site, Jingzao. JD also launched a near-clone of PDD’s app called Jingzi at the end of 2019, which specifically targeted smaller cities with team buying discounts.

Alibaba launched its Teja group buying app in March of 2018. It announced similar C2M initiatives two years later, with plans to integrate consumer data collection for suppliers into every aspect of its products. It's also focused on negotiating exclusivity with these suppliers to keep them off of Pinduoduo’s platform. Its agricultural flash sales platform Juhuasuan launched in 2019, headed by Jiang Fan (below, left), a 34-year old executive in charge of both B2B Tmall and B2C Taobao. Industry experts believed he was a candidate to eventually lead Alibaba and its battle against Pinduoduo; however he was recently demoted due to personal misconduct.

Startup competitors like Taojiji also emerged in late 2018, gaining 11 million users in two months. One year later it had 130 million registered users, but struggled with the retention PDD had mastered, and was out of business by December of 2019. All of these competitors burned hundreds of millions in capital to move down market towards a consumer base Pinduoduo had already secured. At the same time, Pinduoduo was spending to move up. Its entire business was built around this C2M model, meaning its value proposition for suppliers increased as it grew.

One of Pinduoduo’s primary strategies to move up market was TV ads. Many consumers knew of PDD but were skeptical of the product quality. They used exclusive product drops tied to specific shows, celebrity endorsements, aggressive promotions, and in-app countdowns to convince users to try the app. Starting with fruit, this built trust in what has now become 18 broad categories like clothing, household goods, appliances, and electronics - all of which have much higher prices and been long dominated by Alibaba. The below commercial aired on one of China’s hottest TV series, the Voice of China, exposing them to millions.

The strategy worked. In January of 2019, 37% of GMV came from top-tier cities. By June, this increased to 48%. In Q3 of 2019, users in Tier 1 cities spent 3.2x more per transaction than the average user. Young users opened Pinduoduo 89 times per month in 2019 (up from 84 in 2018) compared to Taobao’s 81 times per month. DAU’s were using the app over 20 minutes per day. 

PDD’s marketplace model kept costs flat as incremental revenue on larger orders fell to the bottom line. These new users also aggressively invited their friends. This meant that, while paying similar ad prices, PDD’s ROI was almost always higher than its competitors. It consistently spends almost all of (and sometimes more than) its reported revenue in a quarter on marketing. This is essentially PDD using the two week float provided by the 5.1 million merchants in Q4 of 2019 (up from 1 million in 2018) to acquire customers before paying them out. It was significantly under-earning too. PDD’s revenue in Q1 of 2020 was only 3% of GMV compared to JD’s 28%.

What’s next for Pinduoduo?

Over the next decade, we’ll see Pinduoduo double-down on selling to Chinese consumers. They’ll add new gamified features that predict future purchases while building on top of sticky fruit orders, increasing order sizes, and driving prices lower. It will go deeper with proprietary supply agreements, continuing to build what’s equivalent to a PDD private label. We’ll likely see more products that pool demand to be sold to the highest bidder, and some that require upfront payments or have long payment terms that increase PDD’s float.

Pinduoduo started experimenting with travel deals in 2016, and officially launched a travel business in late-2019. These included tours, vacations, and hotel rooms. In June of 2020, it added trains and domestic flights.

Pinduoduo recently announced a move into real estate. Users paid a ~$1k refundable deposit to take part in a team purchase on a 1,000 unit property under development where 600 units were sold. It appears unrelated to the core business at first, but real estate has a high GMV and typically precedes many other large purchases made by new families. Similar to growing a fruit tree, we may see the company design a complimentary game around the process of building a house or town that ultimately concludes with a purchase.

In 2019, it began heavily emphasizing live streaming. Analysts project China’s live stream market will generate $136 billion in transactions in 2020, up 121% from 2019. Producers broadcast for free and partner with influencers to push garments, cosmetics, and household products. It’s relatively non-scheduled in nature and works well with PDD’s pop-up and flash deals. Consumers put more trust in the product quality when purchased directly from a live stream of the warehouse (below). Many of these warehouses featured foreign goods that had just cleared customs, inviting viewers to ship them directly to their home. And natural to PDD’s model, users earn discounts by inviting friends to watch.

A live streamed event in June of 2019 sold 400 cars over Pinduoduo in 18 seconds. PDD’s recently launched online pharmacy sells both OTC drugs and medical devices. It’s a market fraught with inefficient middle men and pharmaceuticals represent a recurring, sticky product like fruit. In April of 2020 PDD invested $200 million in appliance and electronics retainer GOME, who it now partners with for in-store product demonstrations. It's possible the newly launched DD Bank game precludes financing and insurance related to these larger products. And it will likely integrate consumer usage data directly into many of them, hinting PDD’s custom appliances may one day automatically restock food directly from its farmers.

We may also see Pinduoduo open itself for other retailers to tap into its direct relationship with manufacturers. It also appears to be doubling-down on helping foreign brands sell into China. Amazon may have taken this first step in both of these initiatives, opening a store in November of 2019. It featured 1,000 foreign brands across health, beauty, apparel, and electronics. Amazon has historically struggled in China. Its market share shrunk from 15% in 2012 to less than 1% by 2019. This partnership gave it immediate distribution to hundreds of millions of Chinese consumers. For PDD, it boosted its image within China, giving it instant credibility selling foreign-brands from a trusted source. It also gave PDD access to Amazon’s global network of suppliers and logistics providers.

Pinduoduo recently created its own logistics tracking platform. Similar to Alibaba’s Cainiao which PDD previously used, it won’t own its own fleet or warehouses and instead links together third party providers. It's already the second largest logistics network in the world behind Cainiao. This gives PDD more granular insights on package location and will let them optimize order routing. They’ve opened this up to third parties for free. Couriers will likely be able to bid on demand and earn revenue on previously empty trips like delivering products from an urban center into a village and then bringing produce back into the city. This may allow for faster deliveries and make PDDs produce even fresher. It starts a road down merging online and offline behavior, as described in this video.

Like many other Chinese tech giants, Pinduoduo rolled out its collaborative enterprise software platform called Knock in February. The main features include messaging, to-do lists, basic photo tools, and voice calls. It excludes features popular on other similar products like video conferencing, file transfer, and task / expensing approval. Everything is linked to the back-end of PDD, and hinting it will have benefits similar to Slack’s shared channels for many of PDD’s suppliers.

Colin is personally very passionate about improving the efficiency of Chinese agriculture. Pinduoduo is building 1,000 agriculture plantations which will expand it into coffee, tea, and walnuts among many others. PDD built custom software for farmers to run their businesses and launched Duo Duo University to teach them how to run a business and improve their crop yield. The Chinese labor force is aging, and the new wave of workers will want standardized workflows and quality control on their phones. More efficiency will further reduce PDD’s prices and make its product stickier with suppliers and consumers over time. Its not a stretch to think Pinduoduo could one day provide the software powering most of China’s agricultural and manufacturing production.

Pinduoduo has also started building more in-app games. One in particular is a Farmville-like game called Duo Duo Farm. Another is Duo Duo Crush, a Candy Crush-like game where nearly every action incentivizes a purchase.

One of their newest games is Duo Duo Piggy Bank. Users collect virtual coins by inviting their friends, browsing products, and completing tasks. Once they have enough coins, they can start manufacturing a product in the Dream Factory. This involves performing more tasks to generate electricity to keep the factory running. Finally, they earn extra fuel to expedite the shipment.

These are likely the beginning of many other gamified experiences Pinduoduo will build into its product over time.


Pinduoduo still faces some high-level challenges. China’s discount market has a noted problem with counterfeit goods that are popular with lower income consumers both on and offline. Despite being smaller, PDD generated more complaints than Alibaba’s Taobao and Tmall combined in 2017. A week after its July 2018 IPO, it met with government officials to discuss how to handle these complaints. Short-sellers have made valid claims that PDD’s financial reporting may overstate GMV and Revenue to US investors, and that its hiding employee costs in an undisclosed related entity. The company does not have a CFO (fairly common in Chinese tech), which doesn’t help these allegations. And Colin recently stepped down as CEO to move into a chairman role and began giving his stock back to the company to grant to other employees. Public market investors should be sure to do their diligence on these matters.

Pinduoduo’s rise has been nothing short of impressive. It represents a case study of an excellent team quickly building and scaling a transformational business. Like Amazon used books, PDD used fruit as a wedge to build what became one of the world’s largest ecommerce companies in five years. It went a step further than Tencent building games, or Facebook selling ads, and created an asset-light vertically integrated social gaming company. It built farmers, and eventually manufacturers, a business-in-a-box solution for operating their companies. And it entertained consumers while giving them low prices on products they buy everyday.

If you liked this, please subscribe above for more posts. I’m also on Twitter at @TurnerNovak.

Thank you to Hans at GGV for pointing me to early resources; Ben and David at Acquired for their excellent recent episode on Pinduoduo; and James at Lightspeed + @JoshConstine and @Leonlinsx in the Type House for feedback on early drafts.

Due to the potential inaccuracy and translation errors in some foreign publications, I cannot personally verify every figure shared here. I have linked to all sources wherever possible. This is not investment advice and I currently have $0 invested in securities tied to Pinduoduo.

Snap Partner Summit 2020: Recap and Reflections

Snap opens itself up to developers and adds more features that will increase users, advertisers, and monetization potential

Today’s Snap Partner Summit (replay) gave us new Snapchat usage metrics and more importantly, features that will be core as Snap opens itself up to developers and expands its products beyond the Snapchat app. Many of these new features appear to bolt into Snap’s Ad Platform, giving advertisers access to over 229 million users that use Snapchat every day.

Impressive New Snapchat Usage Statistics

Snap reminded us it still reaches more 13-34 year old’s in the US than Facebook or Instagram, and the 100 million users it reaches in the US is more than Twitter and TikTok combined.

Snap said 125 million people got news from Snapchat over the past year and “hundreds of millions” watch content on Snapchat Discover. Snap announced that 20 Snap Kit apps are now in the Top 100 of the iOS and Google Play Stores, up from 15 in March. The Snap Map now reaches 200 million MAU’s (Monthly Active Users), and Snap Games have been played by over 100 million unique users. They also disclosed DAU’s (Daily Active Users) in India were up 120% year-over-year (YoY), though likely on a base of only 4-5 million.

Snapchat: the Camera, AR, and the Snap Map

Snap reminded us that over 30 times per day, 135 million people create AR with its camera daily, and 170 million interact with it. A big announcement was SnapML. This allows developers bring their own machine learning models into Snap Lens Studio. They highlighted language learning and commerce as early use cases. They also announced 3D face meshing, skeleton tracking, and AR material programming. Importantly, Snap also announced products built in Lens Studio work in all Snap SDK’s - inducing its AR glasses, Spectacles.

Snap also disclosed a new social feature called Local Lenses where users share persistent AR experiences within a neighborhood. It looks like an expansion of the Landmark feature it announced last year.


It also hints a push into supporting local SMB’s, and how Snap may monetize the Snap Map. It suggested the Snap Map will become more contextualized, with plans to offer features similar to Google Maps: hours of operation, reviews, Snap stories unique to the location, and delivery options like Postmates, DoorDash, Uber. This sets the stage for integrating both advertising and miniaturized apps that live within Snapchat (more below).

Surprisingly a little early than expected, Snap also announced Alexa-like voice-enabled lenses called Voice Scan. This builds on its Scan Platform through a partnership with SoundHound, which it also announced is now linked to Lens Studio and thus Snap’s AR Developer Platform. They also threw in some teasers (see below) on how integral commerce will become to the Snapchat Camera.

