JOKR and Personalized Instant Commerce

The inside story behind one of the world's fastest growing and most customer-centric companies

In April of 2021, Ralf Wenzel and his team launched JOKR, a global platform for instant grocery and retail delivery at a hyper local scale. Within six months, the company operated in seven countries and its annualized revenue run-rate surpassed $100 million, showing no signs of slowing down.

I was lucky to invest in JOKR’s Seed and Series A rounds through my investment firm, Banana Capital. Here’s the inside story behind one of the world’s fastest growing and most customer-centric companies.

Spotting Operational Inefficiencies; The Beginning of Ralf’s Ecommerce Journey

Ralf was born in East-Berlin and lived 10 meters from the Berlin Wall. He had family in Cuba and spent part of his childhood there, getting a glimpse of Latin America’s version of communism. Living between both systems as the Berlin Wall came down taught him to spot operational inefficiencies and inspired him to fix them.

He saw the internet coming and studied computer science at the University of Havana in the mid-90’s, betting his career on it. Over the next decade, he worked in engineering at Mercedes Benz, received a Masters in Computer Science at Berlin University of Technology and Economics, started numerous businesses, and even invented a voice-recognition engine.

The first startup Ralf joined was Berlin-based Jamba (Jamster) in 2004. The product distributed news, wallpapers, ringtones, and eventually games on mobile devices (which at the time had tiny black and white screens with text pads). Ralf led Sales and Business Development, expanding Jamba into 45 countries, becoming the world’s largest platform for mobile content. Jamba was acquired by VeriSign and NewsCorp in 2005 and was one of Europe’s first big tech exits, putting Europe’s tech scene on the map. Not only was this Ralf’s first taste of building a company, but it was also where he first met one of JOKR’s co-founders and now-CPTO, Sven Grajetzki.

Two Decades of Company Building

After Jamba, part of the team moved to London to work on Moneybookers together, which later rebranded to Skrill and merged with Paysafe in 2015. Ralf stayed on for five years as COO, working closely with Sven in Product. Skrill, and then Paysafe went on to IPO, becoming one of the biggest payment companies in the world and the largest e-wallet behind PayPal and Alipay at the time. Ralf then launched a Seed-stage venture fund, Tocororo Ventures, to invest in tech and food startups around the world.

By 2013, many of JOKR’s eventual co-founders (Ralf, Sven, + new additions Ben Bauer and Konstantin “Koko” Sorger from Groupon) moved to Singapore to launch and scale foodpanda globally, which was at the time only in South East Asia. The company was initially incubated by Germany’s Rocket Internet and Ralf eventually became foodpanda’s founder and global CEO, with Ben (CCO), Sven (CPO), and Koko (CMO) all taking progressive roles as the company scaled. This was around the time Ralf met JOKR’s now-COO Aspa Lekka. Originally from Greece, she spent seven years playing professional handball across Europe. Ralf convinced her to join foodpanda for four years and then return after her MBA. foodpanda went on to launch ghost kitchens, new brands in each market, make over 12 acquisitions, and capture high market share in over 40 countries across South East Asia, MENA, Eastern Europe, Russia, and LatAm.

Over this time, Ralf became friends with Niklas Oestberg who co-founded Delivery Hero in 2011. Similar to foodpanda, Delivery Hero had captured high market share throughout Western Europe. Initially competitors and both based in Berlin, they combined forces in 2016, citing complementary geographic coverage and a shared investor (Rocket Internet, who had been buying up stakes in numerous on-demand companies around the world at that time). Niklas would focus on all operations as CEO, Ralf on M&A as Chief Strategy Officer, with other future members of JOKR’s founding team taking roles across the globe.

The Delivery Hero / foodpanda merger closed on December 31st, 2016 and the company IPO’d six month later in June 2017. Today, the public markets value Delivery Hero at ~$27 billion. foodpanda’s original markets make up the majority of Delivery Hero’s orders (2.5 billion run rate as of Q2 2021), and over half of its GMV ($33.43 billion) and revenue ($5.78 billion). Delivery Hero continued acquiring companies, making over 26 acquisitions and 22 outside investments that included Rappi in Latin America, Zomato in India, Deliveroo in Europe, and Woowa Bros in Korea (acquired).

Then, Softbank called.

SoftBank, WeWork, and Lockdowns

In 2019, SoftBank recruited Ralf as a Managing Partner to lead its Latin American Tech Hub, an incubator for startups, joint ventures, and strategic initiatives across the region. Fascinated by the opportunity to launch 50 businesses in five years and work alongside SoftBank’s new $5 billion Latin America Fund, the team moved to Latin America.

Part of the role involved managing operations at numerous SoftBank portfolio companies. Market historians will remind us that 2019 and 2020 were challenging years for some SoftBank-backed companies. Spending most of their time on WeWork and Oyo, the SoftBank tech Hub was never realized. Then in March 2020, COVID hit.

Locked down around the world, Ralf reconnected with all the best operators he’d worked with over the past 20 years. As they ordered everything online (meals, groceries, kitchen supplies, furniture) they realized every purchase in each market all had an inconsistent customer experience. Poor product discovery, slow delivery, inadequate customer service, and even the wrong products being delivered all together. Over time, the team that had been building on-demand businesses since 2011 realized there was still a huge gap between customer expectations and what the market was delivering.

The Massive, Outdated Grocery Market

This market they were entering admittedly seems crowded at first glance. It consists of four major categories: Brick & Mortar, Supermarket Delivery, Marketplaces, and Instant Commerce. A non-exhaustive overview:

JOKR specifically operates in the instant commerce space, broadly defined as a service that delivers to customers in 15-minute or less. Within that, JOKR launched with a low-SKU collection of grocery products. Grocery retail is one of the largest categories globally at $9.8 trillion, or roughly ~10% of global GDP, however less than 3% (~$260 billion) of that is currently online.

There are many reasons for groceries' low online penetration, and the simplest answer is that existing offline solutions have generally been good enough for consumers. Things like agriculture and roads first built in the ancient era, food processing and global supply chains established during the industrial revolution, and densely trafficked roads and city plans that evolved around automobiles, all shaped how the global food and retail infrastructure evolved into multiple layers of distributors, consumer packaged goods (CPG) conglomerates, and national chains of physical brick & mortar stores by the 1990’s.

