AboutYou Data-Driven Personalization with ArangoDB
Florian Krause, Head of Development, About You GmbH (Otto Group)
The overall shopping experience improves significantly when customers feel like they have personal guidance. This is true for our 700k+ existing customers and also new users coming from fast moving traffic sources. Both groups require a rather different approach. We do know the taste/habits of our customers very well but have little information about new users. This means that we have to create valuable personalised recommendations with little information available. This was one of our main requirements at AboutYou when we started building our own recommendation engine.
With AboutYou every user can manipulate a variety of settings for his account e.g. Style, favourite colour, favourite brands or modern fashion trends he/she follows. Additionally users are able to “Like” or “Dislike” products. All this user-generated information is being used to further improve their shopping experience.
At AboutYou we create tailored points of contact, such as user’s personal feed with individual recommendations for outfits or real-time suggestions and individual mailings.
Our challenge didn’t stop there. We wanted to integrate external apps for creating additional value for our customers. This lead to the need for a scalable solution which has a general, handy and secure API. Furthermore, we wanted to guarantee high security standards and use just a minimum of personal data.
We noticed quite quickly that common BI solutions and relational databases hit their limits early with these requirements and thus becoming “offline” for us. The development of a proprietary solution which shares the same data with BI and Shop was inevitable.
With AboutYou we want to achieve a quality of recommendations which match the ones of a professional style guide or your best friend. This means providing individually tailored recommendations by taking into consideration fitting style, trends, age, sex and more. The solution is a smart combination of different data models which allows simultaneous usage of documents and graphs.
After an intensive evaluation of different vendors and their technology we decided to use ArangoDB. ArangoDB is a multi-model NoSQL database which allows us to model data as documents or graphs. This database stores all data schema-free in a – NoSQL typical – JSON format and enables us to use the same data collections with document or graph queries.
Meanwhile the amount of data which is handled by ArangoDB exceeds 100GB and 300 million nodes. ArangoDB delivers a query-response-time of less than 20ms for our time-critical recommendations. Therefore we significantly improve the overall user experience with fast page loading and individual recommendations.
By having the opportunity to provide suitable recommendations for our users we were able to reduce the amount of information presented. Too many products irritate people. Furthermore, we could significantly increase the click through rate of our mailings. We recognized an increase in the amount of time users spend on AboutYou.de compared to industry average. This all didn’t only lead to higher sales but also enabled us to identify and target early adopters.
Foxx is directly running in the database which saves network communication between client and DB server. Thus query-response-cycles are being reduced due to less network communication and latency-time decreases. Via Foxx we can precisely adjust all data access of external applications individually while keeping privacy protection on a very high level.
We at AboutYou want to create the most personal experience possible in online retailing. Our database related questions found a compelling answer with ArangoDB.
Big thanks to Florian Krause for investing time to write this use case!