Abstract:
SPAR-Roodeplaat is part of an international group of independently owned and operated retailers who work together in partnership under the SPAR franchise brand. The store does not only offer a wide range of day-to-day grocery products, but they also have their own in-store bakery, deli and butchery. SPAR has been finding inventory management and planning of their ‘departments’ (bakery and deli) challenging as they do not have a set ordering system in place.
Products produced by the bakery and deli are perishable items as the freshness of these products decrease daily and therefore can only be sold to customers for a few days. Consequently, items not sold at the end of the day are considered as wastage and this leads to a loss. However, running out of stock also leads to a loss of revenue and will have an impact on customer satisfaction. An ABC analysis was conducted to investigate the problem and to determine the number of stock-outs occurring.
A literature review was conducted to obtain a better understanding of the problem and to get a general idea of how similar problems were solved. After reviewing the appropriate literature, the project goals have been determined. The project aim is to determine the most accurate and appropriate forecasting technique to be used for the bakery and deli departments. Numerous forecasting techniques were investigated but emphasis was placed on multiple linear regression, Prophet forecasting, autoregressive integrated moving average (ARIMA) and artificial neural networks (ANN) models.
The four models mentioned above were developed using R and were compared using four metrics to determine the most accurate model. Mean average percentage error (MAPE), mean average squared error (MASE), mean absolute difference (MDAE) and mean average error (MAE) were the four metrics used.
The overall accuracy, as per experimentation, and other practical reasons have concluded that Prophet forecasting is both the most practical and accurate of the four algorithms as defined by the methodology of this project. However, recommended improvements to be made include further optimisation of each model, the inclusion of promotional dates and better understanding of the each product unique time series. A proposed implementation plan has also been discussed.
The visual form of a Gantt chart was used as a guideline to complete the necessary activities within the given deadlines.
On completion of this project, it is apparent that an accurate forecasting model will make inventory planning easier within their bakery and deli departments. An accurate forecasting model will not only assist the retail store with production and inventory planning, but will also be useful when making important business decisions in areas of finance and marketing.
