The Value of Social Media Data in Color Trends Forecasting
Operations and Information Management Department; Faculty Adviser: Marshall Fisher
This research explores the value of social media data in forecasting color trends in fashion. It addresses two questions: whether the current social media data can help better predict future color trends in the fashion industry than current practices, and if so, how should retailers make use of the results to improve their inventory management? Results from this research will be interesting to apparel retailers that want to utilize social media data to improve their supply chain operations.
This project analyzes Twitter data and a retailer’s historical sales data. The Twitter data we scraped contains detailed information on every tweet of a list of 360 fashion insiders’ Twitter accounts in a recent five-month period, as well as their account information. The historical sales data from a U.S. apparel retailer contains transaction-level retail sales, operations data, and the product color attribute information.
Initial analyses have shown a significant positive correlation between the number of retweets of a color on Twitter and the three-month-later sales of that color, after controlling for all time-invariant variables, price, stock-outs, and the seasonality. That is, current Twitter data on colors is a significant predictor of three-month-later apparel sales by colors, which enables most fashion retailers to adjust their inventory accordingly. Comparing with the baseline model without social media data, our final model helps explain an additional 7.5% variation in the sales, and decreases the Mean Absolute Deviation by about 25%. Further analyses are needed to incorporate the predictive analytics into the retailer’s current inventory planning process.
Publications, Presentations, & Awards
“The Value of Social Media Data in Color Trends Forecasting,” POMS (Production and Operations Management Society), Washington, D.C. (May 2015)
“The Value of Social Media Data in Color Trends Forecasting,” COER (Consortium for Operational Excellence in Retailing), Harvard Business School (June 2015)