Many of these brick & mortar stores still exist, however are structurally unprofitable selling online as-is and were generally un-incentivized to figure out how to do so until 2020. They built businesses centered around daily two-way commutes from their customers. Their operations, business models, and cost structures are optimized for that in-person, in-store experience. Trips to the grocery store are generally not enjoyable and can take a total of 45 to 90 minute for some consumers. The inconveniences associated with traditional grocery shopping has encouraged users to take fewer trips (once a week on average). This has forced consumers to attempt to plan meals far in advance and make large orders to stock their fridge for weeks in advance, leading to significant time spent on meal planning and high food waste.

Most grocery stores have now begun to offer some form of their own in-house delivery, some of which are profitable. These profitable models generally have higher minimum order requirements, markups on items, non-flexible delivery windows (often multiple deliveries are pooled into one driver trip to save costs), and an unclear view into what products are actually in-stock due to outdated software stacks.

With the rise of mobile phones, a second major category emerged: three-sided marketplaces connecting couriers, local stores, and consumers. Customers order from a local store on the platform, and typically schedule a drop off time anywhere from 2-hours to one week in advance. The platforms dispatch a courier (non-salaried, on-demand worker) who drives to the retailer, walks around the store to locate each item in the order, works through the checkout line, delivers it to the customer, and repeats this multiple times at multiple stores spread around a city until they choose to stop. These marketplaces represented the first wave of mass-market online grocery adoption as they connected consumers with excess capacity at existing retail locations. Their asset-light models allowed them to scale very quickly, and their on-demand nature made them attractive for couriers in a competitive labor market as fast adoption of smartphones made it easy to earn an income.

As these marketplace models scaled, it became clear that a middle, software-layer taking an additional cut between the demand and an already complex supply side was inefficient. Brick & mortar stores traditionally have slim profit margins, as well as outdated software systems that make it difficult to connect their inventory with the marketplaces, providing an inconsistent customer experience (below). The marketplaces also saw limitations relying on existing infrastructure that was built around commuters, which led to longer delivery times and higher prices.

The item replacement process is core to most marketplace products

Long delivery times and high prices were exacerbated in 2020 by the COVID-19 pandemic, giving rise to the latest category in grocery: instant commerce. First proven out by goPuff in the US and Getir in Turkey, this model delivers groceries and other products to consumers in 10-15 minutes from micro fulfillment warehouses. These 2.5k-5k square foot hubs are not seen by the end-customer despite being located within minutes of them. The hubs have floor plans, shelves, packing, and logistics areas optimized for delivery. They pay rent, employ full-time workers, and carry their own inventory - the last of which can have compounding advantages at scale.

When the internet first hit in the 1990’s, companies like Webvan and Kozmo attempted to bring groceries and 15-minute delivery online. Both failed, ultimately due to poor pricing and customer segmentation, unnecessarily complex infrastructure, and simply expanding too fast without discipline around profitability. You could also argue a lack of smartphones for the delivery fleet was working against them. Most investors have since abandoned the category assuming the model doesn’t work, forgetting that Kozmo was profitable in four cities (and ironically raised money from Amazon in March 2000). Pink Dot, founded in 1987, has also operated profitably for over 30 years in Los Angeles, the first two decades operating mostly via phone orders and cash payments.

goPuff has since publicly disclosed its been EBITDA positive since day 1, with 40% gross margins and double-digit contribution margins. Its new markets are profitable within 18 months, and at scale its drivers deliver within 20 minutes and complete an average of 4 orders per hour (often via batching). goPuff also proved the model worked in US college towns, not just dense urban areas, which many believers in the instant commerce model even doubted was possible.

Variations of the model have taken China by storm, as nearly every large Chinese internet platform has at least attempted some form of entry into the market over the past two years. Samokat in Russia was acquired by Russian social network and internet platform mail.ru (assume operating margins similar to Facebook) in 2020, and is now launching in the US under its Buyk brand. Russian search-engine turned internet platform Yandex (reports EBITDA margins north of 40%) has also been investing heavily in its meal delivery and quick commerce brand, Lavka. In its public investor materials Yandex states the Total Addressable Market (TAM) of its Ecommerce and Egrocery business will 3x over the next three years, making it 10x larger than the opportunity it sees in digital advertising and 5x larger than in shared transportation and meal delivery. It recently started to expand beyond Russia.

Finally, public disclosures in Delivery Hero’s investor relations materials also hint the model is more profitable than its marketplace business.

Delivery Hero operates what is likely the largest business in this instant commerce space. Their public investor materials estimate the category will represent 25% of all online groceries and 5% of all other commerce globally by 2030. JOKR has already begun expanding outside groceries and into adjacent product categories to reach the broader $26.6 trillion global retail market, of which ~15% is currently online.

JOKR: Instant Commerce, Powered by Vertical Integration

To differentiate in what some consider a crowded market, the product behind JOKR is an end-to-end experience that takes control of every step of the ecommerce value chain. It combines speed, personalization, local assortment, and sustainability to provide an unparalleled customer experience and give back the one thing every consumer wants more of: time.

Consumers gravitate towards on-demand products because, if done correctly, they get what they need faster than it would take them to travel to it themselves (let alone locate, checkout, and travel back home). The best on-demand experiences drive customer affinity, and there is generally a correlation between delivery speed and customer retention. JOKR’s early metrics show exactly this as it retains customers at twice the rate of the scaled marketplace models.

This superior customer experience is why vertically integrated models will win over the long-term. Vertical integration generally has higher fixed costs and lower per unit costs at scale. In JOKR’s case, these cost savings can fall to the bottom line, be reinvested, passed to consumers in the form of lower prices and fees, or a combination of the above.

JOKR co-founder and CEO Ralf Wenzel at one of JOKR’s early hubs in Mexico

Where JOKR excels in speed of delivery, grocery marketplaces are constrained due to supply constraints (stores, restaurants, etc), which leads to a poor customer experience. The existence of these marketplaces made sense within the context of the market's evolution as they provided immediate scale and product variety, however the initial creation and placement of its supply side throughout cities pre-dated the existence of the internet.

