Saman Modiri, Scott Fay, The Role of Delivery Performance in Shaping Customer Satisfaction in Multi-Marketplace Platforms: Evidence from Brazilian E-Commerce [Draft available upon request]

This paper examines how delivery performance shapes customer satisfaction in multi-marketplace platforms (MMPs), where a single store owner lists products across multiple online marketplaces and manages logistics on behalf of small and midsize business (SMB) sellers. Using transaction-level data from Olist, a Brazilian MMP, we exploit the fact that last-mile delivery is controlled by third-party carriers, not sellers, to causally identify the effect of late delivery on a rich set of customer outcomes. Beyond star ratings, we construct sentiment, emotion, topic attribution, and review quality measures using transformer-based NLP models. We find that late delivery substantially reduces review scores, sentiment, and review quality while increasing frustration, with negative effects persisting after time, location and seller fixed effects. Critically, topic decomposition reveals that delivery-induced dissatisfaction spills over to perceptions of product quality and seller reputation imposing a reputational externality on sellers who had no role in the logistics failure. Heterogeneity analysis using causal forests shows that customers given shorter delivery promises are more forgiving of delays, while longer delivery estimates amplify the marginal penalty. These findings carry implications for how marketplace platforms design delivery promises, how store owners manage logistics quality, and how SMB sellers navigate a reputational environment they partially control.