Can crowdsourcing improve prediction accuracy in fashion retail buying?

Abstract

Fashion retailers’ buyers must decide how much to buy of merchandise well before a selling season. This order quantity decision is always challenging due to limited historical data and high demand unpredictability arising from the novelty of fashion merchandise. Despite many attempts to solve this longstanding problem, industry reports consistently show that fashion buyers’ predictions of product salability and future sales are frequently inaccurate, leading to loss of profits for retailers. In this research, the authors take an “Empirics-First” approach to explore an alternative solution, crowdsourced forecasts from ordinary customers, and investigate whether crowdsourced forecasts would be more accurate than those of expert fashion buyers and if so, how should a crowd be formed in terms of size and composition? After conducting an online experiment, finding that forecasts by a “crowd” of ordinary customers are significantly more accurate than those of expert fashion buyers, the authors test a contingency framework in a second empirical study examining how crowd size and composition impact forecasting accuracy for products of varying fashionability. The results revealed that heterogeneity in a crowd is a key factor in prediction accuracy. Specifically, crowds with more variation in income and shopping frequency made notably accurate predictions. Another key finding of the study pertains to the required crowd size; increasing the size of a crowd at first sharply decreased the crowd's prediction error. However, after a certain point, there were diminishing returns in prediction accuracy. Given the interesting results, the paper concludes with guidelines for implementing crowdsourced forecasting by fashion retailers and directions for future research.