The historical location of the supply that makes up most marketplaces is not aligned with the actual demand of an internet-first economy. JOKR’s smaller micro fulfillment centers (or hubs) strategically positioned and built for its delivery-first model, not in-person, brings that supply closer to the demand, allowing for even faster delivery and a better customer experience.

The team running one of JOKR’s Manhattan hubs

Vertically integrated models ultimately have the opportunity to rebuild the convoluted global supply chains that power much of the retail industry. Existing supply chains are very wasteful and a focus on local products not only removes layers of middleman to increase profits for suppliers, retailers, and consumers; it also provides a better customer experience (speed and quality) and a drastically more sustainable option for consumers (JOKR aims to be carbon neutral by 2022). Local products also lead to faster sourcing and inventory turnover. As JOKR scales its global platform, it will better leverage all its fixed costs and investment in building the underlying technology. And one day, it may have the opportunity to open up this more efficient supply chain to other market participants.

Data Science and Personalization; Taking Verticalization to the 21st Century

To execute this vertically integrated model, JOKR has to be very good at predicting exactly what consumers want. The team refers to this as the singular customer journey. Predicting what each individual customer needs in the morning, afternoon, and evening; on Monday, Thursday, and Saturday; in January, April, and November. JOKR needs to know the exact profile of each customer, in each location, which gets more granular and powerful at scale.

Some of JOKR’s NYC team

To accomplish this, JOKR uses a robust inventory procurement and city management system. It maps each city into 300x300 meter sections. Customers are then segmented by these neighborhoods to predict demand, which has historically been done at an aggregate, high-level globally by suppliers. JOKR predicts what inventory will be needed months, weeks, days, and even hours in advance to optimize the limited space in each hub. Instead of a retail model where suppliers say “what can we produce?”, it becomes JOKR saying “what will we sell?”.

JOKR builds hubs in each neighborhood, supported by a central warehouse in each region. It aims to capture the highest market square within each region, allowing each hub to be profitable on a neighborhood level. Its on-demand riders not only deliver to customers, they also rotate inventory between hubs multiple times per day as JOKR predicts a product may be out of stock in a particular location. This means JOKR only procures the products it needs, increases inventory turnover, maximizes operating leverage, and ensures the best customer experience. Most importantly, it's near impossible for legacy players to replicate within their existing operating models.

As JOKR scaled, it started expanding its share of wallet beyond grocery into adjacent categories like personal care, pharmacy, household goods, pets, apparel, and electronics. Soon a customer will be able to buy any product imaginable on JOKR and have it delivered within 15 minutes. This runs counter to what is possible within the existing infrastructure of incumbent players built for 2-hour and 2-day delivery.

The results of JOKR’s product expansion into 100,000+ SKU’s are already playing out. In September, the total basket size of the first cohort of customers in April was twice as large as that first month. This cohort is also placing nearly 50% more orders per month. Considering that like any new platform there has naturally been some churn, it shows how quickly JOKR’s product has become a habit for its customers.

If companies like Coupang in Korea and Sea Limited in South East Asia are any example, there are other opportunities beyond ecommerce for a platform like JOKR to expand into over time. Capturing a repetitive consumer behavior like ordering groceries allows them to layer on other products over time. This could include, but not be limited to, everything from consumer payments, meal delivery, travel, streaming, healthcare, gaming, financial services, and products for their merchants and suppliers. This expansion into other product lines is generally the holy grail of any consumer internet platform as highlighted in Postmates Series B deck.

Data science and personalization also builds the foundation for a very effective advertising product at scale. Profit margins in legacy grocery average 2-3%. One of the industry’s worst kept secrets is that trade spend, the amount CPG companies pay to promote their products with special placement at retailers, is 5-25% of the average supplier's revenue, representing up to $500 billion globally. This form of offline advertising makes up over 100% of profits in the traditional grocery industry. Historically, this channel has been incredibly inefficient relative to digital advertising due to lack of targeting and measurement capabilities. It is estimated that over $104 billion of trade spend dollars go to waste and up to 75% of offline promotions from CPG brands fail to break-even.

As the industry shifts online, advertising will do for grocery what Facebook did to the ad industry. Suppliers will have self-serve tools to pay for guaranteed demand. This will shift the industry from being supply-driven (“how do we sell what we produced?”) to being demand-driven (“what do we need to produce?”). With up to 35% wastage in a $10 trillion industry, this could cut excess production and trillions of dollars of food waste (and eventually other products) across the globe by only producing what will be consumed, while also being very profitable for JOKR.

A data-driven approach also sets JOKR up to launch higher margin private label products, a common strategy in retail. This combination of private label, high-margin ads, personalization, adjacent product categories, and vertical integration positions JOKR to become a very valuable business over time. Its already playing out: the first hubs launched in LatAm have positive contribution margins despite much lower basket sizes than its European and US markets launched later in 2021.

The Early Innings of Instant Commerce

JOKR is building a personalized, instant commerce platform for consumers across the world. Capturing consumer demand at the point of purchase positions them favorably in the value chain, and capturing repetitive grocery orders positions them to add adjacent products and capabilities over time.

JOKR’s founding team has scaled multiple global companies together and is firing on all cylinders, building one of the fastest growing consumer businesses of all-time. They are still early in their journey and hiring across all roles (here) for anyone ambitious enough to help bring 15 minute delivery to consumers across the world.

If you liked this deep dive on JOKR, please subscribe above for more writings on companies and trends I’m following and investing in. I’m also on Twitter at @TurnerNovak.

Thank you to Armin at Greycroft for help on market research and the team at JOKR for letting me be a part of the journey.

Due to the potential inaccuracy and translation errors in some foreign publications, I cannot personally verify every piece of information shared here. I have linked to all sources wherever possible. My thoughts shared above are not investment advice and I am an investor in JOKR through my investment firm, Banana Capital.

Snap Partner Summit 2021 Recap

Snap opens more of its platform to developers, growth of core Snapchat still accelerating

Snap hosted its 2021 Partner Summit (replay) last week. Here’s my recap and key takeaways, as is tradition.

