CLIENT CASE D'Ieteren Automotive
Clickstream optimization
Identifying the best path for conversion.
About
D'Ieteren Automotive
D’Ieteren Automotive is the biggest car importer in Belgium. Among its brands are Audi, Volkswagen, Skoda, SEAT and Porsche.
In recent years, D'Ieteren Automotive has seen an increase in online research when customers look to buy a new car.
In some cases, cars are even purchased entirely online, especially on the secondhand market.
USE CASE
Identify the best path for conversion for Volkswagen to start with. Research the relationship between the sessions on the website (which pages are visited, the order in which pages are being visited, etc.) and the final conversions.
Key points
Use an LSTM autoencoder
We started with creating groups based on similar paths. Therefore we used an LSTM autoencoder for converting the paths to embeddings. Based on these embeddings (numeric presentations) we used a K-means clustering to effectively divide the paths into groups. This enabled us to identify 7 groups, with clearly different traffic sources (E.g. Some clusters had a bigger share in paid search compared to other clusters, etc.) and a different conversion ratio.
Intermediate conclusions
Traffic sources are sending the traffic to the different paths (clusters) and different paths have a different conversion ratio.
Besides this, we analysed per cluster which paths were the most recurrent ones, what were the different converting paths and how were the different conversion paths being used.
Next steps
- Execute the analysis for other brands.
- Predict which sessions will convert (by using the recurrent neural net), so that the info can be used for retargeting (sessions with a higher change on conversions) and excluding (sessions with a lower change on conversion).
- Expand to a user-based model instead of a session-based model so that marketing attribution can also be included in the model.
CLIENT CASES
More successful cases
