Forecasting demand has become increasingly difficult as TikTok and other social media platforms can create rapid demand spikes that are not captured by traditional forecasting methods. This project addresses that limitation by developing a predictive model for CosmeticCo’s North American Consumer Products Division that integrates TikTok engagement signals into demand planning. The model uses a two-stage approach: logistic regression estimates the probability of entering a high-demand regime within a seven-day horizon, while quantile regression estimates the expected uplift if that event occurs. In the proposed model, high demand was defined as actual seven-day demand reaching at least 1.6 times the baseline forecast, with alerts triggered at a 30% probability threshold and uplift estimated at the 75th percentile. During a 115-day test period, the model achieved 95% overall accuracy and correctly identified 4 of 10 actual high-demand events, resulting in 40% recall and 100% precision for the high-demand class. Although the model did not capture every spike, every high-demand alert it generated was correct, making it useful as a focused early warning signal. Without the model, the baseline forecast failed to anticipate significant incremental demand during high-demand periods, representing a substantial volume of potential lost sales. By incorporating TikTok engagement signals, our model recovered visibility into 72.3% of that uplift, directly reducing potential lost sales exposure by approximately three-quarters of the undetected volume, which planners could now act on. These results demonstrate that incorporating TikTok engagement signals can help anticipate social-media-driven demand surges and improve visibility into their magnitude, enabling planners to identify high-risk periods earlier than the baseline forecast alone. In practice, this framework can support more informed decisions on inventory positioning, supplier readiness, production planning, and escalation of response actions when social media activity suggests a potential surge in demand.