Snapchat Growth Continues Accelerating

Possibly the biggest metric disclosed was Snapchat now reaches 500 million Monthly Active Users (MAU’s). With 280m Daily Active Users (DAU’s), ~56% of users check the app daily. The most successful messaging products like Snapchat are typically higher (Messenger’s DAU/MAU is ~69% in the US), however this lower number may reflect incremental MAU’s coming from Discover, Snapchat’s content product that reaches more US viewers than Netflix, and newly acquired international users (which skews lower). These folks may not be messaging daily, but their usage is highly monetizable. They’re also captive to be pulled into Snapchat’s messaging network over time.

Snapchat’s DAU’s increased 22.3% year-over-year (YoY) in Q1 ‘21, the fastest since its Q1 ‘17 IPO. This was despite dropping hints on earnings calls throughout 2020 that COVID-imposed lockdowns were holding back user growth. Half of all US smartphone users are now on Snapchat, and DAU’s are still opening the app 30x per day. India DAU’s are up 100% YoY for the fifth quarter in a row. Snap also mentioned that 40% of its users are now outside North America and Europe - a trend kickstarted in 2019 by the redesigned Android app.

Snap still has the challenge of monetizing these non-developed market users, something Facebook struggles with as well. Things like AR, premium Discover content, and commerce will likely help with this over time.

The Snapchat Camera & Augmented Reality: Powering Snap’s Business Model

Snap kicked off the summit explaining how cameras typically struggle to accurately capture darker skin tones. It’s commitment to fixing this may seem trivial at first, however considering most of Snap’s future user growth will come in markets with darker skin toned populations, it’s very relevant as explained here:

Snap’s camera usage is still strong, as 5 billion Snaps are created per day, or 18 per DAU. Over 200 million of Snapchat’s 280 million DAU’s (71%) engage with augmented reality (AR) daily. A recent Deloitte study found the pandemic made AR experiences more important to more than half of consumers, and over 75% expect their use of AR to increase over the next five years. In the past Snap has disclosed numbers suggesting users spend ~3 minutes per day in the camera; important considering Snap shared Scan, its camera search feature, now has 170 million MAU’s.

A big Scan capability unveiled last week was “Screenshop”. Point the camera at a friend and scans their outfit to see similar looks to be purchased. This also works with saved pictures in Memories, Snapchat’s cloud photo and video storage feature.

Scan is also getting camera shortcuts, which auto suggests pre-edits to save time when creating a snap, and the ability to recommend recipes based on food in a picture. These rolled out last week, with hints at many more Scan integrations to come. If I had to guess, placement in these visual search tools will be biddable in the future.

Snapchatters share content within the camera 600+ million times per month, and Snap hinted any piece of outside content will soon be shareable inside the camera (this can already be done with Games, Minis, etc). This should increase Snapchat camera usage and opens a new distribution channel for outside apps.

We also saw the next iteration of Local Lenses called Connected Lenses. These allow friends to share the same AR experience at the same time - across the room or across the world. This capability was also opened to lens creators in Lens Studio.

Lens Studio 4.0: Creators, Developers, and Partners

One announcement that seemed to sneak under the radar was Visual Classification in Lens Studio (Snap’s AR development platform), which brings the capabilities of Scan to all developers. I think of Snap’s AR lenses as very early iterations of AR-based apps, and Scan will be how many of these are accessed.

Lens Studio now has multi-person 3D Body Mesh and Cloth Simulation, as well as some new, no-code visual effects. This allows for more realistic, tighter than ever AR. Snap hinted at clothing as big use case, specifically in no code.

Snap also launched API-enabled lenses to pipe in digital assets and product inventory with no integrations needed. Browsable, public profiles for catalogs of AR lenses are also coming to Snapchat (storefronts??). In an AR try-on beta with 30 brands, Snapchatters tried products on 250+ million times, and were 2x more likely to purchase than in other campaigns. American Eagle’s virtual pop-up store generated $2m in incremental sales and one Gucci lens saw 19 million try-ons. All said, Snap continues dropping hints commerce will be a big driver of AR monetization.

Snap unveiled its Lens Creator Marketplace last week, a self-serve portal for creators and Snap’s ad partners. More analytics capabilities are coming to lenses (user demo insights, play time) that mirror its ad tools. This hints at a convergence, and will allow those creators to earn money (200k creators have created 2 million lenses that have been viewed 2 trillion times) while helping brands create AR experiences to be used throughout Snap.

In its try-on beta, Farfetch notably incorporated Lens Studio’s 40 voice commands into its catalog browsing. Snap ML, which allows for Machine Learning models within created lenses, now incorporates sound and audio to recognize phrases, transcribe speech, and make music. This evolution of Lens Studio’s voice capabilities is important as voice will be a key interaction layer in Snap’s AR glasses, Spectacles.

Spectacles 4: the First Built-in AR Displays

This years “one more thing” was Spectacles 4, Snap’s first pair of AR Glasses with a built-in AR display.

The new Specs are 134 grams, 7x lighter than the third generation and ~3x heavier than the average pair of sunglasses. 3D waveguide displays and 15 millisecond latency promise the most realistic AR overlays yet that should work seamlessly in and outdoors. The displays adjust automatically and auto-recommend lenses based on context (hello Scan!). The new Specs link directly to Lens Studio and can render projects in real-time. In addition to bringing back touchpad controls, they have two speakers and four microphones, again hinting how important voice and hands-free will be in AR.

Snap’s biggest advantage in an AR race that includes Apple, Facebook, Google, Microsoft (and many others) is its existing AR usage and friend graph within Snapchat. The new Specs allow you to send snaps directly from the glasses. And if Connected Lenses are any indicator, these will likely work cross platform powering cross-platform, co-located AR experiences.

Alex Heath at The Verge reported Snap acquired WaveOptics, the glass display supplier in Spectacles, for ~$500 million. This is Snap’s largest acquisition to date. Snap has been building Spectacles for seven years, and despite being roughly 10 years away from mass market, Snap is investing aggressively to own this next platform.