Discovery has always limited Snap’s AR products. It was first done through the in-camera carousel, and Lens Explorer (“Browse” in the image above) which launched two years ago. User-created lenses can now be tagged with voice commands. Voice Scan is significant as over 1+ million lenses have been created by the community to date per Snap, up from 900k as of March 31st. The highest viewed UGC lens appears to have over 14 billion views based on Snap’s disclosures.

Snap also rolled out a action bar that will live at the bottom of the app and contextualize based on the user and how they’re using the app. It looks like a subtle change, but undoubtedly makes it easier for older users to use Snapchat as it continues bringing them into the app. It also elevates the Snap Map as a core product, which should become a focal point as Snap’s AR products continue leaving our phones.

All of these new features have interesting applications today, and are also building blocks for the next decade. Snap has become more vocal lately about its intentions to move into spacial computing, or projecting things onto the physical world around us. All of these products, which are already used by most of the youth in the developed world today, are first steps that will eventually converge outside our phones.

Content: More Reach in the US than Netflix

Snap announced a few new content features. Specifically, the Happening Now bar, which now sits at the top of the Discover page (far right). They also expanded on their participatory AR content, allowing users to become part of the show.

It re-highlighted metrics it’s driven home in recent earnings calls. Time spent watching Discover was up 35% and time spent watching its Original Snap Shows doubled YoY. Specifically, it highlighted Will Smith’s show “Will From Home” has now been watched by over 35 million people, up from 15 million in Q1. For context, he has 46m followers on Instagram and 29m on TikTok.

As mentioned above, 125 million people got news from Snapchat within the past year, and “hundreds of millions” now watch Snapchat Discover. Snap also disclosed that Snap Originals (which it described as “lasting franchises”) are now watched by over half of all Gen Z in the US and 70% of people who viewed one episode of Dead of Night came back to watch the entire series. These all seem very relevant compared to the 70 million subscribers Netflix has in the US, per Statistica.

Bitmoji and Snap Games: A Distributed Social Network

Snap officially re-announced Bitmoji Games, a cross-platform SDK to bring users’ Bitmoji into other publishers games on nearly every platform: PC, Mac, Xbox, Playstation, Switch, iOS, Android, and others.

This allows for quick personalization, and continues laying the groundwork for a back-end giving developers access to Snap’s friend graph, easy in-app purchases, and a distributed metaverse, social network, and digital economy that spans Snap’s developer ecosystem. Bitmoji is uniquely positioned to do this as its already used by practically all would-be early adopters and its IP is inherently cross-platform: 2D, AR, and eventually VR.

Snap disclosed that over 100 million people have played Snap Games, and play an average of 20 minutes per day. Ready Chef launched six months ago and has been played by over 25 million unique users. Play sessions are 2x longer when with a friend, proving that no install, built-in monetization, attribution, in-game chat and voice, and a pre-built friend graph can be very powerful for developers. They announced new partnerships with four new studios, including Voodoo, a French publisher that has over 3.7 billion app downloads across its portfolio.

Most importantly, all of Snap Games are built on HTML5, which is likely architected for cross-platform distribution as Snap opens itself up to developers and eventually launches its own hardware.

Snap Mini’s: Other Apps in Snapchat

In arguably the most important announcement, that Snap used the same HTML5 base as Snap Games to launch Snap Mini’s. These are pared apps that live within Snapchat’s chat and accessed from the same interface as Snap Games, which has already been used by over 100 million people. After launching in chat, Mini’s will likely hit the Snap Map and AR lenses next.

Early Snap Mini’s are from a set of curated partners: Headspace, sports betting, booking concert and movie tickets, class schedules on Saturn, registering to vote, and collaborating on flashcards to study. Like most of its products, Snap likely keeps this closed for a year or two before opening it up for any developer to build on.

Bitmoji, which Snap acquired for $64 million and may have north of 300 million MAU’s, also has a keyboard that could potentially plug Snap Mini’s into any other app. They specifically announced a partnership integrating Bitmoji into all Samsung keyboards.

If you’re familiar with Tencent’s WeChat, Snap Mini’s will look familiar. WeChat Mini Programs facilitated over $115 billion in transactions by 300 million people in 2019 (25% of the user base), up 45% YoY. Snap could go beyond WeChat by eventually extending these outside Snapchat, and they’ll likely be interoperable with other pieces of the Snap Platform and the eventual Snap Operating System that will live on its hardware products.

Snap Kit: The Snap Platform

Much of the above ties into Snap Kit. It now has 800 integration partners (~125 added since Q1) and 148 million MAU’s. Specifically, 20 Snap Kit apps are now in the Top 100 of the iOS and Google Play Store’s, up from 15 in Q1.

Snap announced dynamic AR lenses in Creative Kit. This allows developers to use Lens Studio to build and integrate the creation of a Snapchat lens in their own app. Two examples given were sending a snap to a friend inviting them to join a Houseparty room and a 3D model allowing users to swipe up and purchase a fractional share of a vintage car on Rally Rd.

Finally, Snap officially announced Camera Kit. This officially opens up Snapchat’s camera for other developers to bring into their own app. This builds upon the Snap Camera for desktop with Triller and Squad as early launch partners. Its not hard to see how these products arm the long-tail of consumer social rebels and help both Snap and all of its partners generate more long-term value for their businesses.

Snapchat and the Power of Friends

Behind all these product announcements is a reminder Snap is building on a very strong competitive advantage: a messaging network of hundreds of millions of close friends in markets with high per capita income.

In the short-to-medium term, all these products increase Snap’s monetization potential. More video ads in content and games. More AR ad inventory and commerce opportunities as it opens up its camera. Monetization from the long-tail of business across the world as they open up the Snap Map and Snap Mini’s. It’s likely these features also drive new users into the app, which continues to build on the user acquisition flywheel Snap created with content and AR.

Over the long-term Snap Kit, the Bitmoji SDK, and the Bitmoji keyboard spread Snap’s products into every other app. We didn’t get any updates on Ad Kit, Snap’s Third-Party Ad Network built into Snap Kit, however, it will likely integrate with all Snap’s other distributed products over time as well. And voice will become increasingly more important in the context of hands-free computing with Spectacles.

One of the biggest tests for Snap will be figuring out a monetization model beyond advertising. No one has cracked this outside of China yet, and Snap’s messaging network of close friends makes it possibly the best positioned to do so.

Attack of the Clones: TikTok’s Rival Kuaishou Lands in the US

Plus: Developers can build apps inside Snapchat, Microsoft opens up Office

On Wednesday May 27th, a new app hit the top of the US App Store: Zynn. I threw some initial thoughts on Twitter suggesting this may be a big deal since it was backed by Kuaishou (and ultimately Tencent) which was later confirmed by The Information.

As I mentioned, Zynn is essentially a bare-bones TikTok clone with a reward system. The camera tools feel much more like Instagram than TikTok and there is no ability to search by video sound (or search for anything), a key part of TikTok’s DNA.

The initial content appears to be ripped directly from TikTok. The upload dates on these videos go back as February and March, showing how the app was seeded for three months before launching. Most accounts are related to themes like snowboarding, street racing, or food (for data targeting purposes?). Some accounts (1, 2, 3, 4) appear to have re-uploaded the most viewed videos of a top sound on TikTok. Some combine posts from multiple TikTok influencers (Zynn, TikTok). There’s an account re-posting videos from a beauty influencer as she uploads to TikTok (Zynn, TT). Many of the initial accounts follow each other, and most profiles like exactly 6 or 7 videos. Clicking on these Zynn links, the videos can’t be viewed on desktop and the mobile web experience forces you to download the app. This is a stark contrast to TikTok’s approach.

Zynn began climbing the iOS charts in May. It's unclear when exactly it launched on Android, but it hit the top of both the Google Play and App Store (#1 overall) charts as of Friday, May 29th. It doesn’t seem to be slowing down either: total downloads on iOS were ~241k on the 27th and 28th, and Android increased from 19k to 28k (per App Annie).

The app’s aggressive reward system likely propelled its rapid rise. Users earn small amounts of cash watching videos (around $1.20/hr), bonuses for hitting certain thresholds, and as much as $20 for each friend they invite.

The very big catch behind all of this is the minimum withdrawal amounts and that you can never actually redeem all your rewards. There are some big totals, like $50 gift cards to Amazon, Walmart, and Domino’s. This practically requires users to invite their friends and stick around for while if they want their cash.

Normally I wouldn’t suggest something like this was sustainable, but The Information confirmed what I figured out: Zynn is owned by Kuaishou, ByteDance’s (which owns TikTok) closest rival in China. And Kuaishou itself is backed by Tencent (one of ByteDance’s other largest rivals), Baidu (a Chinese search engine), and Sequoia (an investor in ByteDance) - and just raised $3 billion from them in December.

An Overview of Kuaishou

Kuaishou was founded in 2011 by Su Hua and Cheng Yixiao. They previously worked at Google and Baidu. It launched as a way to create and share GIF’s and quickly evolved into short videos. By 2013, it had over 100 million DAU’s (Daily Active Users); three years before the Chinese version of TikTok (Douyin) launched in 2016.

ByteDance and Kuaishou have a long rivalry. I wrote about ByteDance two weeks ago. One recount of the story describes Kuaishou as having a similar multi-app strategy as ByteDance, although moving slower. Kuaishou’s initial users were rural and working class, often called “the forgotten 80% of China”. Most of the content in its early years was these non-urban users doing weird stunts. Analysts often described Kuaishou as “Jackass for mobile”. One of its most popular creators during this time was a woman who ate and drank random objects.

Combined images show the homepage of ‘Gourmet Sister Feng’ (left) and a live stream of her eating a light bulb (right).

By 2016, Kuaishou was at the top of the Chinese short-video market. Kuaishou’s revenue lagged ByteDance’s first breakout app Toutiao because its advertising product didn’t launch until Q3 of 2016, two years after Toutiao’s. This likely impacted how much revenue and cash it had to re-invest in growth. It also struggled gaining traction outside China in major markets like India and Indonesia where TikTok (launched in 2017) now thrives. Like ByteDance, Kuaishou went through many periods where its products were blocked by WeChat (nearly six months in 2018!), which impacted its growth as well.

A pinnacle moment in this rivalry came in 2017. Kuaishou was reportedly out negotiated by ByteDance for after it refused to also buy News Republic and Live.Me, two other products backed by Fu Sheng, the founder of Cheetah Mobile and influential angel investor. ByteDance bought all three; and subsequently merged and TikTok while de-prioritizing the other two. TikTok went on to consistently top the download charts in nearly every international market. It was late to enter Latin America, and Kuaishou’s international app Kwai eventually gained traction in Brazil. Latin America still appears to be a focus. In June of 2020, its Snack Video app cracked the top 20 in Indonesia, Peru, and Colombia.

By Q1 of 2020, Kuaishou’s short-video product reached over 300 million DAU’s in China, up from 200 million in December. This was lower than ByteDance’s 400 million DAU’s in China and 1 billion globally, but still impressive relative to other large digital platforms. It reported revenue of $7.2 billion in 2019 (beating its initial target of $4.36 billion) in its eighth year in business - better than Facebook. This was primarily through taking a cut of live streamed commerce, a market that’s projected to reach GMV of $136b in China in 2020, an increase of 121% year-over-year.

Coming back to Zynn, suggesting we take it (or at least Kuaishou’s attempt to enter the US) seriously isn’t that crazy. Kuaishou’s core product likely has operating margins in the range of 30-40%. This would have generated nearly $2 billion in cash in 2019 to re-invest elsewhere - on top of the $3 billion it just raised at a $28.5 billion valuation.

