Analytical Assessment of Retail Sales of Dairy Products Using Data Mining

Accurate forecasting of sales amount plays an important role in effective production and sourcing management of manufacturing companies which ultimately leads to high profitability. The paper proposes an effective use of data mining approaches to analyze and extract clusters of purchases that best characterizing the purchasing behavior of the dairy customers in different regions of Iran in a dominant dairy production company. The paper applies clustering and classification algorithms for differentiating the purchasing behavior across Iran. It employs six phases of business understanding, data understanding, data preparation and pre-processing, modeling, evaluation, and explanation of predictions for the managers of selected company. The data gathering and modeling phases are complemented by analytical predictions and conclusion. The research proves that for the dairy products that are highly perishable and should be produced on a daily basis, unexpected fluctuations in consumer demand may pose significant problems to the commercialization and sales of such products. It also categorizes the products sales behavior in different seasons and provides new directions for practitioners on the dairy products that should be selected for each season and city based on seasonal fluctuations. Due to the limitations in accessing the central databases of dairy companies and products, the researchers were required to gather data from daily databases for a limited period of time and for specific products. It is recommended to gather a full dairy data set for a broader range of products for the purpose of a more comprehensive validation and examination of results throughout the industry. The paper includes a practical analysis of appropriate sales policies for each category of sales behavior for the purpose of high profitability. The final forecasting solution demonstrates that demand in dairy market is very sensitive to changes in weather condition while it does not show a quick reaction to price changes. The paper concentrates on the need to maintain a balance between meeting customer demands and controlling inventory costs and also fulfills a major longtime need for enterprises seeking good sales performance and profitability through accurate analytical predictions.