Ramesh Johari
Professor at Stanford (Operations Research and Data Science) and long-time technical advisor to Airbnb, Uber, Stripe, Bumble, and Upwork. One of the foremost academic practitioners at the intersection of marketplace economics, experimentation, and data science.
Appearances
Key ideas
Marketplaces sell friction removal. The product is not rides, stays, or jobs — it is the elimination of transaction costs that prevent markets from clearing. This framing determines monetisation: the platform’s fee model should align with the specific friction being removed at each growth stage.
A marketplace never starts as a marketplace. Before achieving liquidity on both sides, a platform must articulate its value proposition without scale. UrbanSitter removed payment friction; oDesk solved remote-work verification trust. The early business model is often different from the mature one.
Prediction vs. causal inference. ML finds correlations; decisions require causal reasoning. Sending promotions to high-LTV customers is a fallacy unless the promotion causes more spend. Rigorous data scientists think in terms of interventions and counterfactuals.
Whac-a-mole principle. Consequential marketplace changes reallocate inventory and attention, creating winners and losers. The right question is not “did a metric go up?” but “do the winners matter more than the losers?”
Rating systems and distributional fairness. No platform has solved ratings. A first negative rating correlates with an ~8% revenue drop and platform exit. Bayesian priors and double-blind reviews (the Airbnb model) improve information quality and fairness for new entrants.