US consumers are generally the holy grail for internet businesses. As shown in the chart below of Facebook’s ARPU (Average Revenue per User), North American users bring in 3x more revenue than Europe, 9x more than Asia, and ~14x more than the Rest of the World. This could be further defined as “developed markets” since RoW and Asia-Pac includes countries like Australia, Japan, and South Korea, which have similar income levels to the US. The US also typically drives Western culture, which makes it the perfect market to seed content for similar markets.

For a Chinese company printing cash and looking globally for growth, cracking the US is likely a high priority. The big US players like Facebook and Google don’t exactly move fast, and neither them nor Amazon or TikTok have business models built on live streamed commerce. ByteDance saw success in 2018 and 2019 with TikTok, and we’re now seeing Kuaishou attempt the same with Zynn. The majority of Kuaishou’s Chinese revenue is generated via live streamed commerce, and its user base primarily lives in smaller cities and rural areas of China. This contrasts with ByteDance’s beachhead in China’s richest cities, starting with advertising and expanding into ads and commerce. This difference may foreshadow the strategy each company takes in the US.

Is Now the Perfect Time for a Cash-Based Referral Strategy?

In the past 11 weeks, the US has seen an unprecedented 40.8 million unemployment claims. While many jobs will come back as the world leaves lockdown, many Americans are currently stuck at home all day with no job.

Zynn’s referral and “pay to consume” user acquisition and retention strategy has been tried before by another Chinese company QuToutiao, or “Fun Toutiao”. If you’re familiar with ByteDance, you know its initial product was a hyper targeted news app called “Toutiao”. QuToutiao (launched in 2016) is a Toutiao clone that pays users to use the product. They ended 2019 with $778 million in revenue and 46 million DAU’s who spent over 60 minutes per day on the platform. Its publicly traded (QTT) and valued at $750 million, or ~1x its 2019 revenue - not exactly a premium compared to Kuaishou’s 4x, Facebook’s 8x, or Snap’s 15x. This is likely because investors believe QTT will always have to pay out the majority of its revenue to users and won’t be able to increase its margins. Its sales & marketing expense was 80% of revenue in 2019.

This “pay to consume” model has also been tried by Kuaishou itself. It launched Kuaishou Speedup in August of 2019. It paid Chinese users to watch videos, very similar to Zynn, and reached over 10 million DAU’s within the first 20 days. We could be watching the same playbook here in the US. Many Chinese companies see FinTech as a key initiative, and Kuaishou recently launched a red packet product that allowed Chinese users to send monetary gifts to celebrate Chinese New Year. This combination of rewards, payouts, and P2P gifting could be early moves in a digital wallet effort.

The irony of Kuaishou copying ByteDance, by copying the rewards strategy used by another app that copied a different ByteDance product, while copying TikTok that ByteDance initially copied from Kuaishou, shouldn’t be lost.

We should also not forget that TikTok got to where is today through an unprecedented user acquisition strategy that hadn’t been tried yet at scale: dumping money into competitors’ ad products to grow its own user base. As I wrote:

ByteDance flooded the world with marketing. It reportedly spent $3M/day on user acquisition and PR throughout 2018 and 2019, beating short-video competitors in each market by simply outspending them. TikTok spent over $300m on Google ads in 2018; and $10m/month in India alone. In Q1 of 2019, 13% of all ads seen by users of Facebook’s Android app were for TikTok (Apptopia). At the peak of the ad blitz in September of 2018, 22% of all ads seen on Facebook’s properties by Apple devices in the US were for TikTok.

ByteDance gets lots of criticism for spending so much on ads, but it's worth resurfacing Elad Gil’s 2010 reminder that sometimes it's worth spending money to kickstart a valuable network. Facebook used a similar strategy with Instagram by favoring it in the feed, which helped it quickly convert new users… 50-75% of Instagram downloads ultimately came from Facebook in the years following the acquisition.

TikTok acquired’s 60 million MAU’s in 2017 and was likely much closer to Product Market Fit when it started its marketing strategy than Zynn is today. The initial TikTok product essentially started where Vine left off, and the flood of marketing allowed it to quickly level up its content and user base.

Public reports estimate TikTok paid ~$3-5 per user, which is 4-5x less than the $20 Zynn is promising per referral. Referrals are also an interesting strategy at this moment in time when considering digital ad prices decreased ~30% in Q1.

Zynn: A Mobile-First Live Streaming Platform?

Less than one month after launching, there are early signs Zynn’s strategy could be working. As of June 4th, I saw videos in my feed from real users, not the initial seeded accounts. The comments, which initially consisted entirely of referral code sharing and “friend for friend” requests, are slowly starting to resemble TikTok and YouTube. I found one user with more followers on Zynn than TikTok. This account registered on May 15th and now re-posts some videos a few hours after TikTok. Raina Huang, a TikTok / YouTube / Instagram influencer who eats food, launched her account on May 22nd. She re-posts videos from TikTok, and occasionally sees similar engagement on Zynn.

Live streaming will likely be a high priority for Zynn. Looking at the picture above, “Say something” looks much more like a comment on a live stream, not an uploaded video. Kuaishou disclosed over half of its 400 million registered users had live streamed at least once as of May, 2017, and that 10% of streamers have sold goods. Most TikTok power creators don’t have that skill set, and its arguably still open for the taking. One things for sure: distribution (or content?) is king, and China is willing to pay for it.

The Power of Friends

Paying for content and distribution starts to look much more like traditional media than social media. Products with friend graphs and true user-generated content have structural competitive advantages vs those that don’t. One of the reasons Facebook and Snap are such valuable businesses is because they can place ads beside the content their user base creates for free.

Over the last 4 quarters, Facebook generated an average of $186 in revenue per US Facebook app DAU. It paid all-in costs (expenses and capital expenditures) of $28 per global “Family of Apps” DAU. Facebook’s disclosures with the SEC make it difficult to understand its costs by market, but assuming they’re even around the world, its operating margins are roughly 85% in the US. That’s very good. We see this with Snap as well: roughly 70% of each $1 in new revenue it generated over the past two years has converted to cash on its balance sheet (or Snap’s case, less cash was burned).

Starting with Facebook’s numbers, let’s assume Zynn spends an extra $20 per user for the initial referral bonus. This increases the total cost to $48 per user, or an operating margin of 74%. Kuaishou will likely keep paying bonuses to users to retain them, at least for now, which will also eat at its margins.

It doesn’t seem to have a strong recommendation algorithm, which hints at a weak targeted advertising product. It also generates most of its existing revenue (in China) through live streaming, which means it will likely take a 50% cut of “tips” and pay out a majority of the live streamed ecommerce GMV it generates to creators. Using Facebook’s $186 US ARPU as an anchor and assuming the average marketer targets a 3x ROAS (Return on Ad Spend), Facebook facilitated roughly $558 in commerce per US user in 2019. Kuaishou had over 100 million daily live stream viewers in 2019, and Zynn’s lack of editing tools I described earlier could hint at a long-term focus on live streaming. It disclosed over 19 million users had earned income live streaming on Kuaishou through September of 2019, up from 10 million at the same point in 2018. An announcement of its 2019 funding round also mentioned strategic initiatives in gaming.

Zynn / Kuaishou’s US operating margins may look closer to 30-40% at scale. That’s still impressive and worth paying up for today, provided it can retain its users. They’ll need to get them generating content to boost their margins, which may have already started happening, and could be accelerated if they pay influencers to post, like TikTok.

This ties back into my original point: Broadcast-based, or one-to-many social networks, are starting to look much more like traditional media companies that pay for content and distribution. Jeff Bezos famously says “Your margin is my opportunity”, and digital advertising companies are among the highest margin businesses of all-time. Snap cloned Facebook’s self-serve advertising platform in four years, and TikTok will likely do it even faster. Why wouldn’t their competitors give up margin to build similarly massive businesses? Maybe instead of thinking of these platforms as winner take all, we should think of them as TV Networks 2.0; the next iterations of Comcast/NBC, Disney/ABC, CBS, Viacom?

The large platforms like Facebook, Instagram, YouTube, Amazon, and TikTok will continue to get larger; but newcomers will be able to beat their distribution advantages with capital efficiency and superior product decisions. Products that rely on friends and true user-generated content will continue to have the strongest competitive advantages and highest margins over time.

Links I’m Reading

Snap will let other companies build versions of their apps within Snapchat. This appears to be an extension of its SnapKit developer platform, and may look similar to WeChat’s popular mini-programs. Considering Snapchat is used by over 88 million North Americans for over 30 minutes per day, there will be lots of opportunities to build on top of. Its no secret Snap sees its future as a camera-based computing platform. Snap Kit, the Snap Map, AR Lenses, and now mini-apps are strategic pillars that will converge as Snapchat eventually launches a phone, AR glasses, or another piece of hardware that isn’t on anyone’s radar yet.

Microsoft took the lid off its Fluid Office Framework. Its an open source, web-based framework that embeds Office into other products with real-time collaboration. Its launching on and Outlook. Teams integration will come next, and non-Microsoft products soon after. Microsoft has been building this for 18 months with a team of approximately 140 employees. This should increase the stickiness of Microsoft’s bundle and likely creates many opportunities for other products to build on top of it. It will also bring up strategic questions that other companies building in the space will need to answer.

Instagram to start showing ads in IGTV. 55% of revenue will be shared with creators. IG also announced users will be able to purchase badges that appear next to their comments during a particular accounts live stream. This feels like a long-time coming, and puts even more pressure on TikTok and other platforms to launch creator monetization.

I write about the intersection of business and technology from my perspective as a venture capitalist. If you’d like this in your inbox twice per month, please subscribe! If you’d like to read what I’m thinking about more often than that, follow me on Twitter at @TurnerNovak.

The Rise of TikTok and Understanding Its Parent Company ByteDance

In November of 2017, Bytedance acquired for $1 billion. In August of 2018, Bytedance officially rebranded and TikTok into one app.

Shortly after, ByteDance raised $3 billion and was valued at ~$75 billion. Analysts estimate its 2019 revenue was anywhere from $16-20 billion, up 100% from 2018. Its market share of the Chinese digital ad market has grown from zero to 22%, up from only 5% in 2017.

TikTok has become the best way to create and consume short videos on mobile. It rode the wave of AirPods and audio memes to over 1 billion DAU’s (Daily Active Users), and is likely worth ~$200 billion. This would make it not only the world’s most valuable "startup” but one of the world’s most valuable companies, period.

Introducing ByteDance and Toutiao

ByteDance is arguably the most serious competitor to Google, ever. It reaches consumers in the two largest middle classes in the world, China and the US, at a scale larger than any other digital advertising company. Like its Chinese counterparts, its seen success with monetization models beyond just advertising, which US-based competitors have struggled with to-date.

The company was founded by CEO Zhang Yiming, the son of civil workers in the southeastern city of Longyan. He studied microelectronics and software engineering at Nankai University. ByteDance’s first product Jinri Toutiao (“Today’s Headlines”) launched in August of 2012. Its best described as the Facebook news feed with no content from friends that serves hyper targeted content and ads. It peaked in mid-2018 at around 200 million DAU’s using it for an average 74 minutes per day; nearly twice as long as Facebook, Instagram, and Snapchat.

According to reports, Zhang had the idea in 2008 when he was still running the social site with Wang Xing (who founded Meituan). He was using his computer to book a train home and set up a system that sent him a text when a ticket became available. After Hainei, Zhang founded multiple other companies. He was CTO of online travel booking provider Kuxan, CEO of Twitter-clone Fanfou, and CEO of 99Fang, a real estate listing portal that reached 1.5 million users. He also launched a collaborative enterprise software service in 2003 while still in college and spent a short-time at Microsoft in China in between his entrepreneurial endeavors.

When Toutiao launched, China’s mobile news market was dominated by censored state controlled media and web portals like Sina and Sohu that served lengthy, boring text that was written and optimized for desktop. Toutiao’s initial product crawled content from around the long-tail of Chinese web-based media and reformatted it for mobile. For a time, it went as far as removing the original ads and replacing them with its own.