Extending the 30 minute battery life, bone conduction, and chip miniaturization (Snap will acquire or ultimately be acquired for this) will eventually increase the use cases and open up wider scale consumer adoption of AR glasses over time. And unlike VR, AR overlays on top of real life. Instead of isolating us from the world, AR promises to immerse us. And similar to Apple’s approach to mobile, the initial consumer use cases more likely translates to enterprise than vice versa.

Snap Kit: Snap’s Developer Platform

A platform that’s starting to see immediate results, Snap revealed its 2.5 year old developer platform Snap Kit now has 250k+ developers. Last week it released Sticker Kit, which opened 34+ million Bitmojis, Snap-made stickers, and GIFs to developers. Snap disclosed its creator tools are used 1+ billion times per day, or 3.6x daily per DAU. More importantly, it adds more features to Snap Kit, making it more compelling for developers to adopt. Pieces of Snap’s platform now help with initial conversion on installs, increasing engagement and retention, and monetization. And for Snap, it gets developers building on a platform that all appears to be converging and increasing in utility.

A few partnerships were highlighted, including a new sticker integration with existing Camera Kit partner, Bumble. This puts Snap’s products in Bumble’s camera and chat.

We also learned about what looks to be the beginning of many Disney partnerships. First are location-based AR lenses in Disneyland, including fixed permanence on iconic physical objects around Disney Parks.

Last year, Bitmoji was integrated into all Samsung phone keyboards. This year, Snap mentioned it’s Camera Kit product was coming to the native camera of all Samsung phones. This brings (all?) features of Snapchat’s camera (and ability to send snaps?) to 20% of all smartphones sold globally on an Android operating system it historicvally struggled on.

Bitmoji: Snapchat’s Decentralized Social Network

Snap casually mentioned Bitmoji now has 200 million DAU’s, another very under reported stat coming out of SPS. Aqcuired for $64 millioni in 2016, Bitmoji is a digital avatar tied to each Snapchat account. Snap has hinted at social features coming to Bitmoji, and we shouldn’t be surprised to see more parts of Bitmoji interlink between all the places its starting to show up. Bitmoji saw over 5 billion digital try-ons over the past year, hinting at what a decentralized advertising and virtual economy might look like.

Bitmoji for Games is also coming to Unity. After announcing the Snap Audience Network would plug into Unity’s supply in December 2020, this further proves how strategic of an asset Bitmoji has become. It’s a portable digital character that can be used across Snapchat, various consumer apps, and in mobile, PC, and console game environments, setting the stage for a decentralized Snap metaverse.

Games and Minis: Snap’s Ecosystem

Over 200 million Snapchatters have played with Games and Minis, up from 100 million one year ago. Snap Games now have 30 million MAU’s, and 30 total Snap Games have been published over past two years.

Games and Minis are built on the same platform and offer developers one engine that works cross-platform across iOS and Android. Snap is also turning on revenue generation via ads, tokens, and commerce.

French-based Voodoo’s Aqua Park has reached 45 million players since launching last fall and is now working on 5 more Snap Games. Snapchat is serving as an additional distribution channel for new app downloads + gives them incremental revenue as they can re-use assets they’ve already built for their own apps. If you’re not familiar with Voodoo, they’re a game publisher with a portfolio of “mini-game”-like apps. Voodoo also has a publishing platform that powers 100’s of other game publishers.

A standout early Snap Mini example is Poshmark. The Mini will feature daily shopping events, viewing Poshmark’s 200+ million secondhand item catalog, sharing with friends, and checkout - all within Snapchat. We’ll likely see similar concepts launch with other early test partners that include Target, Adidas, Ralph Lauren, and Dior. Gifting platform Givingli powered 2 million digital Valentine’s Day cards, and Turbo Vote helped 30 million users register to vote. With products like Bitmoji that can weave through these experiences both in and outside Snap, Minis will likely be a channel that increases revenue, installs, and retention for partners.

A big change coming to Snapchat’s UI is Games and Minis are getting their own home in the app. Swiping down from the camera will access what might look like, if you squint hard enough, the very early days of an Snap App Store.

Snap Map: Bridging Snapchat’s Friend Graph with the Real World

A product that ties together Snapchat’s core camera, messaging, and Minis products is the Snap Map. Its a personalized digital map that shows (and lets you message!) your friends around you. Snap just announced its crossed 250 million MAU’s, up from 200 million in Q2 ‘20, and now features 30 million businesses.

Snap announced a new Layers feature last week, which changes the map based on:

  • Memories: User photos and videos taken around the world

  • Explore: Public content from other users around the world

  • Infatuation: Restaurant reviews

  • Ticketmaster: Explore concerts and events - and jumps into Ticketmaster’s Mini

We’re in the early days of Snap opening the Map, and we’ll likely see more use cases built on it.

Content, Creators, and Spotlight

Snap shared Spotlight, Snapchat’s TikTok-like vertical video feature, now has 125 million MAU’s (up from 100 million in January). Viewers watching over 10 mins per day grew 70% in Q1, and it will soon roll out globally (Middle East launch was May 17th). Flexing the power of Snapchat’s built-in distribution, this 100 million MAU milestone took less than three months, compared to the 12 and 18 months it took Douyin / TikTok’s to reach 30 million / 100 million, respectfully, when first launched in China in 2016.

Snap initially made waves with Spotlight by paying out $1 million to content creators per day. Since launching at the end of November 2020, 5,400 people have earned over $130 million. In what felt inevitable, Snap recently changed the messaging around this payment to “millions per month”.

Snap also launched story replies allowing creators to insert user responses within new content. This looks similar to TikTok’s comment replies feature. Snap also announced users will be able to gift creators, of which Snap will take a cut, set to roll out later this year. Snap is notably adding web-based PC editing tools for Spotlight. This hints brands will start entering the Spotlight feed, and soon, ads.

More on the editing front, Snap announced Story Studio, a standalone mobile app for editing content that will launch on iOS later this year. It borrows much from TikTok’s editing suite (clipping, browsable sounds, effects) and taps into Snap’s more robust AR ecosystem.

