Social Media Sentiment as a Leading Indicator of In-Store Product Demand: An Empirical Analysis

Conventional retail demand forecasting systems rely predominantly on historical point-of-sale(POS) transaction data, which creates an inherent temporal lag that prevents retailers fromresponding to emergent consumer demand signals before they materialise in sales data. This paperinvestigates whether real-time social media sentiment constitutes a statistically significant leadingindicator of in-store product demand across retail product categories. Employing a simulation-basedempirical design, this study generates 1,680 hourly observations spanning 14 days across fiveproduct categories with distinct sentiment trajectories — viral positive, plateauing, complaint spiral,recall event, and stable positive. Aspect-based sentiment scores (SDDS) are computed using theVADER lexical model and correlated with the simulated POS demand series using cross-correlationanalysis, ordinary least squares regression with lagged predictors, and Granger causality tests.Results indicate that social media sentiment Granger-causes retail demand for all five productcategories (p<0.0001), with an average optimal predictive lead time of 6.8 hours (range: 3–10hours). Asymmetric effects are confirmed for products experiencing negative sentiment trajectories,with statistically significant demand suppression preceding observable POS decline by 4–10 hours(p<0.05). These findings have direct implications for retail inventory management, suggesting thatsocial media sentiment signals can provide a viable, sub-daily forward indicator of store-leveldemand changes in both positive and negative directions.