Cake processes a growing number of bank transactions thanks to data engineering pipeline
Mobile application enriches data from different bank accounts and shares profits with its users
The Belgian mobile app Cake bundles and links the bank accounts of different banks in one mobile app with the aim of improving its user’s financial well-being. Using the app also yields financial rewards. In addition to a clear overview of the financial activities, all of the transactions are categorised and enriched with insights. The developers were looking for a solution to process and categorise these transactions coming in as raw, unstructured data in real time. Together with support from Dataroots, a robust data engineering pipeline was developed. Thanks to this model, the growing number of Cake users can continue to benefit from a scalable and stable user experience.
Cake was founded in 2018 by Davy Kestens and a small team of data scientists. Since then, the team of the start-up has grown to about 20 people and the public beta of the eponymous mobile application exceeds 6,000 registered users. These people link their various bank accounts to the app, where all transactions are neatly displayed and enriched with categories and spending insights. The app thus acts as a financial coach. Users automatically receive discounts or cashbacks from affiliated merchants, as well as a portion of the application's revenues. As a result, a typical Cake user already earned an average of 8.79 euros, just by using the app.
Since launching Cake's public beta at the end of 2019, over two million transactions have already been processed by the platform. These initially arrive at Cake as raw, unstructured data and are converted into insightful transactions for the user. It is essential that this is conducted in real time in order to have immediate insight into expenditure or to be able to obtain discounts. This requires a model that on the one hand can process this raw data very quickly and efficiently and on the other hand can continue to do so with an ever increasing number of users.
Cake's data science team initially worked out a viable solution themselves. Anticipating the rapidly increasing volume of transactions and users, it is fundamental that such a solution is scalable. In order to meet this requirement, external expertise was sought. Cake's model was placed behind an API, a software interface and Dataroots developed a technical element that acts as the link between this API and the front-end. The component sends the incoming transactions to the Cake model that processes the raw data and then to the application interface. This gives the user access to the enriched data in real time. For example, when the user buys a loaf of bread with a bank card from a connected account, this transaction appears in the app as soon as it is available at the bank. It is given the category 'Food' and the user gets an overview of the weekly or monthly food expenses as well as the complete expense history at that merchant.
Co-founder and head of data science from Cake, Jessica Ruelens explains: "Although our own solution was intrinsically functional, we soon ran into the capacity limits. As soon as the number of users grew and the amount of data increased, the system would crash regularly. Dataroots created a solution with the Spark framework on an AWS cloud environment, which guarantees excellent scalability in data processing. The solution is fully tailor-made, but is based on Dataroots' design principles. This involves disconnecting the model from the technical implementation, making it possible for developments on both sides to be carried out independently".
Part of the team
At first, a number of workshops were set up to analyse the current technology and the state of affairs. The outlines were mapped out collectively and from then on, Dataroots’ machine learning engineer Nick Schouten started working exclusively in-house for Cake. In addition, Cake could count on continued support from the Dataroots expertise centre.
Davy Kestens, founder and CEO of Cake deliberately chose to adopt this approach: "In terms of data science we are well versed in developing our models, but embedding this in the broader technical platform is a different matter. As we intend to remain lean and nimble with Cake, we found an excellent match in Dataroots as an external partner. In the meantime, Nick has truly become an integral part of the team and together we continue to develop new projects".
As Cake users grow, the amount of available data increases. This opens up opportunities in the field of data science. After anonymising this data, Cake can generate reports for retailers on insights into purchasing behaviour that are suitable for in-depth refinement. Half of the revenue generated by these reports goes to the users via a monthly deposit. For the development of these functionalities Cake also collaborates with Dataroots.
"Dataroots also possess the expertise in the field of data science, so we can count on them for this too. It's truly a pleasure to work with a team that brings so much know-how to the table," says Jessica Ruelens. "They are as invested in our success as we are.We crawled into the code right from the first workshop, while other parties stay very high level in the initial stages and don't go into depth. This fits in perfectly with our doers spirit as a start-up".