Sales Forecasting, Polls vs Swarms

Abstract

Sales forecasts are critical to businesses of all sizes, enabling teams to project revenue, prioritize marketing, plan distribution, and scale inventory levels. Research shows, however, that sales forecasts of new products are highly inaccurate due to scarcity of historical data and weakness in subjective judgements required to compensate for lack of data. The present study explores sales forecasting performed by human groups, and compares the accuracy of group forecasts generated by traditional polling to those made using real-time Artificial Swarm Intelligence (ASI), a technique which has been shown to amplify the forecasting accuracy of human teams in a wide range of fields. In collaboration with a major fashion retailer and a major fashion publisher, three groups of fashion-conscious millennial women (each with 12 to 15 participants) were asked to predict the relative sales volumes of eight clothing products (sweaters) during the 2018 holiday season, first by ranking each sweater’s sales in an online poll, and then using an online software platform called Swarm to form an ASI system. The Swarm-based forecasts were significantly more accurate than polling such that the top four sweaters ranked using Swarm sold 23.7% more units than the top four sweaters as ranked by survey, (p = 0.0497). These results suggest that ASI swarms of small groups can be used to forecast sales with significantly higher accuracy than a traditional polling.

Keywords: Swarm Intelligence, Artificial Intelligence, Collective Intelligence, Sales Forecasting, Product Forecasting, Customer Research, Market Research, Customer Intelligence, Marketing, Business Insights