The Porsche Interactive Forecasting Dashboard enables comparing different predictive models in a visual easy-to-grasp manner to optimize demand forecast and adapt to market needs in-time. Improved forecasting at Porsche leads to reductions in inventory and logistics costs and improved on demand fulfilment. The car buyer benefits from shorter delivery times which leads to a higher customer satisfaction.
A big challenge in the automobile industry, but especially for car importers and dealerships, is to adjust to the rapidly changing market demands. Therefore, involved stakeholders need an on-demand forecasting model to base their decisions on. Together with Porsche Austria, we analyzed market performances data of new car models in the past with multiple algorithms and methods, which included new-, used- and tactical car registrations, sales data and additional market data sources like working days per month and results of market investigations. A direct comparison of multiple predictive models showed that, depending on the data aggregation level (brands, segments like SUV and specific car models), certain algorithms performed significantly better than others. The two most accurate models, a recurrent neuronal network and the seasonal ARIMA, were then implemented in an interactive dashboard giving insights on the past behavior and future predictions of the market. The dashboard allows users to investigate the predicted values for each brand on multiple error metrics over time and for the whole segment itself to gain clear understanding of the model performance. The benefit for the project is recognizable for both Porsche and the car buyer (end user).