Prediction of offline sales through online behavior
The sale of cars still takes place offline. Nevertheless, car manufacturers invest in online marketing measures, build up digital skills and / or purchase them. After all, it is a matter of wooing the target group in the digital space. Despite this, those responsible often struggle to justify the added value of digital efforts.
The aim was to generate concrete added value for Sales and Product Management Teams by examining the correlation between on-site behaviour and incoming orders. Specific goals were:
- Determining the time-lag between onsite behaviour and incoming orders/sales
- Identification of possible differences between countries/ markets and series
- The confirmation of a (presumed) connection between sales in the stationary trade and the onsite behaviour
Strategies and Actions
The relevant questions were identified via user stories, which were collected from various stakeholders. Descriptive and predictive analysis methods based on available data sources were used to answer these questions. Our approach included:
- An iterative and agile approach based on a prioritised backlog of requirements in the form of user stories
- Time lag quantification through Granger Causality Tests and auto- and cross-correlation analyses
- Training and optimisation of a vector autoregressive model (VAR) and subsequent evaluation
Strategic questions and concrete requirements from the MINI Product and Sales Management Team were answered. In particular based upon the onsite behaviour of potential customers, a statistically reliable forecast of incoming orders was made possible. The following findings were obtained:
- A high correlation between product showroom visits and sales (especially for short-term time lags) was proven.
- The ratio of onsite behaviour to sales differs greatly between different countries and also from series to series.
- A forecast of incoming orders was established for short periods of time.