Toutiao initially seeded the app through Weibo influencers (the Twitter of China). It used aggressive push notifications and frequently nudged users to share content (more later), which helped it grow to 10 million users within 90 days. New users logged in with their Sina or Weibo account, which Toutiao scraped for initial interests and friends. It then used each user's individual usage data (how they tapped, swiped, or paused, time spent per article, their comments, location, time of day, and much more) to serve each user the most relevant content. It changed the titles, cover images, and even shortened most articles. This got users to an 80% read rate on each article in less than one day, which contributed to the 45% lifetime user retention it boasted early on. This was all instant and free compared to human editors and gave Toutiao a 10x better product at a 10x lower cost.

Traditional media companies of course hated all of this and Toutiao was in constant legal disputes. It claimed it was hard to attribute original sources of any particular story and that many sites crashed the app due to bad formatting on mobile. Toutiao eventually let users bounce to outside sites, and Toutiao often made up the majority of publishers' traffic. In 2014 Weibo’s traffic started declining, and it invested in Toutiao’s $100m Series C round in exchange for guaranteed traffic.

Toutiao soon convinced publishers and curators to create content directly in the app in exchange for a revenue share. A majority of the read-time on Toutiao quickly shifted to partner contributed content. Toutiao’s targeting gave its contributors a larger, more relevant audience, much faster than any other platform. This further incentivized contributors - and it had 1.2 million by 2017. State-controlled entities embraced Toutiao because, unlike any other services like Weibo or WeChat, they did not need followers to distribute public statements or updates on health epidemics. Tencent, Alibaba, Baidu and many startups eventually launched similar news products, but Toutiao’s high-margin business model (its core business was reportedly profitable in 2015 with $220 million in revenue) helped partners monetize their content and kept most contributing directly to Toutiao.

Over time Toutiao added relevant comments from friends (made on other platforms) to articles, the ability to share pictures to a feed, job listings, buttons to launch into fitness, music, and podcast apps; and live streamed shows, interactive Q&A channels, and a platform for full-length movies (Xigua). It followed other Chinese tech giants and introduced mini programs in 2018, which allowed developers to build app-like experiences within Toutiao. This introduced third-party delivery for restaurants, groceries, and pharmacies. Last, Toutiao had a culture of quickly shipping new apps with names like "good-looking pictures", "connotation comics", “funny embarrassment”, “must-see videos tonight”, and "getting pregnant when you laugh too much" and shutting them down within two months if they didn’t take off.

Most importantly, Toutiao incorporated video all over the app. Similar to Facebook, this allowed it to start inserting video feed ads. Toutiao initially went after big brands just starting to invest in mobile ads. Despite being much smaller than competitors Baidu and Tencent, brands were convinced to try Toutiao’s hyper-targeted ad product that reached its wealthy user base in Tier 1 and 2 cities. The targeting worked so well that data-driven marketing companies like Meituan were spending 85% of their ad budgets on ByteDance properties by July of 2019.

Toutiao’s reach with Chinese consumers and ability to hyper target video ads propelled ByteDance to an estimated $16-20 billion in revenue by 2019. This was more than twice Google in its eighth year of existence (2005).

In September of 2016, ByteDance used its core recommendation algorithm and the distribution it had built up with Chinese users and advertisers to launch short-video app Douyin in China (originally called, and TikTok to the rest of the world in 2017 (important to note these are two completely separate products). Not only did this expand Toutiao’s video ad inventory, but also created an avenue for introducing commerce and expanded outside its largely-male user base. Most importantly, it became evident by 2018 that Toutiao had reached maturity and ByteDance’s future growth would need to come from new products.

Zhang, ByteDance’s founder and CEO, has stated his primary strategy is to eliminate the need for search - how Google and Amazon serve their very profitable advertising products - and immediately serve users exactly what they want. All digital ad products have evolved to serve recommendations in some form or another; however their core DNA’s rely on products originally designed for desktop, not mobile.

TikTok: YouTube Built for Mobile

Its hard to argue with TikTok’s success. It disclosed it had over 1.5 billion MAU’s (Monthly Active Users) in June of 2019 and TikTok’s US leader suggested it was surpassing 1 billion DAU’s in October of 2019. Besides a ruthlessly executed growth strategy that I’ll explain later, it rode a few specific UI decisions to become a global phenomenon.

Video-First UI

TikTok’s key feature is a video-first interface, the “For You Page”, that takes up the entire screen and starts playing immediately, sucking users into the app. There’s a one-screen onboarding at the very first open, with no login required as it creates shadow profiles based on device ID. Many users don’t actually have accounts. Every video loops by default, and users swipe up and down to navigate the endless stream of videos, sideways to jump into a creator’s profile for more content, or tap one of multiple small buttons around the edge of the screen.

Just like Toutiao, TikTok uses app behavior to quickly build each user a data profile. Other digital media companies serve content using algorithms, but their mobile products carry design debt from feeds built for desktop. Large portions of the valuable, limited screen space on mobile is wasted. Here’s a comparison of three different core use cases: Do you think someone is more likely to watch a video or engage with an ad on WeChat (16% of screen), Instagram (31%), or TikTok (100%)?

The cross-cultural nature of UGC video may also have stronger network effects than traditional news or text-based social media. Purely entertaining bite-sized content may stay relevant to the TikTok algorithm for years. And a silly video uploaded in Thailand is potentially interesting to someone in Romania and the US.

Short-Form Video

Content on TikTok is very short - originally 15 seconds, and now up to 120 on Douyin in China. Short video is optimized for mobile by default, and naturally leads to a very different product DNA than long video.

Short-form reduces the friction of both creation and consumption. Most TikTok videos are produced by the creator alone, and many post multiple videos per day. A ByteDance advertiser deck from 2018 claims 34% of US users shoot content daily. Longer content on other platforms is typically not shot or edited on mobile, and creators often employ large teams around them.

Editing content themselves also puts TikTok creators much closer to the action. They must constantly consume the product themselves and understand the nuance of the community. An assumption could be made that the younger, mobile-first creators on TikTok are more motivated and innovative than their older predecessors on YouTube.

From a consumption perspective, an experience of entirely short content has less commitment each time a user opens the app. Each dopamine hit happens fast, and many TikTok’s climax at the very end (the algorithm favors watch-length). All videos replay by default, which often leads to multiple re-watches (which also has heavy weighting in the algorithm). This trains each users algorithm very fast: in the time it takes to watch a 10-minute YouTube video, TikTok can capture data from 40, 15-second videos. Training users to actively sit through full videos (that play with sound by default) makes them more receptive to ads, and TikTok’s bite-sized content will provide frequent, natural ad-breaks as its US advertising business ramps up.

YouTube, TikTok’s closest competitor, is optimized for large screens and simply not competing on the same playing field. It just signed its biggest star, PewDiePie, to a Live Streaming deal - and over 100 million people watched YouTube on a TV in March.

Snapchat Stories evolved from the constraint of sending picture and video messages to friends with the camera, and worked perfectly built on top of its core messaging use case with very strong network effects. When following Snapchat, Instagram failed to capture the greater opportunity by copying the format exactly and bolting it at the top of the feed, even though it did not have the same messaging constraint. Instagram placing this artificial constraint on creation, opening to the feed by default, and its emphasis on a follower graph, all capped the potential of any full-screen video product it launched.

Anti-Social Social Media

TikTok is the largest social media company that has little reliance on a social graph.

Facebook had a growth mantra of “7 friends in 10 days” that made the product sticky for new users, but TikTok doesn’t require any friends, followers, or even an account. The hyper-personalized algorithm recommends content based on thousands of objects and tags analyzed in each individual video, along with an individual user's view history, re-watches, likes, comments, shares, and even post-view activity. a16z partner Connie Chen called TikTok the first mainstream consumer app where artificial intelligence IS the product. This runs in stark contrast to the algorithms of legacy social products based largely around interactions from other users.

This anti-social approach may have more staying power than traditional models that suffer from context collapse, where the value of the product eventually goes down as a user adds too many friends. By default, TikTok separates young users from their parents as the user base matures. The switch to non-chronological feeds allowed Facebook, Instagram, and Twitter to smoothly transition into boosted and sponsored posts, and TikTok will have no chronological component whatsoever from the first moment it turns on self-serve programmatic advertising.

TikTok’s Expansion Strategy

TikTok used a very deliberate expansion strategy in each new market:

To start, it created the best tool to edit short-videos on mobile. This included easy clipping and scrubbing, a variety of filters, and a large library of music. Initially, many creators used it to post on other platforms. TikTok forces videos to be posted in order to export them, and also inserts a watermark on every exported video. Influencers on Weibo, Instagram, YouTube, and Twitter all drove traffic back to Douyin and TikTok, whether they realized it or not. Many creators who initially used it as an editing tool would deliberately start posting more when they noticed fans started following and interacting with them on the app.

Additionally, it made creators first-class citizens, giving them full-service support that included weekly content suggestion emails, 1-on-1 demos with TikTok’s staff, and live in-person events to create a community and encourage collaboration (an important aspect explained below). They even sent phone stands to keep their phones steady while recording.

It went after each market's biggest social influencers, and quickly built them social capital on the app. In the early days, everyone on its operations teams was tasked with roping in influencers (and long-rumored to have found ways to pay them to post). ByteDance partnered with (paid) celebrities to seed Douyin with content in China. One early partnership was with Baidu-backed streaming service iQIYI, where Douyin curated audition videos for China’s first hip-hop talent show, growing DAU’s from 290k to 1.73m between April and July of 2017. A partnership on a Michael Kors catwalk event in mid-2017 led to another spurt of user growth. Its speculated both Douyin and TikTok ripped and re-uploaded popular videos from other platforms to seed themselves with interesting content (Kwai is likely guilty as well), however some believe these were independent creators growing an audience. Finally, buying for ~$1 billion allowed ByteDance to enter North America and Europe by acquiring all its users.

Deliberate product decisions compounded this process. Extremely short videos, algorithmic discovery on the FYP, the muted importance of a follower graph, and constant traffic into the app via heavy marketing, cross-platform posting, promotion by influencers, and (eventually) habit allowed TikTok to deliberately control the allocation of social capital across the network to its most talented creators, both new or old, as new users poured into the app. Alex Zhu, founder and now Head of Product at TikTok, likens the process to creating a new country and giving a greenfield of opportunities to a new class of creators. Hyper fast user onboarding and no friend graph let it use the entirety of time spent in-app allocating this social capital. Casey Newton described this perfectly:

"In previous eras, most of the spoils went to the platform’s earliest adopters - mining value gets harder as the platform ages. TikTok, on the other hand, promotes all content regardless of who made it or how many followers [social capital] they have."

In business, superior distribution (reach with customers, users, etc) usually wins in the long run. A common way to beat a competitor with a distribution advantage is to build a better product that eventually surpasses their distribution. Despite fewer users, the product decisions above gave new TikTok creators instant distribution advantages over existing platforms. This was similar to when Instagram added Stories and gave creators instant distribution advantages over Snapchat. Creators with ~100k followers on Snapchat eventually switched to posting the same content for their ~1 million followers on Instagram. On TikTok, the best videos could instantly reach the entire user base.

TikTok also localized the content and creation tools for each market. Eventually, content in markets got so good that local influencer agencies began frequently scouting on TikTok, which created a flywheel as the best platform for creators. TikTok also allowed comments to be read in the bottom half of the screen while videos played. This encouraged more audience interaction with creators, participation with each other, and lots of inside jokes that spurred future content creation.

