This study provides a framework for the evaluation of third-party datasets that can serve as valuable forward indicators for demand forecasting at Tempur Sealy International. Using a multi-step methodology involving exogenous feature selection, lag analysis, and baseline LSTM modeling, the study evaluates the extent to which third-party datasets enhance forecast performance across various product and channel combinations. While some indicators, such as consumer sentiment and retail activity, exhibit predictive value, the overall results are mixed with improvements in accuracy depending heavily on variable selection and context. These findings underscore the need for rigorous feature evaluation, model customization, and ongoing validation when incorporating exogenous data into demand planning processes.
