Facoltà: Ingegneria Gestionale - classe LM-31
Collocazione: 19841
Autore: Rubino Rodolfo
Data: 20/12/2021
Titolo: Implementation of sales forecast models in food and bevareges industry: taste for luxury case
Relatore: Creazza Alessandro
Autorizzazione per la consultazione: NO
Le tesi si possono consultare unicamente in sede
The aim of this thesis was to implement and develop the most accurate forecast models in order to project the future sales for Taste For Luxury. Taste For Luxury is a company which has been exponentially growing in the last years, so it has been necessary to improve its processes so to keep up with the increment of business itself. For this reason, one of the first change that was needed to be implemented was focused on the forecasting process. As a matter of fact, TFL used to rely only on the experience of the managers for the estimation of future needs, so they decided it was time to have a better tool than just experience. Since the business is growing, the need for a more accurate forecast is absolutely essential so to avoid possible out of stock situations but also to avoid wasting perishable foods. With the goal of providing to TFL a tool able to project future sales, the first step was to perform a deep data and history sales cleaning. Once all set of data was cleansed, an analysis for each code present in the product portfolio was made so to understand if the historical sales belonged to the continuous or sporadic series. According to the type of historical series, the various forecasting models were applied: Brown, Holt, Holt-Winter, and Decomposition models for continuous series, and Poisson, and Croston for the sporadic ones. In order to understand the most suitable model for each product, the error analysis was performed. Unfortunately, the analysis of the historical series, although it has allowed to extrapolate the typical components that characterize the trend over time of the sales of a code, it has not allowed to generate valid forecasts for the totality of the articles of study. This consideration is made according to the error analysis carried out in the thesis. Nonetheless, it has not been possible to see the actual application of the models yet, since TFL has started to implement them only starting from October. A first step towards a better forecasting system has been made thanks to the developed Excel models, however the firm will improve even more the demand planning process by investing in specific forecast software which will provide a way more accurate forecast.