TikTok also aggressively promoted sharing of content. After two loops of a video, the share icon lights up (green icon on the left), which pops up an internal and external share page that takes up half the screen. TikTok is optimized for mobile web, and users get sucked into more videos they receive from a friend even if they don’t have the app. These off-platform shares between friends allowed it to start building both data profiles and a friend graph for any future social products based on device ID.

Finally, ByteDance flooded the world with marketing. It reportedly spent $3M/day on user acquisition and PR throughout 2018 and 2019, beating short-video competitors in each market by simply outspending them. TikTok spent over $300m on Google ads in 2018; and $10m/month in India alone. In Q1 of 2019, 13% of all ads seen by users of Facebook’s Android app were for TikTok (Apptopia). At the peak of the ad blitz in September of 2018, 22% of all ads seen on Facebook’s properties by Apple devices in the US were for TikTok. The watermark on videos posted to other platforms also fed this loop.

At first, TikTok’s 30-day retention of new users in the US was as low as 10%. This was likely due to targeting such a broad audience with content still very similar to and Vine. Retention improved over time, and has hovered between 28-40% since early 2019 per App Annie. Even as early users churned, the massive flood of traffic gave TikTok views and social capital to allocate to new creators that were just joining the platform.

ByteDance gets lots of criticism for spending so much on ads, but it's worth resurfacing Elad Gil’s 2010 reminder that sometimes it's worth spending money to kickstart a valuable network. Facebook used a similar strategy with Instagram by favoring it in the feed, which helped it quickly convert new users. As seen below, it's evident 50-75% of Instagram downloads ultimately came from Facebook in the years following the acquisition.

ByteDance didn’t have Facebook, but it did have over $7B in high margin advertising revenue from one of the fastest growing businesses of all time. This allowed it to raise an additional $3B from SoftBank’s Vision Fund that let it play by similar rules. In China specifically, it consistently dealt with dominant social network WeChat blocking access to Douyin videos shared by users. By necessity, cross-promotion in its own apps, aggressive sharing, and heavy marketing was part of its DNA.

As the growth of legacy platforms plateau, TikTok’s exploding user base and time spent in-app allows it to continue strategically allocating social capital. TikTok has clearly eaten into Instagram’s downloads...

… and the “monthly reach” in the US shown in Instagram’s ad platform was FLAT in the US between January of 2019 and April of 2020, the most valuable market in the world, during a time most of the population was stuck inside.

Investors and marketers haven’t been too spooked by Facebook and Instagram’s user struggles yet, but they will. App Annie data shows US users spend more time in aggregate on TikTok than Instagram (with a user base half the size), and reported time spent by Chinese users the first week of February was up 130% from the weekly average in 2019. ComScore reports total time spent on TikTok in the US increased 93% between October of 2019 and March of 2020, and 55% within the app specifically (this shows how frequently TikTok videos are shared and viewed outside the app). While ComScore usually under reports metrics on mobile, they’re directionally correct and show that total time spent on TikTok by US users has nearly doubled in the last six months.

TikTok’s downloads also accelerated significantly in Q1 of 2020 for its best quarter ever. This was presumably with much less ad spending, proving its strategy of smothering the world in marketing to kickstart the network worked.

A combination of many forces above also allowed TikTok to leapfrog YouTube in social relevance. VidCon 2019, traditionally focused on YouTube influencers, was dominated by TikTok stars. “The older generation doesn’t realize how important TikTok is yet… more people came here for TikTok people than YouTube. Big YouTubers get ahead of themselves. They think they’re too good to be around their lower fans.” I believe the biggest factor to this disconnect is YouTube’s product decisions that position it for larger screens, and specifically its lack of mobile editing tools.

Music and Audio Memes as a Wedge to Create YouTube 2.0

Social networks generally evolve from a powerful initial content creation tool, or entirely new content format, that reduces the friction of creation and consumption in a way that accumulates social capital for the creators. If done right, it builds a flywheel that compounds on itself as the network grows. TikTok’s flywheel is certainly spinning.

Video as a format has existed for a century, and is arguably the most intimate storytelling medium invented by humans. Despite the enormous success of YouTube and its perceived position of strength, the end-product (horizontal video) is built for desktop and TV - and most YouTube creators don’t edit their videos on mobile. There was always a hole in the market for a mobile-first (vertical), bite-sized video product. This was made apparent by the success of Snapchat and Instagram Stories.

Memes, defined as iterations of shared cultural beliefs or behaviors, have evolved as the default way many young people communicate - and even consume news. Typically, the more iterations to a meme, the more momentum it gets. Full subreddit’s have emerged that predict the velocity of new memes. Instagram disclosed memes are shared 7x more often than non-meme posts. Until recently, memes were mostly images with iterations of text; they were created with a keyboard and mouse.

TikTok rode two trends to dominate the past two years: memes and AirPods. Initially,’s lip syncing use case still dominated the app post re-brand, and most new users, who saw it as a joke, rebelled. By mid/late 2018, the app felt very similar to Vine: most videos were ironic jokes. In a stark contrast to the Millennial perfection of Instagram, irony become the language of Gen Z, and TikTok was built for it.

Vine users often remade popular videos, and TikTok lip syncing tools made it simple to remake videos using the same sound. TikTok’s UI lets users search for other videos with the same audio - something no other platform allows (Vine did briefly before shutting down). As TikTok was ramping up in the US, AirPods were proliferating. Apple sold ~16 million in 2017, and 35m, 60m, and 100m were shipped in each of the next three years. There was a captive audience of early AirPods adopters ready to consume memes that required audio at any time of the day. Audio memes were born. One of the first mass market audio memes on TikTok was the “Someone Like You” meme, with many others before it. A recent favorite is the Home Depot meme.

Sticking with Musically’s roots in music, many of the top memes on TikTok were built on songs. “Old Town Road” by Lil Nas X was the most successful of them all, taking over TikTok for months and breaking the record for the most time at #1 on the Billboard Hot 100 for 17 weeks. Similar to memes, most of the most successful musicians rely on remixes, sampling other songs, and features from other artists.

The focus on memes as a distribution strategy isn’t surprising. One of ByteDance’s first products was the joke / meme app Neihan Duanzi. It’s been taken down and re-launched numerous times, with the most recent iteration named after one of China’s biggest memes in 2017.

TikTok quickly laddered from lip syncing, to memes, to many other types of content: Comedy, how to’s, art, etc. Expanding the content broadens the user base; it also expands the interest graph it can build on users. This is crucial as it starts targeting users with other ByteDance products to go from challenging YouTube to taking on its parent company Google.

Leveraging TikTok to Launch New Products

Just like in China, ByteDance will leverage the AI capabilities, content, and user base of TikTok to launch other products around the world. Its pace of employee hiring reflects this ambitious strategy.

This fits Elad Gil’s “Product to Distribution” philosophy: Build the best product on the market, grow the user base aggressively to create a direct relationship with the network of users, and then use that network to launch new products. Facebook is a classic example of this with both Instagram and Messenger. Google has done the same with Search, YouTube, Email, etc; and compounded that with Android.

In TikTok’s case, layering on more content types over the past two years and expanding the scope of its interest graph is just the start. We should think of TikTok as not only a data collection and ad delivery vehicle, but also a top of funnel to push users towards other products.

Longer-Form Video

Starting in mid-2019, Douyin began allowing some creators to upload videos up to 15-minutes long. YouTube’s biggest stars were initially seeded by Vine, and some are now coming from TikTok. It feels inevitable TikTok launches products that prevent this leakage.

ByteDance operates Xigua Video (“Watermelon Video” in English) in China. Its a Netflix-like product for watching full-length movies used for an average of 70 minutes per day by 55 million DAU’s. Netflix’s ability to buy content it knows users will watch is a huge advantage, and its likely ByteDance uses TikTok behavior to nudge users into similar long-form content. Xigua recently started licensing content from BBC, PBS, and kids publisher Moonbug. Most recently, ByteDance hired the ex-Head of Disney+ as its new COO / CEO of TikTok, who grew Disney’s new subscription service to over 50 million users in less than a year. He also led many of Disney’s acquisitions over the past decade, including Marvel, Lucasfilm, Pixar, 21st Century Fox, Club Penguin, and Maker Studios. ByteDance is positioned to do a lot in long-form video over the next few years.

Music Streaming

Like Spotify playlists, TikTok memes drive which songs hit the top of the charts. Some songs now release on TikTok before streaming services.

ByteDance recently launched its own music streaming app, Resso, in India, Indonesia, and Brazil. It marries the UI’s of TikTok and Spotify with the spirit of YouTube. The app borrows TikTok’s swiping UI and emphasizes commenting, creating, and sharing content within the app - its also possible users will be able to post directly to TikTok. Lyrics are prominent - which is extremely important as it helps TikTokers understand the songs / memes as they get stuck in their heads. It's not far-fetched to think TikTok knows exactly what music its users like, nor that they could tap into Resso and listen to a song they just heard seven times in the TikTok feed.

Analysts project China’s karaoke / live streaming industry will hit $19b by 2023, which made up 72% ($2.6b) of Tencent Music’s revenue in 2019. ByteDance’s music product may incorporate live karaoke, growth loops within TikTok, and undercut competitors like Spotify, Apple, WeSing, and Kugou on price as it launches around the world. Resso also gives ByteDance better control of  point-of-stream data and monetization, and it could tie this in with TikTok performance.


ByteDance has been very vocal about entering gaming. It has bought multiple game publishers over the past two years and its gaming division now has over 1k employees. Three of the top five mobile games in China during the 2020 Chinese Lunar New Year were from ByteDance. Its first launch in Japan spent nearly a week at the top of the App Store.

Building up a large social property to distribute games rhymes with Tencent’s strategy. Consumers spent $86b globally on mobile games in 2019. And ByteDance’s users evidently play games - 63 of the top 100 Chinese mobile gaming ad spenders in 2019 reportedly spent over half their marketing budgets on Toutiao. It has an opportunity to vertically integrate and eat the enterprise value of its largest gaming advertisers. You could argue ByteDance could go a step further than Tencent’s in-app purchases model to incorporate in-game ads with its existing self-serve ad network, or even open up a third-party ad network to other game publishers.

Consumer Finance

Many Chinese tech companies see fintech as the holy grail of monetization, ByteDance appears no different. It’s rolled out various forms of consumer and business loans, insurance products, and wealth management products to over 23 million users. It’s still early, but ByteDance will likely leverage its distribution with billions of users around the world to experiment with things like payments and a mobile wallet.


Education is a massive priority for ByteDance. It recently announced plans to hire over 10k new employees for education initiatives, or ⅓ of all 2019 new hires. It's unclear exactly what success here will look like, but the latest rumors include a hardware product for institutions, an AI-based tutor, a tutoring portal, and a consumer product for paid-courses. None of its historical pushes into education have caught on. Gogokid, which matched Chinese students with English teachers abroad, was an early foray into education but struggled to gain traction. Chen Lin, one of the earliest employees and ex-Head of Toutiao, was recently put in charge of “finding the next 100 million DAU product”, which is speculated to be in either education or messaging.


Tencent has a history of blocking competitive products from both WeChat and its Android app store, YingYongBao. This happened to ByteDance's Snapchat-like messaging app Duoshan after it hit 5 million downloads in one month. It's also experimenting with Flipchat, a Reddit-like feed-based group chat app that also facilitates things like payments, car-hailing, and food delivery.

Figuring out a messaging product has always appeared to be a high priority. I think ByteDance could see the most success in messaging through Douyin and TikTok, as 51% of Douyin users “share videos frequently”. TikTok has started nudging US users to message others in-app and Douyin has experimented with a Connection feature that matches strangers to play games together with selfie filters over a video call.