Snap now has a total of 500 premium content partners, and content partner payouts have grown 100% over the past year. Launching less than two years ago, India had 70 million Snap Show viewers in 2020, and reach in the Middle East grew 300%. Not only do these premium shows have ads that bring in revenue to be re-invested in other areas of the product mentioned above, but much of this premium content acts as user acquisition. Content and creators act as a growth flywheel that helps kickstart Snapchat’s messaging network in each market, which then reinforces Discover’s content network.

Conclusion

This year’s Snap Partner Summit reminded us how far Snap has come over the past few years: build anything with Lens Studio, find anything with Scan. Augmented reality and Snap Games will start meshing with Spectacles and the Snap Map, and the advertising network originally seeded by the content business will be a monetization engine that weaves through it all.

Most importantly, its become clear Snap has taken a key lesson from Amazon. Each piece of the Snapchat platform - the Camera, Messaging, the Map, Games, Bitmoji, Stickers, Discover, Snap Ads - has quietly opened itself to developers and third parties through products like Snap Kit, Lens Studio, and Snap Audience Network. Historical cost centers like messaging, which sucked cloud resources and user time spent, are now positioned to drive material revenue for Snap through commerce and advertising in products like Minis, Games, and the Snap Map.

Snap’s business is evolving into a mix of messaging, content, advertising, software, and hardware. What started as an app to send disappearing pictures is starting to look like pieces of WeChat, Disney, Google, and Apple. Time will tell if Snap is ultimately an enduring business, however it appears to be positioning itself to be just that.

Launching Banana Capital

Announcing Fund 1

I recently closed Fund 1 of my new firm, Banana Capital. I raised the oversubscribed $9.999m fund in Q1.

Banana Capital is an internet-first sector, stage, and geography agnostic tech-focused investment firm. Fund 1 will write $150k-$300k initial investments in early-stage companies (pre-seed through Series A) with select later stage investments. This will expand over time from pre-seed to pre-IPO and into the public markets. I’ve made 15 investments (some new, some follow-on’s to prior investments), with five set to close in Q2. Fund 1 will have a total of 35 to 40 core portfolio companies. To date, 65% of Fund 1 founding teams include non-white founders and 50% of the teams include more than one gender.

Banana has the DNA to be a founder’s first outside investor or their newest supporter as a multi-billion dollar company. During this fund, I won’t lead rounds or take board seats, allowing for more collaboration with founders and other investors. Like most investors, I can help with things like marketing, PR, recruiting, fundraising, bug squashing, and memes. What I can uniquely bring to founders is high independent conviction in what they’re doing and speed of decision making.

Thank you to all the founders and LP’s that chose to partner with me on this journey. I am forever grateful for your advice, support, and trust.

I talked with Natasha at TechCrunch, and you can learn more about Banana Capital here. If you’d like to follow along, find me on Twitter, subscribe to my Substack below for longer-form writing, or check out bananacapital.vc.

WeeCare, Carebnb’s, and the US Child Care Epidemic

The LA-based startup fixing America's child care deserts

My partner and I welcomed our second daughter into the world seven weeks ago. Raising a child is possibly the most rewarding experience in the world, but 2020 feels like a strange time to do it. Every family is adapting to the changes brought by the COVID-19 pandemic, which has emphasized the many problems facing the US child care industry.

I first caught on to how LA-based WeeCare was addressing these structural issues head-on when I connected with co-founder and CEO Jessica Chang in 2018. Their team built a system to instantly spin-up a new daycare anywhere in the country. They recently launched a corporate benefits offering, allowing any business in the country to offer every employee a local child care solution overnight.

Child Care: A Massive Broken Market Hiding in Plain Sight

US child care is a $57 billion market (60% in-home, 40% in-center). An additional $22 billion is spent via government subsidies, and the amount of time spent and forgone wages by parents and family members providing unpaid care is orders of magnitude larger than the actual dollars spent. The largest historical and future driver of the market size is the participation rate of women in the workforce. If you read through investor presentations from industry leader Bright Horizons, a publicly traded operator of large child care centers valued at ~$10 billion, most of the industry’s profitability is driven by price increases, cost cuts through consolidation, and adding non-child care product offerings inside its centers.

It’s not a particularly large market, but it's often the largest expense for any household with children. These households spend more annually on child care than on housing or college tuition, and more than transportation, food, and healthcare combined.

60% of children under age six have parents who are both in the workforce, which (putting aside the often difficult decision to have someone else raise your kid) can make child care more affordable for working families. The unaffordability problem is more pronounced for single-parent households who spend an average 37% of their income on child care - up to 70% in some states like Massachusetts. It's a difficult reality that many single parents work all day only to pay most of what they earn to someone else to watch their young child(ren) for 10 hours.

Government subsidies have historically helped offset some of these costs. Unfortunately, the number of US facilities receiving subsidies decreased 59% between 2005 and 2017, with the number of small in-home centers legally operating without all proper licensing (and most prevalent in low income and communities of color) receiving subsidies down 75%.

There are also hidden costs of child care that are not captured by any of these metrics. 43% of professional women leave the workforce within two years of having children. 74% of women eventually return at least part-time, but only 40% ever return to full-time. 75% of women who leave the workforce say the top reason is child care related.

It’s estimated families without access to affordable care forgo $29 billion in wages annually. US businesses lose $4.4 billion annually due to employee absenteeism as a result of child care. In Louisiana alone, a lack of child care leads to $1.1 billion in economic losses. A survey also found 40% of parents regularly missed work due to child care issues in 2017, 16% of respondents quit their jobs because of child care issues, and 14% actually turned down a promotion in order to keep the same schedule or to ensure they did not lose access to a child care subsidy due to a pay raise.

Shrinking Supply and the Rise of Child Care Deserts

The state of the US child care market can be traced back to a massive supply / demand imbalance. There are roughly 690,000 licensed (60,000 unlicensed) daycares in the US, with 1.3 million Americans working in the child care profession. 8.3 million kids are enrolled in care: 900,000 with a nanny and 5.8 million with family, friends, or a neighbor. Approximately 7 million children under 5 are watched in a home: 4 million by a relative or friend, 3 million in a paid daycare.

Experts estimate there’s a 3:1 ratio of kids needing care to kids in care. The highest quality options often have waitlists of a year or more. This lack of supply primarily impacts women, and specifically those with low incomes, single parents, and people of color.

