Baseline prediction of point of sales data for trade promotion optimization

Abstract

Baseline prediction is an important to devise marketing strategy for a consumer goods product. Simulation techniques, time series algorithms are often used to generate baseline for the future. However the algorithm that fits a particular point of sales (POS) data varies according to the datasets. Sample set of point of sales data were simulated under different conditions and constraints incorporating seasonal and non seasonal trends. This study has compared the performance of two time series models namely Winters model and linear exponential smoothening on the simulated datasets. Winters model was found to be a better fit for the point of sales data that were used for testing.