Demand forecasting for companies with many branches, low sales numbers per product, and non-recurring orderings

TL;DR

This paper introduces the Top-Dog-Index, a robust non-parametric measure derived from real-world sales data, to classify branches by supply-demand deviations and improve forecasting in companies with many small branches and irregular ordering patterns.

Contribution

The paper presents the Top-Dog-Index as a new, statistically robust non-parametric tool for classifying branches based on supply-demand deviations, aiding demand forecasting without estimating branch-specific demand directly.

Findings

Top-Dog-Index is statistically robust in real-world data.

It effectively classifies branches into oversupply or undersupply clusters.

The approach improves supply-demand alignment iteratively.

Abstract

We propose the new Top-Dog-Index to quantify the historic deviation of the supply data of many small branches for a commodity group from sales data. On the one hand, the common parametric assumptions on the customer demand distribution in the literature could not at all be supported in our real-world data set. On the other hand, a reasonably-looking non-parametric approach to estimate the demand distribution for the different branches directly from the sales distribution could only provide us with statistically weak and unreliable estimates for the future demand. Based on real-world sales data from our industry partner we provide evidence that our Top-Dog-Index is statistically robust. Using the Top-Dog-Index, we propose a heuristics to improve the branch-dependent proportion between supply and demand. Our approach cannot estimate the branch-dependent demand directly. It can, however, classify the branches into a given number of clusters according to an historic oversupply or undersupply. This classification of branches can iteratively be used to adapt the branch distribution of supply and demand in the future.