News Feeds

ByteDance will likely continue experimenting with feed-based news products that borrow from Toutiao to put pressure on Facebook and Twitter. This includes TopBuzz and News Republic in Western markets, BaBe in Indonesia, and a hybrid feed/TikTok-like product called Helo in India. It invested in Daily Hunt in India, and tried to buy Reddit in 2016. I doubt ByteDance sees much success here for awhile due to the competition, and would not be surprised if it ultimately gets into this space through a large acquisition.

Enterprise Software

In 2019, ByteDance launched an enterprise software product called Feishu (Lark in English). The team had over 1.7k employees at the beginning of 2020. It combines email, chat, video call, calendar and cloud document storage, borrowing from US-products like Slack, Microsoft Teams, and Google Suite.

Most large enterprise software in China is owned by the Chinese tech giants, and many companies build tools internally once they reach a certain size due to a fear of giving up access to their data. Lark was similarly first built for internal use. Surprisingly, ByteDance has stated they’re targeting the US, Europe, and Japan (not China) as Lark’s initial markets. Its key differentiators of translating all communication into the reader's native language, easy coordination across time zones, and expense reimbursement hint its targeting large enterprise customers that operate globally.

Cloud Hosting

ByteDance started opening data centers in India and the US in 2019. This may be strictly for internal purposes, but its possible they plan to launch a cloud hosting service. ByteDance acquired Terark in 2019 which allows databases to run 200x faster by reading compressed data at 50% of the cost. If they launch an external service, the selling point could be faster processing and access to some of ByteDance’s AI tech.

A cloud service combined with enterprise software (and eventually a phone or other hardware) would take a page out of Microsoft’s approach of bundling multiple products and selling them top down through the C-Suite.

Other Bets

Some other products we may see from ByteDance outside of China include live streaming, search, more traditional e-commerce, e-books, and other various social products. Search would boost ad inventory and build a Search + YouTube intent loop similar to Google. A web browser launched off links out of TikTok profiles could be an interesting product to capture more mobile time spent, as well as a podcast or audio product seeded by TikTok creators.


Last, ByteDance launched a low-end Android handset in China in 2019. It comes pre-loaded with its suite of apps and possibly the first iteration of a ByteDance “AI-first” operating system. On top of hardware revenue, this provides an additional distribution channel for all its apps. Its mini programs in China work on both Toutiao and Douyin, and it’s possible they’re also built to run on “ByteDance OS”.

What Will TikTok Do Next?

Circling back to TikTok, it likely adopts many of the features from its Chineses counterpart Douyin over the next two years. The app is built around sourcing and elevating the next generation of creators, yet you could argue the minimized importance of a follower graph allowed lots of social capital to leak off TikTok. Long-term, we’ll see TikTok features that help creators make money while launching new products mentioned above that increase ByteDance’s touch points with consumers and their time spent on mobile.

One we’ll see soon is ecommerce integration in videos. Chinese creators on Douyin can create shoppable videos linked through Alibaba’s Taobao store. Shopify could be a partner in the US (like it just launched with Facebook). ByteDance may also build this internally embedded directly into TikTok videos, further driving its viral sharing loops. It appears to be testing this in China while taking only a 1% cut. We could expect early ecommerce use cases similar to this ironic reaction video or product demos like this:

TikTok recently rolled out donation stickers which will start training users to transact over the app: purchases and tipping creators. Live streamed commerce is very popular on Douyin, and ByteDance recently started pushing streams within TikTok. There are cultural differences in China that favor live streaming, but TikTok’s young user base is familiar with the format and the likeliest early adopters in North America and Europe.

TikTok wants brands to be first-class citizens in the feed. TikTok’s Creator Marketplace connects companies and content creators. Influencer marketing is a fast growing $10+ billion industry that’s largely coordinated over email and social media DMs and this will be a selling point as ByteDance works to onboard North America and Europe advertisers. Brands like Square are using it to promote products and artists and labels may even use it to promote music releases.

TikTok likely opens up its self-serve ad manager, which looks to perform on par with other programmatic channels. This would make it the only self-serve platform that hits users in both the US and China at scale.

TikTok may start providing brands and creators AI-generated content recommendations based on what’s performing well: what effects to use, changes to make to a video, or what content to produce all together. TikTok will also likely double down on AR camera effects (pushing an ad unit priced on video impressions), and may also open its camera and effects to developers.

Considering other products in the pipeline, we may eventually see a cross-product creator tool that spans ByteDance entire ecosystem: posting videos, music, ecommerce, podcasting, analytics, and even financing. ByteDance has already consolidated the back-ends in China, allowing creators to use Douyin as a distribution channel for multiple products.



In less than four years, TikTok became the most culturally significant product in the world. It rode the tailwinds of AirPods and memes to reach over 1 billion DAU’s and is now the best way to create and consume video content on mobile. Everyone should be watching to see what they do next.

If you’d like to watch with me, subscribe for more posts like this. I’m also on Twitter at @TurnerNovak.

Turner's Fantasy VC Portfolio II [Originally published May 9th, 2019]

I originally tweeted this on May 9th, 2019 before I set up this blog. Re-posting it here in an easy to read format.

I’ve gotten lucky since I put out my first fantasy portfolio in Aug ‘18 (link), with five of the seven companies I picked raising follow-on rounds since then, and two raising twice. For my second fantasy VC portfolio, I picked 17 companies accompanied with a one-page thesis for each.

I stayed as early as I could (less than $10M previously raised) as that’s what I’m most interested in. I prioritized positive long-term social impact, companies with unique business models, clear revenue prospects, opportunities for defensible distribution channels, and hints of a large market with incumbents that are un-incentivized to react.

My sourcing process is a combination of friends, Twitter, Reddit, other social channels, proactively helping founders and investors, some Product Hunt-ing, a little Hacker News, and following too many blogs, newsletters, and podcasts.

My due diligence is admittedly held back doing this remotely, in my free time, with no budget, and almost entirely blind to metrics. I dig things up on Google, Reddit, and YouTube, listen to podcasts, read blogs/reviews, scan social, and ask people smarter than me too many questions.

I've loved applying my background in credit, PE, and as an allocator to venture, and know I'm just scratching the surface with lots still to learn. Appreciate any and all feedback on this!


Stage: Series A / Total Raised: $12.6M
HQ: Los Angeles / Underrepresented Founder: No

Ordermark created a system to consolidate orders and hardware from all food delivery providers (Uber Eats, Postmates, DoorDash, etc) into one PoS tablet and order ticket printer. Born from a fourth-generation restaurateur who added 14 delivery providers all at once, saw it boosted revenue but increased complexity, and built an in-house solution to manage all the inbound orders. There’s a lot you can do with delivery order data, and Ordermark is in an interesting spot seeing all the orders from every provider. I also believe food delivery companies are simply last mile delivery providers that used food to build out their networks, and will continue expanding into other categories. Ordermark could be positioned to consolidate last mile delivery solutions for more than just restaurants, or simply become the restaurant PoS system for both in-house and takeout, and use that wedge to become a focused, vertically integrated restaurant SaaS solution.

The incentives seem well-aligned as it helps restaurants generate incremental revenue by tapping into unused kitchen capacity by plugging in multiple delivery providers, while making it easier for delivery providers add extra restaurants options to their networks. Also plays into the rise of Cloud Kitchens - kitchen-only restaurants designed for delivery-only. It feels like the large chains that haven’t utilized much delivery to-date will be some of the big long-term winners in the proliferation of food delivery, and Ordermark already has Subway, Sonic, Little Caesars, Buffalo Wild Wings, Popeye's, and Papa John's (as a few examples) as customers. I can’t find any competing products that are as well-designed or with as many big customers on their sites.

The CEO Alex Canter grew up in the restaurant industry, was previously COO of a food truck ordering app, and had launched various startups as a student at the University of Wisconsin. Uber Eats and DoorDash have experimented with their own hardware, which could make this a tough business over the long-term if they move further into restaurant management and Ordermark is unable to provide more value for customers. On-demand delivery networks will significantly change the restaurant industry over the next decade, and I’d love to hear how the founders are thinking about how they fit into the space. Specifically, are they working with franchisees, or chainwide? How quickly do they see adoption spread chainwide?


Stage: Series A / Total Raised: $9.8M
HQ: Paris / Underrepresented Founder: Yes

WeMaintain is a platform that connects elevator mechanics and property managers and landlords. 60% of the world's elevators are located in 30 cities, and there are four major elevator operators who hold all the customer relationships and power over the elevator technicians and property managers and landlords in the US and Europe. WeMaintain allows technicians to connect with clients over its mobile platform instead of through the elevator operators, letting them take charge of their own schedule, compensation, and which jobs they fulfill. For property managers, WeMaintain provides price transparency and a faster, digital-first service, of which the elevator cartel has had no incentive to do due to the oligopoly structure of the market.

The global elevator market is expected to reach $130B by 2025. Installation growth will be driven by urbanization in emerging markets. EMEA (41%) and China (33%) make up the bulk of the current global install base, however China is adding roughly 63% of global supply each year. Though smaller, the US and Europe represent the largest service markets, are the most mature / consolidated, and represent WeMaintain’s immediate opportunity. The US elevator servicing and installation market will reach $21B in 2019, with growth to be driven by higher safety requirements and interest in energy efficiency / cost savings. The elevator industry is beginning to adopt technologies like IoT to diagnose and pre-empt maintenance, something WeMaintain is positioned to take advantage of.

Prior to co-founding the company, the CEO Benoit Dupont finished an MBA from Harvard in 2016, and spent the past 12 years in progressive roles at Otis Elevator (one of the elevator majors); four years in France, eight years in China. Here's his (literal) elevator pitch. Seems like there’s eventually an opportunity to do more than just elevator maintenance once it builds up its customer and technician base, and I’d love to talk to the founders about the long-term strategy. I'd also want to know what sort of contracts the existing elevator companies have with clients, because while maintenance is likely higher margin and more attractive than manufacturing, I could see there being service exclusivity contracts with the big players. China and other emerging markets will largely drive installs in the immediate term, but will become large servicing opportunities as those new units age and will likely become key growth markets around the time of a hypothetical IPO. I’ve been following WeMaintain for almost a year, they just raised a Series A as I was putting this together, and think they're uniquely positioned for a big opportunity.


Stage: Series A / Total Raised: $9.2M
HQ: San Francisco / Underrepresented Founder: No

Darkstore operates a network of third party urban fulfillment centers that enables 1-hour shipping for brands (5 min pitch from the CEO in 2016). Faster shipping reduces the # of orders that fail to convert / die in cart, which boosts sales for e-comm partners. 1-hour shipping has interesting implications with exclusive drops, like the Air Jordan III shoe drop they partnered on with Nike and Snapchat in 2018. Darkstore also allows any storage facility, warehouse, courier depot, physical store, mall, etc. to monetize unused space by becoming a Darkstore fulfillment center, which appears to be largely automated and can be fulfilled by workers via mobile. It also partners with last mile delivery partners like Uber and Postmates, which are integrated with Darkstore's product. This gives Darkstore a capital-light business model by not needing to build its own fulfillment centers or operate a last mile delivery network.

Darkstore says its all-in fulfillment fee for retailers is near 3% of the retail price (compared to what it says is a 25% average for e-commerce brands), sharing a cut with fulfillment partners, also giving additional revenue to delivery partners. The business model seems to be a win-win-win for customers, real estate owners with excess capacity, and delivery providers. I could see the current product scaling up as-is, but Darkstore could also build tech to allow customers to bring Darkstore’s automated fulfillment solutions in-house to their own space.

