“Every decision at Revolut has to be made with data.” So said the lead data scientist of the digital banking app, Abhi Thanendran.
Due to its philosophy of democratised data throughout the company, everybody is able to get the data and everyone has the same understanding of the definitions of the KPIs and metrics. The data team that Thanendran heads up has to enable this to happen by ensuring that the platform for accessing or analysing data runs smoothly.
He said: “The platform has to be clear for new people to understand what's going on and figure out how Revolut works. It also has to be able to support multiple dashboards and the different things people want to analyse.”
"The data department provides the brain."
Thanendran also said that the company's goal for data is for it to be applied as a function to every single team in Revolut. The teams at this three-year-old start-up are like “little squads” that have a product owner and team of developers and data scientists that help that team to accelerate. “Where the data data department comes in is we provide the brain for whatever they are trying to do.”
Previously, Thanendran and his data team were using a Postres database which meant that they had to write custom code to take the data and prepare it in such a way that it could be queried in a short amount of time. The disadvantage of this was it was not scalable.
“Anytime people wanted to do any new analytics, you had to have dedicated data scientists or data engineers writing this code, so you have to keep hiring people just to do that rather than focusing on big problems.”
This strain on resources led to the decision to get an analytics database which would do the number crunching quickly and eliminate the need to prepare the data. Thanendran initially used BigQuery to solve this problem, but found that this too had drawbacks. “BigQuery it is hosted by Google and you don't actually have control over the servers.”
"It would take a week to get the data ready, now it's minutes."
In May 2018, he and his team decided to use Exasol as a central data repository for the entire business. The trial period ran from May until October, after which Revolut began the discussions to get the paid version.
For Thanendran, the main benefits of the new technology were related to speed and control. He said: “If someone asked for something, it would take a week just to get the data ready and now it is minutes where they can write a query quickly and get the data. Moreover, it has allowed us to better control over the permissions of access in terms of who can see what - that has also improved.”
He also said that there were quantifiable gains in regards to fraud detection, customer satisfaction, engagement, customer acquisition, financial reporting and growth. Better access to data has led to better insights.
An example of this was with the data from the hiring pipeline. This was analysed to see which interview questions were most successful in identifying candidates and which interview stage had the highest number of drop-offs. Looking at customer satisfaction, they can now more easily take data from everything including chats, queue times and external review sites like Trustpilot.
Although Exasol offered to send out an engineer, Thanendran said that it only took one or two members of his “highly technical team” to handle the migration, and the main cost involved was that of getting the servers on Google Cloud.
"Data knowledge is spread out like bacteria."
Throughout the whole process, the new technology has been running unobtrusively in the background, making working with data more efficient. “People don’t really need to know about the database we’re using, but we have seen very good responses ever since we implemented it.”
The organisational philosophy of giving all employees access to data has been in place since day one, said Thanendran. This means that there is no need to hire data analysts. They wanted to avoid of the pitfall of “keyman risk” or “tribal knowledge”. This is the phenomenon where data analysts are hired and focus on doing analysis on one or two specific things in a certain team.
When that person leaves the team, all of that knowledge is lost. In contrast, data knowledge at Revolut is not kept in isolation but instead is “spread out like bacteria.”
"The end goal is to apply data to every single thing."
Thanendran’s team is made up of data scientists and data engineers who will look at the complex analytics and focus a lot on building algorithms, whereas general employees will look at simple metrics and aggregation.
“We use Exasol to get the data and then we'll we do a lot of things on Python or other programming languages to analyse the data in the way that we need, whereas normal regular employees will just write a SQL query on it and answer whatever questions they want. The end goal here is to apply data to every single thing and just go from there.”