In Minnesota, the number of family child care providers dropped 25% between 2006 and 2015. The total slots available in Minnesota and Wisconsin dropped 5% between 2015 and 2020, and California’s total available slots dropped 9% between 2014 and 2017. Looking at data on US child care facilities, its apparent most of these closures are in-home daycares. No one knows why for sure, but one explanation is that older in-home providers are retiring without new educators replacing them. It’s a hard job, and stressful enough watching children all day, let alone running a business that typically isn’t very profitable.

This lack of supply has led to child care deserts: neighborhoods with little or no access to affordable child care. While this tends to skew towards low income and communities of color, a 2017 study found child care deserts are prevalent across the socio-economic spectrum. As few as 41% of top quintile, suburban US census tracts and as many as 60% of bottom quintile, urban neighborhoods are classified as child care deserts. This lack of supply means parents pay more, classes are fuller, and higher child to provider ratios generally lead to a lower quality of care.

Further explaining a lack of new educators entering the profession, child care isn’t generally a financially lucrative career. According to the BLS, as of May, 2019 the average wage for US child care workers was $11.65 per hour, or $24,230 per year. A Center for the Study of Child Care Employment (CSCCE) study found that 86% of providers who work with infants and toddlers are paid less than $15 per hour. The child care workforce is composed mainly of women, 40% of whom are of color. Nearly 15% of child care workers live in households with an annual income below the federal poverty line, and 36% below 200 percent of the poverty level. Half of all child care workers receive food stamps, welfare, or Medicaid.

Working as a child care professional has also historically been poor for someone’s health. A 2017 study of North Carolina family care providers found 29% did not have health insurance. Less than 50% met guidelines for physical activity, fruit/vegetable consumption, and sleep. Over 50% reported “high” stress, largely due to low wages that impose high financial constraints, 12-hour shifts, and no access to adequate space to take safe breaks.

WeeCare and the Carebnb

WeeCare created software built from the ground-up, entirely customized to simplify the process of running an in-home daycare. Founded in 2017, it launched its first daycare in January of 2018 and helps early childhood professionals manage:

  • licensing and regulations

  • curriculum creation and planning

  • launching, leads / marketing, virtual tours, and parent communication

  • billing, insurance, and taxes

  • government subsidies, and

  • all other back-of-house functions and paperwork

It’s essentially an in-home daycare “business in a box”. Customized software saves time for a child care professional spending an average of 56.5 hours per week watching children and reduces churn for WeeCare as they acquire clients for their customers. There’s often a significant disconnect between millennial parents and members of older generations providing care. Most don’t have reviews online, or even a website. WeeCare helps bridge that gap. And like any vertical SaaS business, there are lots of ancillary features (payments, insurance, etc) WeeCare can layer on over time to increase monetization and provider and parent stickiness.

WeeCare educators earn an income 2-3x higher than if they were working in a senior role at a large child care center - up to $100,000 per year, or 4x the national average. They can leverage their personal mortgage / rent payments for business use, and those with young families can spend the day around their own children while earning income teaching others.

WeeCare's in-home model often results in tuition prices 30-40% below market. The company is quickly scaling its subsidy program too. Over 80% of WeeCare locations have opted-into accepting families paying with subsidies, and over 40% have subsidized families currently enrolled.

In-home child care arguably provides better care for students. In home day cares average a 6:1 student to teacher ratio, compared to 12:1 in traditional centers, and safety and background checks are more strictly enforced. Homes also tend to be the primary option for families of color. It’s more affordable, convenient, and parents tend to be more comfortable with members of the community watching their children.

During the course of building and scaling the software product, the WeeCare team realized their customers needed help acquiring their own customers and turned it into a full marketplace that (after a lot of hard work) had acquisition flywheels sucking up both supply and demand sides of the marketplace, anywhere in the US, with the flick of a switch.

A Team Dedicated to the Problem

All of this comes back to the strength of the team. Co-Founder and CEO Jessica Chang was originally born in China. She landed a marketing internship at Cisco’s China division the summer before moving to San Francisco in 2002 to study Economics at Cal Berkeley. Three summers later, she was covering semiconductor equities at Citi while taking a full-time course load. She worked in M&A at Deutsche Bank after graduating and made the jump into a private equity operations role at Macquarie in 2007. Over the course of the next decade she started a family, bought and managed three preschools, and eventually landed at RadPad in 2016.

It was at RadPad where Jessica first met Matt Reilly. After a six month overlap, Matt was recruited to join Honey as Head of Growth. After a decade in growth and SEO roles, Matt kickstarted Honey’s initial growth flywheels. He left Honey, ultimately acquired by PayPal for $4 billion, to join WeeCare a year later. The duo then recruited friend Jesse Forrest, who was working as Director of Engineering at Tradesy, to join as CTO. The team’s growth, technical, and hands-on operational experience from previously scaled and exited startups sets the stage for today.

On-Demand, Scaleable Child Care

Fast forward to October of 2020, WeeCare built a product that can instantly bring supply anywhere in the US onto its network. Child care centers typically have a 20%-30% vacancy rate, and WeeCare can also flip on demand acquisition to fill any existing, non-WeeCare center’s unused capacity at scale. It takes less than a week to onboard a qualified provider onto the platform. This is (often more than) 10x faster than the national average of 2 months.

WeeCare’s home market of Los Angeles is a perfect case study of its model. It currently has over 900 schools in the Greater LA area, or approximately 18% of all in-home daycares in Los Angeles County. Digging into individual sub-markets on the map, it’s clear they have a deliberate supply acquisition strategy that hits specific sub-neighborhoods as needed.

Detroit and New York highlight two different phases of WeeCare’s growth within a market. Detroit is a newer city with WeeCare locations predominately in lower income areas, and New York reflects how its model, first launched in low income communities in LA, has began scaling into higher income neighborhoods like Manhattan.

WeeCare’s deliberate, productized approach has created a very capital efficient model. Using public data compiled in mid-March and a very basic estimate of supply compared to total capital raised, WeeCare was adding capacity roughly 10x more efficiently than its closest startup competitor, Wonderschool. Other startup competitors like MyVillage and NeighborSchools appear to have just started scaling outside their home markets. Compared to Bright Horizons, which operates 705 large child care centers, WeeCare adds new capacity 86x more efficiently.

