GOAT: building trust in a resale marketplace
How GOAT uses machine learning and human experts to build trust, and how to cut fraud costs without slowing sign-up.
Role Co-authorWhen Carnegie Mellon, technology strategy courseTeam Five authors
GOAT is a marketplace for sneakers and streetwear. Counterfeits are common in resale, so GOAT authenticates items before they reach the buyer. Our report covered how that works and where the strategy could be stronger.
How GOAT builds trust
Sellers apply before they can list, and they upload standardized photos from several angles. Those photos train image-recognition models that learn what authentic items look like and flag inconsistencies. When a sale happens, the item goes to a GOAT distribution center first, where trained specialists check things a photo can’t show, like stitching, texture and feel. GOAT calls this “ship to verify.” Buyers are covered by a protection policy that gives an authentic replacement or a refund. The platform reported catching $72 million in counterfeit shoes in 2019.
GOAT’s seller commission is 9.5% against 8% to 10% for StockX, 13.25% for eBay and 15% to 80% for The RealReal.
Where we saw an opening
Authentication and buyer protection deal with counterfeits after they’ve been listed. They don’t remove the people who list them, and they cost a lot to run. Our main recommendation was to make it harder for bad actors to get in without hurting sign-up.
Sign-up is a growth metric for a marketplace, so we didn’t want to add friction at the start. We proposed asking for a valid photo ID later in onboarding, after a seller has already put in significant effort and is unlikely to leave. Automated ID verification services exist and are cheap compared with the cost of one counterfeit item going through the entire process. For a counterfeit seller, the extra effort makes listing less worth it.
Other recommendations
- Partnerships and creators: GOAT’s $50 million partnership with Paris Saint-Germain put the brand in front of millions. We suggested continuing that with rising athletes, and adding video and live-stream selling through creators
- Community: build on its 40 million members with events and content
- Data: use data from over 320,000 listings to give sellers pricing and inventory guidance, and to predict price movements
- Sustainability and resilience: focus more on second-hand luxury as demand grows, and keep a flexible fee structure that holds up when the economy slows
What I’d note today
GOAT is a clear case of human-in-the-loop AI. The model handles volume, and people handle what a photo can’t show. I use the same pattern when I design AI products in healthcare.
The report
