Sales Forecasting Automotive

Challenge

Planning and forecasting sales figures for markets, model series, and variants is a central challenge for all OEMs. These plans guide decisions in production, financial planning, and other business-relevant units. The closer the planned figures align with the actual sales numbers, the more solidly the business can be planned. Our client, an international automotive corporation, faced the challenge of their current planning figures, which involved significant manual effort, not achieving the desired accuracy.

Approach

At the project's start, available data sources were surveyed and consolidated. Various internal and external data sources were used and structured for further modeling. After data cleaning and validation, different statistical tests and models were applied to determine the seasonal and autocorrelative structure of the sales time series. Based on these insights, a state-of-the-art machine learning model was developed to accurately forecast monthly sales at both the overall and type-class levels.

Result

The model delivers an accuracy of 98% and reduces the planning error by approximately 48%. Based on the monthly forecasts, planning processes will be shifted to data-driven metrics in the future. The model was implemented into the client's infrastructure and provides all forecasts monthly as database entries, which are fed into an interactive dashboard for the sales planning department.