Another way to frame WeeCare’s growth: Before WeeCare’s 2018 launch, Bright Horizons had a total of 695 US centers (page 21). By October of 2020, WeeCare had 2,700 US schools compared to a net of 13 additional centers brought online by Bright Horizons.

COVID-19: Closures, Panic, and Chaos

COVID unquestionably accelerated many of the aforementioned issues in the child care market. Attendance dropped as much as 90% nationally, and 80% of daycares in the SF Bay area closed during the peak of the crisis. In a March, 2020 survey of 6,000 providers, over 70% of home-based daycares had only one week of financial runway, and 63% didn’t think they could survive closing for more than 30 days. Experts predict loss of profitability and no federal funding could close 4.3 to 4.5 million slots in 2020, or roughly half the total supply of child care in the US.

Preschool capacity has disappeared as well due to licensing limitations and closures, while many K-12 schools (which allow most parents to work) have become virtual or staggered in the fall. And for many child care workers earning near minimum wage, going back to work actually means a pay cut from unemployment benefits in addition to risking their lives on the frontlines.

The anxiety for new parents sitting on waitlists exceeding a year have been exacerbated during this year's global pandemic. They’re now avoiding large centers with large class sizes. Since home daycares aren’t grouped by age, each house consists of only a few families and leads to less cross-contamination. Parents no longer commuting to work, mostly millennials raised in the sharing economy, are also finding convenience in local in-home options.

Worst of all, COVID has accelerated the number of parents (specifically women) leaving the workforce. Some due to job loss, many due to zero child care options. This accelerated when 865,000 women dropped out of the workforce when school resumed in September. According to a McKinsey study published in September, an additional one in four of all women are considering reducing hours, moving to part-time, switching to less demanding jobs, taking leaves of absence, or stepping away from the workforce altogether according.

During the onset of the pandemic in March, 12% of child care chains remained in operation. Home daycares fared slightly better at 25%. Over 250k child care professionals were laid off. In a corporate update, 21% (150) of Bright Horizons US centers remained open in March, with each classroom operating as self-contained classes (not the most efficient use of their massive centers). This number increased to 30% (210) by the end of June (page 26) and 55% (390) to end July.

90% of WeeCare’s centers remained open through the height of the Pandemic, and per Time, it continued adding capacity to reach to 2,700 in October. It also launched two new product lines: Fever Free and a corporate child care benefits offering.

Unlocking the Next Level of Growth with Corporate Benefits

COVID ultimately accelerated WeeCare’s end-game: the ability to quickly launch a daycare anywhere. It’s new corporate child care benefit allows any employer to provide child care for any (or every) employee as fast as one day. This includes support for a distributed team, an employee relocating across the country, or Instacart’s 200k essential on-demand workers in thousands of US cities.

Child care is an essential infrastructure for the US economy. In today’s work-from-home environment, most working parents (and moms, if we’re being honest) have not only continued working, but also became full-time teachers, caretakers, and often therapists for loved ones who have plunged into depression. Nearly 33% of the US workforce has children at home, and 10% of economic activity can’t happen if schools and daycares are closed. According to a 10-year study published in 2016, 54% of employers reported workplace child care benefits had a positive impact on absenteeism, reducing missed workdays by 30%. It also reduced employee turnover by 60%, and women who received assistance for child care costs were 40% more likely to remain at the same job two years after giving birth.

Employers paying for child care could expand the size of the market significantly as more unpaid, family, and friend-based care shifts to a professional setting. And WeeCare offers a solution to quickly spin-up the exact supply needed to support others in the child care benefits like Kinside, Winnie, and Care.com.

WeeCare’s ability to quickly spin up supply also benefits Bright Horizons. 66% of it’s 705 US centers are built on-site, and often funded by, its base of over 1,000 B2B customers.

Bright Horizons discloses these centers cost an average of $2.5m to build-out, and are unfortunately not often convenient for employees working from home, nor useful for distributed workforces. WeeCare is perfectly positioned to partner with large center operators like Bright Horizons to spin-up supply and fulfill contracts with their clients.

WeeCare’s Fever Free product launched in July represents another addition to its product stack. It’s the internal tool it used to detect family fevers from their homes pre-drop off, leading to zero COVID spread in any of its 2,700 locations.

Parents take a brief video of themselves and their kids using a digital thermometer. AI verifies the temperature reading and that it’s a real thermometer, and parents tap symptoms family members are feeling. I was surprised that competing apps are all self-reported, and WeeCare’s was the first product in the market to do this automatically.

WeeCare rolled this out to all their daycares in March, and opened it to other customers in Q3. Before launching, they were fielding inbound requests from daycares, then preschools, entire public and private school districts, and even universities. Inbound interest has also opened up channels to government entities, healthcare systems, and private sector businesses like film production and manufacturing plants.

Fever Free allows WeeCare to continue acquiring customers and building distribution if either schools remain closed and parents need in-home care, or if the world reopens and businesses are in-need of temperature monitoring. As WeeCare scales its instant child care + Fever Free solution, it becomes a very compelling partner in the child care benefits space.

Closing Thoughts

While the global COVID-19 pandemic of 2020 put the child care industry into turmoil and closed the majority of supply, WeeCare kept 90% of its centers (now near 100%) in operation. As you can imagine, more product lines and increased urgency in the problem it's solving have positioned it to grow even faster to end Q3 and into 2021. As WeeCare expands beyond in-home daycare, it’s on a path to positively impact every family in the US.

If you, your business, or someone you know is looking for a new child care experience, check out WeeCare. It’s likely better and more affordable than their current experience, and its model should become even more affordable over time. The team is working towards an incredible mission and I couldn’t imagine a more exciting company to be involved with.

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 Yongping, 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, Ouku.com, 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 yqphh.com, 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, Musical.ly 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.

Conclusion

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. If you’re building a company that borrows elements from Pinduoduo, I have invested in numerous companies with similar models from pre-seed to pre-IPO, and would love to learn more at turner[at]bananacapital.vc.

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.

Loading more posts…