Year: 2026
Type: dissertation
Abstract: This thesis asks a practical question: can ordinary spreadsheet software forecast demand for a perishable product well enough to run its inventory day to day? The product is bananas, and the data covers a full year of daily sales in one store. Because bananas spoil fast, the store needs short-term forecasts that keep shelves stocked without ordering more than it can sell. Small and medium firms rarely have specialist forecasting software, so the study sticks to tools they already own in Microsoft Excel. Three methods are tested and compared: Simple Moving Average, Exponential Smoothing, and Linear Regression. Mean Absolute Deviation measures how far each forecast falls from actual sales. Accuracy turned out to depend heavily on the day. Weekends were the hard part, with errors above 30%, while weekdays stayed calmer and easier to predict. On those calmer days Linear Regression was the most accurate. The wider lesson is modest but useful. Everyday spreadsheet tools can forecast well enough to lift service levels, cut spoilage, and guide reordering. They cannot do everything, though. A simpler model reacts more slowly, so good inventory control for perishables needs a sensible forecasting method backed up by rules such as safety stock.
Author: Sadasivan Pillai, Anjana
Institution: University of Maribor
Language: English
Field: Decision Sciences
Domain: Social Sciences
SDG: Zero hunger
Open Access status: green
OpenAlex ID: W7221221384
