"The modern data stack propelled a pioneering digital bank on a journey towards profitability and becoming data-driven." Leon Van Dyk, Head of Data and Decisioning, Kuda
- Fivetran replaces the work of 5 data engineers, allowing them to focus on other tasks
- Kuda now benefits from valuable metrics including customer acquisition, products, and service monitoring
- Company’s C-suite executive use dashboards daily for a wide-range of insights including customer acquisition and account activity
- Their next step is to build out predictive analytics and explore Intercom for more sophisticated customer analytics
Sources: Azure SQL databases, Intercom
Destination: Google BigQuery
Business Intelligence: Looker
Cloud Platform: Microsoft Azure
Launched in Nigeria in 2019, Kuda is a digital bank that offers the country’s residents a simpler way to look after their personal finances. At the core of the service is an app that includes tools for tracking expenditure and money management. On a spectacular growth trajectory, the bank had a fourfold increase in customers in the last six months alone.
As an agile, mobile-first digital company, Kuda knew it needed to be data-driven and identified the modern data stack as essential for achieving its goal. At first, the data team of five people was manually building data pipelines and relied on SQL Server Reporting Services (SSRS) to extract insights from transactional databases. Their data split into 12 disparate Azure SQL databases with no way to successfully join the data across their internal sources.
When Leon Van Dyk, Head of Data and Decisioning joined Kuda, the team was looking to move from running OLAP queries on an OLTP database, which was hard work.
“It was manual, very limited, very narrow and it impacted on the operational performance of the servers, we went back to the drawing board and looked at what was best practice in data architecture.”
Leon wanted a more scalable solution that would relieve his team of having to build and manage data pipelines. Fivetran was chosen as the data pipeline, Google BigQuery as the data warehouse, dbt for data transformation and Looker as the BI tool.
“That combination is currently best-of-breed for what we do, I don't have any problems thinking about scalability now, because I can create more connectors as I need them.”
The decision was taken to go with Fivetran for ELT, with dbt performing data transformations in BigQuery. At the other end, Van Dyk resolved the ‘build vs. buy’ debate by engaging Fivetran as a fully managed ELT service.
“Building is one thing, but operating and maintaining is where a service like Fivetran thrives,”
The source that delivers the most valuable data comes from Kuda’s customer-facing core banking transactional systems. This includes metrics around customer acquisition, products, and service monitoring. Kuda also uses advanced analytics for credit scoring, mitigating risk by making more informed overdraft offers to customers based on data.
Analysing data from Intercom, a customer conversation platform, helps the support team allocate resources and plan for the type of questions they are most likely to be asked. The goal of getting Intercom data into BigQuery is to improve the customer experience and use insights to create better customer engagement models.
Having identified data as the lifeblood of its business, Kuda invested in a best-in-class data stack that has more than met expectations. The most important metric of success is the bottom line, according to Leon Van Dyk, and the modern data stack strategy has delivered.
“Our revenue is increasing on a continuous basis as a direct result of better visibility through data of what our customers want, how our products perform, and how we can optimise production processes,”
The company’s C-suite executives use dashboards to uncover wide-ranging insights, from customer acquisition to account activity. Sophisticated visualizations show when customers sign up, how they engage, and transactions categorised by merchants. Growth and marketing can see when prospects fall out of the sales funnel and when customers are eligible for an overdraft. Data made available also allows Kuda to satisfy its regulatory requirements, e.g. credit information that needs to be submitted to a national credit bureau on a regular basis.
“Fivetran has helped us get where we are with data management today, building the pipeline is one thing but managing the operational and monitoring side was important for us, getting somebody well known in the business for being reliable to take over that pain.”
Van Dyk estimates that it would take at least five internal hires to cover the workloads that Fivetran automates and runs in the background. He also values the way Fivetran takes responsibility
”I recall how the Fivetran team proactively responded on a Google BigQuery service alert, and reran the jobs affected at no cost to Kuda, and only notifying us when the problem was resolved.”
Kuda is well advanced on its journey to being data-driven but there’s much more to come. Kuda is already doing descriptive analytics, but its data team wants to move forward on more advanced analytics, specifically predictive and prescriptive analytics, to allow Kuda to be able to provide the right product to the right customer at the lowest price point and lowest risk to shareholders. Having a best-of-breed data stack is the start of the journey, not the final destination.