The founder Lee Hnetinka previously founded Wunwun, a 1-hour delivery provider for restaurants and retailers, which was acquired by HelloAlfred in 2015. Looks like he’s hired some ex-Amazon logistics / operations folks and others with others with same-day delivery expertise. Biggest question I'd have is, since they don't have massive warehouses, how do they figure out how to stock inventory at each fulfillment center? Do they rely on their customers shipping them the right mix of inventory? Do they have enough density in each urban market to allow them to stock different inventory types in strategic locations around each city? I’d also love to hear a little more on how the math works with pricing so low, but could see the model working since they aren’t spending much on leasing or in-house delivery.


Stage: Series A / Total Raised: $8M
HQ: London / Underrepresented Founder: Yes

Scape built a cloud-based, city-scale localization engine (their “Vision Engine”) using only cameras and no LIDAR. By mapping on the server, not on the device, they've taken a very different approach from 6D, Placenote, Jido, etc, avoiding CPU, memory, and storage limits. Server-side also allows them to build a single city-scale map, with every camera querying and adding to the same map in the cloud. This means everything is geo-positioned, allows them to localize in rain, wind, snow, etc, and their product is essentially a hyper-accurate GPS.

The first product is a camera API for city-scale AR placement, with plans to apply its Vision Engine API in self-driving cars, delivery robots, autonomous drones, etc. It seems as if most non-stealth companies working in the space have targeted smaller-scale localization (on-device, not on-server) that allows them to add AR geometry to the environment, making them better-suited for room-scale games. Scape focused on city-wide localization from the get-go (beta is live in 100 cities) and is more suited for things related to autonomous navigation and Pokemon Go-type games.

The founders have complimentary histories in autonomous robotics and VR/AR/360 video, which gave a unique skillset to build both the tech and developer-facing ecosystem. Computer vision is capital intensive and they will likely need to raise multiple rounds to scale up. I’d want more clarity on the product roadmap and how they specifically see themselves staying competitive against others in the space who may have more capital and look to implement similar camera-based localization.


Stage: Series A / Total Raised: $7.1M
HQ: San Francisco / Underrepresented Founder: Yes

Farmstead operates a network of farm-to-fridge grocery stores built from the ground up optimized for delivery. Farmstead takes an AI-first approach by combining historical sales, current trend data, consumer recommendations, holidays and product sell-by dates to predict customer demand and stock just the right amount at each warehouse. Intuitively a delivery-first grocer could have lower costs, faster speed, etc, compared to partnering with physical stores delivering out of buildings and inventory / employees base, etc that are not optimized for delivery. Farmstead gives discounts/free delivery when customers pool deliveries with neighbors, which reduces delivery costs and minimizes the carbon footprint (will eventually benefit from EV's and AV's, of which it is already partnering with Udelv). Global food supply chains are extremely inefficient and wasteful, and Farmstead reduces food waste (8% compared to industry average of 35%) and allows for fresher local food. This combination of 1) delivery-first infrastructure, 2) pooled delivery routes, and 3) less food waste gives it a lower cost structure compared to traditional grocers.

Farmstead was growing 20%-30% MoM as of December, and nearing contribution margin positive. With groceries as a multiple-times-per-week order for most families, they could use it to wedge into other household products, clothes, etc to become a delivery-first retailer. I also believe consumer preference will favor retailers and manufacturers that adopt closed loop / reusable / zero-waste packaging solutions, and Farmstead appears well-positioned to do this with the frequent, direct-to-household delivery channels it’s developing.

The co-founders met at Yahoo, CEO Pradeep Elankumaran spent time as Head of Driver Growth at Lyft, and is a 2x YC grad. I like groceries as a wedge into becoming a larger retailer, and they have a delivery-first approach that I think other grocers will have trouble competing with. Biggest thing I don't understand is how they order from local suppliers, and how well that will scale as they expand. One of the articles I linked notes their six-month retention is "over 50%", so I wonder how that’s changed over time and if they see it improving as they add more categories.


Stage: Seed / Total Raised: $6.7M
HQ: Los Angeles / Underrepresented Founder: No

Postie is bringing digital ad targeting and machine learning capabilities to the $46 billion direct mail industry. Postie’s data management platform integrates with most CRM’s, websites, and mobile apps, and uses the same A/B testing, custom and lookalike audience modeling, retargeting, and performance attribution digital marketers have become accustomed to. Postie also designed a printing system that lets advertisers create and ship out individually customized direct mail campaigns at scale within 24 hours, much quicker than legacy processes that take weeks. Postie has been profitable from day one, reaches up to 320 million consumers, and early customers are seeing ROI as strong as top performing digital campaigns.

Direct mail is a $46B industry that is just now starting to adopt digital best practices. It’s quantitative, highly optimizable, and directly measurable, but has not changed much in decades. Physical mail elicits more emotion in consumers and generates high-intent conversions. Response rates across the industry are up 60% over the past decade, with 20-30 year olds as the highest converting age group. Publishers have learned the importance acquisition channel diversification, and recent Facebook Ad Manager outages have caught digital marketers flat footed. The combination of plateauing user growth and rising CPM’s on the large digital platforms will lead savvy marketers to continue looking for emerging, high ROI advertising channels.

The team is distributed across the US, and the two founders previously worked on D2C marketing campaigns for Dollar Shave Club and numerous other brands while executives at venture studio Science Inc. Postie’s customers seem to be concentrated in e-commerce, and it may be vulnerable to a pullback in the early stage consumer funding environment or a broader recession. While direct mail alone is a large market, I wonder if the core tech will allow Postie to add other channels and become more sticky? It does not appear that they have much proprietary data, and I’m wondering how proprietary their printing system is? Direct mail would be an interesting add-on to Facebook’s Ad Manager, but I’m not sure the opportunity would be large enough for Facebook to take seriously until an acquisition makes more sense.


Stage: Series A / Total Raised: $6.3M
HQ: San Francisco / Underrepresented Founder: Yes

Anomalie designs, manufactures, and sells custom wedding gowns directly to consumers. It has a team of stylists that use a data-first approach to work directly with brides to design a dress to fit their unique style and budget. It’s product is a faction of the cost and has less lead time than traditional bridal gowns.The dress is both the central piece of modern wedding ceremonies and a key pain point for most brides, and Anomalie offers an affordable, custom solution for brides who’s style or body types don’t fit the narrowly defined box of the industry. Eventually, Anomalie could build out its own network of boutiques and design workshops, offer pre-made, customizable designs, and potentially adds-on other pieces of the wedding process that flow from the design of the wedding gown. There’s also an opportunity to use their tech and manufacturing process to create other categories of custom garments.

I’ve grown up in the wedding industry, raised by a single mother who designed and hand-sewed hundreds of custom wedding gowns out of our home, and (unsuccessfully) launched a similar concept in 2012. The US wedding market is over $100B+, and the US wedding gown industry was over $8B in 2018 (roughly $34B globally). The industry is largely brick and mortar, very fragmented and localized (with the largest player going through bankruptcy), and pricing is very opaque across the industry. Most gowns are bought off-the-rack, however most brides seek certain aspects of customization and boutiques often charge for post-production alterations that Anomalie largely eliminates. Most of the world’s silk, and over 80% of wedding gowns, are made in Suzhou, China. By going straight to the source, Anomalie is able to undercut boutiques that have low inventories and no manufacturing expertise. Most brides have the vision in their head, but can’t find it in person.

Anomalie was founded by couple Calley Means (ex-Strategy at Zenefits) and Leslie Vorhees, who met while doing their MBA’s at Harvard. Leslie applied her work experience managing factories in Asia for Apple and Nike to source, design, and manufacture her own wedding gown (similar to wedding gowns, both Nike shoes and the Apple Watch she oversaw production of were handmade). Friends asked her to do the same for them, and the company was born. They are attracting top talent, recently hiring the ex-Head of Production for David’s Bridal (oversaw 35% of US wedding dress production), data scientists from Stich Fix and AirBnB, and operations managers from Google and Goldman Sachs. I believe Anomalie’s full-stack approach will be much more difficult to copy than other D2C products because it’s not a generic item that can be drop-shipped from hundreds of different Chinese manufacturers. While Anomalie’s product is very differentiated, it’s ultimately competing on cost as more sales increases Anomalie’s leverage with workshops to re-design and optimize their manufacturing process, allows them to negotiate lower material costs, gives them more data, and ultimately increase virality and allow them to lower the price (which both increase demand).

Titan School Solutions

Stage: Series A / Total Raised: $5.2M
HQ: Los Angeles / Underrepresented Founder: No
Website / App Store

Titan developed a browser-based HTML5 PoS system for school lunches that works on any hardware. The product includes a mobile app for students with menus and nutrition information, and allows parents to add funds to accounts and see their child's purchase history. Lunch is a daily use case for every single student, and it hits many different parts of the back-end infrastructure of a school district (accounting/finance, compliance, inventory, purchasing, etc). Seems like they’re positioned to use what they built for school lunches as a wedge into becoming a school ERP system that bridges both families and districts.

30.4M students participated in the National School Lunch Program in 2016, which spends over $15B to serve 5B+ meals per year in over 100k k-12 schools. Titan serves 1M students in 150 districts (primarily in AZ and CA) as of December of 2018, up from 350k students in 70 districts as of May of 2018. Buying food for a school district requires dozens of non-standardized processes, learning state and federal regulations and reporting requirements, and sorting through product catalogs and annual trade shows. Also requires menu planning, contract negotiations with food distributors, running / hiring staff for daily cafeteria operations. Budget size significantly impacts how many resources, how much purchasing power, and what sort of supplier access a school district has. Titan simplifies this process and offers districts real-time reporting, with an an opportunity provide an inventory management platform / ordering marketplace at scale, to eventually transition to a district-wide ERP solution. Titan’s website mentions a 100% retention rate.

CEO Brad Blankenship previously worked at Harris School Solutions, a subsidiary of Constellation Software (the Berkshire Hathaway of software). Looks like he brought over some of his senior management team from Harris based on scanning LinkedIn. Harris seems to be one of their big competitors, and would love to hear Brad’s perspective on the industry. It looks like Titan goes a step further than Harris’ products by, simplifying the process for students and parents, speeding lunch lines, and integrating inventory / supply chain process for districts. Titan’s reviews have an “it just works!” reception from districts, and it smells like the “ex-employee coming over from the broken, unwilling to fix it” incumbent.


Stage: Seed / Total Raised: $3.6M
HQ: London / Underrepresented Founder: Yes

Floom is building an operating system for florists. It’s bringing creativity and design back into the floral process, minimizing the environmental impact caused by global competitors, while still offering the same day delivery consumers expect. The company was built upon the thesis that flowers are an emotional (and most universally accepted) gift, and existing competitors are built around centralized, bland global product catalogs that treat florists as order stuffers. The company started as a marketplace for local, independent florists selling unique, seasonal bouquets. Floom is using its relationships with florists to create a full-stack SaaS solution for florists that includes everything from local order and inventory management, online web management, seamless checkout, an in-store PoS system, and delivery. This allows florists to better manage their time, cut costs, and provide better experiences for customers.

The US floral gifting market is projected to reach $16B by 2023, and is north of $100B globally. Growth will largely be driven by a growing consumer preference for local, custom, and more expensive flowers. Local flowers are not only typically one-off designs hand-prepared by independent florists, but also have more sustainable supply chains. Global competitors are run as centralized hubs with generic pictures that push out orders for local networks to fill, and they are structurally unprepared for the creativity and self-expression the Instagram Age has brought back into the floral industry. Currently, local florists in the US are a very fragmented, $7B industry, and must do weddings and large events to generate high enough margins. There’s an opportunity for a platform to not only simplify the process of running an independent floral boutique, but also facilitate the relationship between local suppliers, creatives, consumers, and wholesale local floral bouquets to supermarket shelves.

