Context

Grant LaFontaine co-founded Whatnot, a live-shopping marketplace, in December 2019, starting with an authenticated Funko Pop marketplace before pivoting into live video commerce after watching customers try to sell items to each other over social media. He's interviewed by a16z General Partner David George, a longtime investor in the company. Whatnot now processes over $8 billion a year in sales and users spend an average of 95 minutes a day on the platform.

The Big Idea

Users don't care about the marketplace, they care about whether they'll enjoy the experience, and treating commerce as entertainment rather than a transaction is what let Whatnot expand the total market for live shopping instead of just moving existing e-commerce online.

Live commerce is already 30 to 40% of all commerce in China versus single digits in the US, and Whatnot's bet is that the gap closes because discovery-driven shopping creates demand that a search-driven marketplace never surfaces in the first place.

Key Insights

Customers don't buy a market, they buy value

Users never think in terms of "this is a live shopping market," they only ask what value a product provides them right now. Building toward "the market" as a category, rather than a specific compelling experience, is what distracts teams from the actual user experience. LaFontaine credits Whatnot's lack of market sophistication at launch as an advantage: knowing a market too well tends to make founders iterate on what already exists instead of building something new.

Discovery-driven commerce expands the market instead of just moving it

Traditional e-commerce requires knowing exactly what you want, which only digitizes existing offline sales. Live commerce is demand expansionary: users discover things they didn't know they wanted, the way a shopping mall historically expanded retail demand rather than just relocating it. That's why LaFontaine expects live commerce's China-sized share to eventually apply to a market that's also larger in absolute terms, not just redistributed.

Trust and safety is 40% of the company

• Headcount: roughly 40% of Whatnot's employees work on trust and safety, which LaFontaine compares to running the police force and legal system for a city of tens of millions of people.
• System: an automated rules engine processes billions of data points in under a second to flag harassment, late shipping, or a high refund rate, and can suspend or ban sellers programmatically.
• Why: this is the actual product investment that makes strangers trust each other enough to buy fresh fish or put a credit card on file with a stranger's shop.

AI is pointed at seller efficiency, not human replacement

• Rejected: seller avatars and other tools that would substitute for the human interaction buyers actually return for.
• Adopted instead: auto-generating listing metadata, inferring stream tags for discovery, and surfacing analytics on where a seller's business is over- or under-performing.
• Why: the platform's core value is that buyers know the specific shop owner and come back to them, the same reason people return to a favorite market stall.

Small business economics work differently on Whatnot

• Whatnot deliberately surfaces individual sellers and their stores, unlike marketplaces that obscure who a buyer is purchasing from to maximize platform efficiency, so buyers build a relationship and return to the same person.
• The largest sellers on the platform generate over $100 million in annual revenue with 20 to 40% EBITDA margins.
• Whatnot intentionally keeps its own take rate low so businesses keep the majority of the value they create.

The product pivot came from watching users hack around limits

Whatnot's original product was an authenticated Funko Pop marketplace. The live-commerce pivot happened after the founders noticed their own customers already trying to sell items to each other over live video on social media, with broken checkout, shipping, and discovery. Rather than researching the Asian live-commerce market first, they built a better version of the workaround their existing users had already invented.

Mental Models & Frameworks

The shopping mall analogy for demand expansion

Live commerce is closer to the historical invention of the department store or shopping mall than to e-commerce: both expanded what people bought by making browsing and discovery enjoyable on their own, not just efficient for a known purchase. Apply it by asking whether a product is optimizing for a known-intent transaction or for expanding what users didn't know they wanted, since the two need fundamentally different design and metrics.

Trade-offs & Nuance

Low take rate versus faster monetization

Whatnot keeps its cut of GMV deliberately small rather than maximizing near-term revenue, on the reasoning that seller success compounds into platform growth over time. This trades slower near-term revenue capture for a stronger long-term flywheel, a bet that only pays off if seller retention and repeat buying are strong enough to grow GMV faster than a higher take rate would have grown revenue directly.

Practical Application

Audit whether you're building a market or an experience

Before adding another marketplace feature, ask whether it makes the product a better standalone experience users want to return to, or just makes "the market" more complete on paper.

Watch for users hacking around your product's limits

Treat a workaround your users have already invented, the way Whatnot's founders treated customers improvising sales over social video, as a stronger discovery signal than researching what competitors elsewhere are doing.

Size trust investment to the actual risk surface

• Do: if your product depends on strangers trusting each other with money, treat trust and safety as core product investment, not a support cost.
• Why: Whatnot put roughly 40% of its headcount there because the platform's growth was gated by how much buyers and sellers trusted it, not by acquisition.

Point AI at seller efficiency, not human replacement

When adding AI to a marketplace, apply it to tasks like metadata generation and performance analytics that make existing sellers more effective, rather than tools that substitute for the human interaction buyers actually return for.

Bottom Line

Whatnot grew by treating commerce as entertainment and trust as a core product investment rather than a market to compete in, and its low take rate combined with a 40%-of-headcount bet on trust and safety only makes sense if you believe discovery-driven experience, not transaction efficiency, is what actually expands a market.