Impressing the Algorithm

Abstract

We study how platform ranking design shapes sellers’ off-platform pricing when the platform learns product appeal from on-platform sales. Because those sales depend on cross-channel pricing, ranking creates a career-concerns force: by restraining offplatform discounts, the seller shifts demand on-platform and improves future exposure. Thus, even without conditioning directly on off-platform prices, a sales-based algorithm can reproduce the incentive effects of explicit price-parity restrictions. But because the platform also relies on on-platform sales to infer seller appeal, a sales-based algorithm may distort promotion decisions and reduce welfare. We compare sales-based algorithms with rules that condition directly on off-platform prices and show that a price-parityclause algorithm can attain the first best.