Founder Lana Elie had zero background in the industry before founding Floom, and came up with the idea after many years of trying to order unique floral bouquets during her time in content, design, and marketing at Vice and Burberry. She realized the incumbents model, which required seasonal flowers to be held in stock year-round by every order partner and treated them as order stuffers without a consumer-facing brand, was structurally unable to compete with a local-first model. Lana realized there was an opportunity to empower florists to order their own inventory from suppliers, use that inventory to create their own stock, and then created a dashboard for them to manage their own product catalog. This not only gives florists more control of their business, but also allows Floom’s customers to sell products the global chains are unable to distribute.

For Days

Stage: Seed / Total Raised: $2.8M
HQ: Los Angeles / Underrepresented Founder: Yes

For Days is a circular manufacturing and commerce platform. It launched as an annual subscription product (a capital efficient, upfront cash infusion) that allows customers to order plain cotton t-shirts and return them once worn-down directly to For Days’ to be de-threaded and re-spun into new products. For Days utilized a waitlist incorporating Robinhood-like pre-launch referrals and social engagement to drive demand. The company designed a zero-waste manufacturing system in LA built on renewable energy and water reclamation to support that initial product, and is now opening up its capabilities to other companies. For Days will need to generate significant scale to drive down costs and make a closed loop manufacturing process affordable for consumers, and opening its platform to the industry gives it an opportunity to scale the business faster and onboard more brands to the circular economy.

The global apparel, shoewear, and jewellery industries are roughly $2T, and have long been considered large contributors to global pollution. Regardless of the true impact the industry has on the environment, consumers are increasingly demanding more sustainable solutions. Consumers have become conscious of the environmental impacts of the industry, and have incorporated second-hand and thrifting into their consumption habits. Even still, the majority of donated clothing ends up in landfills. As awareness of the waste currently produced by the apparel industry continues to increase, I believe closed loop manufacturing will eventually become the norm. This will largely be made possible by the rise of direct to consumer e-commerce channels and last mile delivery networks. It’s possible that closed loop systems are inherently slower than the fast fashion model that has driven the industry in recent years, however I believe a closed loop re-threading process will eventually allow for customized products that can be recycled and re-customized for the next customer.

Co-founder Kristy Caylor has nearly two decades of experience in the industry, and previously founded Maiyet, one of the first sustainable fashion brands. Prior, she worked in progressive roles at the GAP, and founded an online membership-based reservation system for LA restaurants in 2000. Co-founder Mary Saunders previously launched Maiyet with Kristy, and prior to that worked in progressive merchandising positions at GAP after completing her MBA at Harvard. They’ve brought on early employees with impressive textile backgrounds, including a Chief Manufacturing Officer with nearly four decades of experience in all aspects of textile operations, most recently running three dyeing and finishing factories for American Apparel. It’s hard to gauge exactly how For Days cotton de-and-re-threading process works, but I’d want to learn more about how it can be utilized on other materials and in other categories outside just apparel. It seems like the founders have brought on the right talent, but haven’t talked much about global expansion, which I think will be important as most industry growth going forward will come outside the developed markets where they’re currently based.


Stage: Seed / Total Raised: $1.8M
HQ: Cork, Ireland / Underrepresented Founder: Yes

ApisProtect retrofits bee hives with data sensors and provides beekeepers real-time, mobile monitoring to better-manage their colonies. The sensors create a black box of previously non-existent hive data (temperature, humidity, CO2, sound, and movement data) and provide actionable messages, such as "this hive is sick" or "no queen", for each hive, allowing beekeepers to proactively respond to problems as they arise. This allows beekeepers to monitor their hives during times they can’t check manually (in the night, in poor weather, in the winter), while also removing manual, periodic hive checks from their work flow, which take up most of their time and are a large cause of disease, pests and other issues that deteriorate hive health and contribute to hive loss.

Crop pollination adds an estimated ~$175B in "economic value"  globally per year. The US Beekeeping Industry is estimated to be $327M in 2018. This is largely driven by the almond industry as nearly 73% of US colonies are used for almond pollination during the summer, contributing ~$313M of revenue in 2018. Colony health is important when negotiating terms with almond growers, and ApisProtect provides beekeepers with insight and added value. Beekepers lose anywhere from 30-50% of their hives every year, yet bees are responsible for pollinating most fruits and vegetables in the global agriculture supply chain. The total number of managed honey bee hives is up 45% over the last 50 years, but supply has not kept up with demand as there has also been a 300% increase in agriculture production that relies on pollination. Long-term consumer preferences moving towards local, natural, and sustainable beauty products and food sources (and eating more than just corn and soybeans, which don’t need to be pollinated), plus the adoption of indoor farming, could lead to significant expansion of the pollination and honey markets over the next few decades. I’d expect network effects as colonies are on-boarded, logging more hive activity to monitor patterns both regionally and globally. Benefits could extend into other forms of insect management, and there may be an opportunity to create an IoT platform that spans the entire agriculture industry.

Founder Fiona Murphy studied Electrical Engineering in undergrad and completed a PhD in Embedded Systems (IoT) in 2013, both at UCC in Ireland. Her doctoral research on the application of sensors and networking in honeybee hives initially attracted the beekeeping community, which is how she got involved in the industry. There are 91 million managed beehives worldwide, and ApisProtect currently monitors 144, with plans to increase to 200 next year, and 300k hives over the next five years. They’re manually installing themselves today, I’m assuming they’ll need to figure out at-home installs to ramp up to 300k. They raised a $1.8M Seed round in November (Atlantic Bridge + some ag-focused VC’s) to open an office in Salinas, CA, considered the center of the industry due to the local almond industry’s reliance on bee pollination.


Stage: Seed / Total Raised: $1.6M
HQ: San Francisco / Underrepresented Founder: No

Keeper is a monthly subscription product that automatically finds tax write-offs for independent contractors (real estate agents, freelancers, scooter chargers, online sellers, rideshare drivers, etc). Uses Plaid to pull spending info, sends SMS texts each day to track expenses throughout the year, and gives an export at tax-time. Plans on a full-suite tax filing solution for the 2019 filing season next year.

The US tax prep industry will be ~$11B in 2019, and accounting services in the US and globally were $120B and $507B in 2018. I think there’s an opportunity for a monthly subscription tax product that’s incentivized to give customers high quality year-round service, whereas incumbents focus on confusing consumers and charging high lump sums upon filing. The average user reports $2,073/year in write-offs ($173/mo compared to the $10/mo price) they would have missed without Keeper. I could see Keeper evolving into a full-suite SMB accounting and tax provider, a new-age Intuit.

I like their GTM which seems to be partnering with and offering discounts to workers of other on-demand startups. One of the co-founders (Paul Koullick) has a background in data science at Square, created the tax/expense tracking app for Stride (health benefits platform for independent contractors), and studied Applied Math / CS at Harvard. Other co-founder (David Kang) went to U of Chicago and spent the past five years as a options trader at a hedge fund in Chicago. Feels like a team that could build this, but would want to learn more about how they’re thinking about the space and how they’ll localize the product if they decide to scale globally. They just raised $1.6M in April, and I’d assume they want to staff up for tax season if they see good traction over the next few months.


Stage: Seed / Total Raised: $1.2M
HQ: Boulder / Underrepresented Founder: No

Zestful provides a physical debit card that allows companies of all sizes to adopt employee perk and alternative compensation programs. It allows for hundreds of options (food delivery, subscriptions, ride hailing, events and experiences, fitness, and education and charities) that can be customized by each employer at the employee level. Zestful makes a compelling employee perk program economical for small businesses, and enterprise clients benefit from real-time monitoring and elimination of cumbersome 10-week reimbursement processes via Zestful’s debit card. Zestful was born from the insight that some employees value a lower salary with unique perks because it has a higher impact on their lives. Zestful provides an interesting value proposition as it allows its customers to promote employee happiness while also being more capital efficient by lowering their all-in compensation costs.

Employee recognition is a $5.8B industry, projected to grow at a 10-year CAGR of 15.7% to $29B by 2028. Roughly 89% of companies have employee recognition programs, and nearly all have some sort of recognition, reward, discount, or bonus program. Zestful falls within multiple mega-trends, as younger demographics are trading salaries for unique benefits and experiences, and companies are realizing the value of employee happiness. The customization of consumer products is spilling into the enterprise, and the rise of decentralized work will not only allow employers to reduce CapEx spend on in-house perks like cafeterias and fitness centers, but present a need for the modularized perk program that Zestful provides. There are legacy providers in the space, however most are not positioned for the future of work and have participation rates of around 20% compared to the 94% of Zestful users who are active every month. There are other corporate charge card providers, however they are largely optimized for enterprise, not employee perks. Zestful charges a flat fee per card, plus a small % fee per use.

The founder Mat Vogels initially took the company through YC in 2017 and is a YC and Techstars mentor. He has a background in finance, design, and product marketing that shines through in all aspects of the product, from Zestful’s digital presence to the physical debit cards. Mat has recruited a team that includes CTO Malte Muenke, who co-created GoToMypc, GoToMeeting, and GoToWebiner, one of the original suites of enterprise SaaS products. Zestful’s bright colored debit card attracts free, organic attention at every use and may lead to structurally low long-term customer acquisition costs. The back-end is built to on-board a long-tail of local business in under 24 hours, which means more perks, more clients, and more data to eventually recommend customers situational out-of-the-box perk packages. Managing a debit-card facilitated employee perk program also presents an opportunity to eventually move into other aspects of employee engagement, human resources, or financial services.


Stage: Pre-Seed / Total Raised: N/A
HQ: NYC / Underrepresented Founder: No

Insycle is a tool to bulk-manage, sync, and update data across multiple SaaS platforms like Salesforce, Hubspot, Zendesk, Intercom, Yext, etc. Enterprise software is projected to continuing proliferating, and Insycle is intriguing because it lays on top of its customers SaaS platforms. I also think an inevitable increase in mobile, voice-first, and other data input methods (contractors taking photos of invoices; transcribing customer service calls, etc.) will increase data inconsistencies across different products, especially in the short/medium-term. Insycle currently uses a manual process to bulk-approve changes, and I think there’s an opportunity for AI to dramatically speed up the process over time.

The founder Yonatan Lee is originally from Israel, and spent over a decade building wall street trading software (Bank of America, Citigroup). He’s recruited a small team, just wrapped up the Acceleprise accelerator in NYC, and has been very hands on with customers planning out the product road map.


Stage: Angel / Total Raised: N/A
HQ: Hamburg, Germany / Underrepresented Founder: No

Standups is a Snapchat-stories inspired take, 60-120 second video updates, on standup meetings for distributed work forces that span multiple time zones (“asynchronous work”). I came across Standups in August of 2018 from this post on Hacker News (he had just launched at the time), followed by their Product Hunt launch, where JP the founder does a good job responding to feedback. He’s been a developer on distributed, asynchronous teams for the majority of his career, and understands exactly what asynchronous teams need.

The use of distributed/remote teams has continued gaining momentum over the past decade (Stripe just made a big announcement here) because it opens up a global talent pool and has evident cost savings. I believe efficient communication will be a core component of distributed teams. Owning the daily update, a crucial piece of the remote-enterprise stack, could give an opportunity to expand into other aspects of asynchronous/remote enterprise work. I also think the remote workforce trend is still new enough that there’s room for products specifically designed for it. JP doesn't see his product limited to distributed teams, he thinks he's found a need for all teams that require constant updates, whether they're all in one room or spread out over the world. He also thinks there’s an opportunity to apply data science to provide much more than daily standup video updates.


Stage: Angel / Total Raised: N/A
HQ: Seattle / Underrepresented Founder: No

Weel is building autonomous bikes and scooters that deliver themselves.


Stage: Angel / Total Raised: N/A
HQ: NYC / Underrepresented Founder: Yes

Eternal is an avatar-first digital identity company. Their first product is the social game Icon.

Loading more posts…