Tchibo GmbH is an international trading company with headquarters in Hamburg, Germany and more than 10,000 employees worldwide. Founded in 1949, Tchibo began as a coffee mail order company and today is known for its unique business model: In addition to high-quality specialty coffees, Tchibo offers a weekly changing assortment of non-food products, ranging from clothing to household items and electronics. With over 900 stores, international online stores and island stores in supermarkets, Tchibo is represented in numerous countries and achieved a turnover of 3.2 billion euros in 2023.
The project at a glance
- Conducted the first successful A/B test for product recommendations in the Tchibo mobile app in years
- Significant increase in the conversion rate for cross-selling products in the mobile app
- Provision of various recommendations data to different clients via a central interface
- Creation of a robust infrastructure that ensures continuous development and smooth operation
- Increased development speed thanks to a powerful platform and well-functioning product teams
Background
Tchibo has been operating digital commerce since 1997 and regularly addresses the question of how it can generate optimal product recommendations for its customers. These recommendations are issued via the website, the company's mobile app and the customer house newsletter. Tchibo has observed that sales success in the newsletter was greater than in the app.
In order to catch up in the app area, Tchibo has developed various prototypes (minimum viable products — MVPs) in the past and evaluated them using A/B tests. These calculate the individual product recommendations per customer and provide the recommendations data via an interface.
Some of these MVPs were successfully tested, which subsequently increased the interest of various internal development teams in the data. In order to meet this demand and ensure stable operation of the interface, it was now necessary to “capture” the detached prototypes and convert them into a central microservice.
Besides, the prototypes were developed on a platform that is primarily available to the Business Intelligence (BI) department and was not designed for such microservices. The aim of the project was therefore to outline and implement a solution that leaves the calculation of the recommendations data in the BI platform and at the same time locates the provision of the data at the point where all other productive microservices, such as the Product Data Service, are provided.
Solution
To calculate the recommendations, data pipelines that run every night were created with different approaches. These comprehensive data sets are then made available in different BigQuery tables.
A mirror service then has the task of synchronizing the data between BigQuery and a database suitable for productive purposes. As the existing service was slow and very unreliable, it was rebuilt together with codecentric as part of the project. The new service was developed in Golang, a language developed by Google that addresses the challenges of scalable, distributed systems. It is now much more stable and takes a fraction of the time of the old service.
The interface will eventually be used by the Tchibo webshop, the Tchibo mobile app, some internal services and the newsletter tool. This means that a very large number of requests per second can be expected, which is why Golang was also used here. Development was based on the API-first approach and in accordance with the Open API standard.
Result
At the end of the project, Tchibo was able to conduct the first successful A/B test of product recommendations for the Tchibo mobile app in years. Also, the recommendation engine is ready to play out several million personalized recommendations at a time. As soon as all systems are connected, they can access the new engine and issue personalized recommendations in a conversion-optimized way.
The successful test has also aroused the interest of various internal teams in the data. The interface is now robustly set up for this run and can withstand internal requests for data. The high-performance platform has also increased the speed of development.
The project uses the retail sector as an example of how big data technologies can help to increase sales and improve the customer experience. Tchibo now has an advanced data analytics solution that has a direct impact on sales and makes its own benefits immediately measurable.
Marcel Schließer
Data Engineer, Tchibo GmbH
"The development of such a service is both fascinating and challenging. Together with codecentric, we were able to make a major contribution to improving the shopping experience of our app users. I am convinced that the new big data solution will contribute to the company's continued success."
Marcel Schließer
Data Engineer, Tchibo GmbH
Any questions about the project?
Would you like to know more about the project? Are you interested in a similar solution for your company?
More projects of codecentric AG
Find out about other successful projects that we have completed with our customers. Perhaps you will find inspiration for a use case in your company.
Jimmy Nelle