Regulators are responding to growing platform power with curbs on platforms’ potentially biased exercise of power, creating urgent needs for both a workable definition of platform bias and ways to detect and measure it. We develop a simple equilibrium framework in which consumers choose among ranked alternatives, while the platform chooses product display ranks based on product characteristics and prices. We define the platform’s ranks to be biased if they deliver outcomes that lie below the frontier that maximizes a weighted sum of seller and consumer surplus. This framework leads to two bias testing approaches, which we compare using Monte Carlo simulations, as well as data from Amazon, Expedia, and Spotify. We then illustrate the use of our structural framework directly, producing estimates of both platform bias and its welfare cost. The EU’s Digital Services Act’s provision for researcher data access would allow easy implementation of our approach in contexts important to policy makers.