Forecasting the volume of sales of goods of a trading company

Abstract

The research carried out clustering and forecasting of sales volumes of products of the enterprise "Polyprint" LLC. The entire assortment of "Polyprint" LLC is divided into two groups: sets of stickers with one sheet and sets with two sheets. In each group, a matrix of distances between objects was calculated using the Euclidean method and clustering was performed by full enumeration. Using full enumeration clustering, clusters were formed in each group, which included similar products in demand. Five clusters include only one product. The other four clusters contain three to four products. Also, in the 9th cluster of Group II, the initial data is better described by a linear trend. By summing the sales data for each cluster and adding them to the graph along with the second-order polynomial, a better approximation of the original data to the second-order polynomial was obtained. Nine sales forecasting models in the form of second-order polynomials for 19 products were developed. Based on them, the optimal size of production was found. The economic effect of implementing the models was calculated. It showed that by using them, Polyprint LLC would be able to save 7% to 38% on monthly production costs. The company also developed an automated information system based on MS Excel, which allows to model the demand for the company's products, build a sales forecast and calculate the economic effect of the forecast implementation. The algorithm for forecasting sales using MS Excel can be used by analogy not only at the above-mentioned enterprise, but also in other sectors of the national economy.