Multi-Channel Sales Forecasting in a Kenyan Retail Environment

The fast and continuous progress in technology is significantly impacting not only the dynamics of retail and e-commerce but also the innovations in sales forecasting. Identifying what could potentially be the most effective method available for predicting sales across online and offline sales channels in this rapidly growing market is becoming more and more crucial to business owners. This study on multi-channel retailing examines the performances of conventional time series forecasting models ARIMA/SARIMA and the more recent approaches like Facebook Prophet and RNN. For this purpose, Kenyan retail time series data of three years was utilized. Our findings highlight that among these methods, LSTM emerges as the top performer, demonstrating superior accuracy in capturing the complex characteristics of the sales data. It is also important to note that based on the results of the experiments, SARIMA still has potential especially in short-term predictions of time series with strong seasonality.