Lunar at a glance
Launched in 2015, Lunar is the leading Nordic 100% digital bank. More than 850,000 users in Denmark, Sweden, and Norway use Lunar, including 20,000 entrepreneurs who use Lunar Business to manage their finances.
Industry | Data stack | Company size |
---|---|---|
Fintech, B2C | dbt, Looker, Redshift | 501-1000 |
Getting executive buy-in through the data governance framework
Lunar’s company-wide strategy is to differentiate considerably from other Nordic banks by offering a simple and easy way to handle day-to-day finances and relevant products through third-party integrations. Simple on the surface, complex under the hood.
Lunar’s 25-person data team is essential to making this happen, as data is deeply ingrained in the company’s top priorities.
- High standards against financial crime – grow the business without compromising the high standards against financial crime by having data-driven controls to catch fraud and money laundering.
- Personalisation driven by AI – give customers proactive tips tailored to their lifestyle, helping them stay ahead as they spend, save, and grow their money.
- Reliable external and internal reporting – deliver transparent and accurate data to regulators and customers to build trust through each interaction.
The data team’s objective is straightforward: build world-class data governance with less than 10% of traditional banks’ data headcount. To deliver on this, Lunar developed a data governance framework consisting of eight principles, placing an emphasis on securing C-level sponsorship.
Lunar’s Data Governance Framework
Lunar kicked off multiple initiatives to bring the Data Governance framework to life across the organization, including using data contracts, annotating GDPR data, and data observability in Synq. The data observability requirements were clear: Define the critical data elements, establish sufficient monitoring, and implement an ownership model. Building similar capabilities in-house would require several data engineer FTEs.
“Every three months, we meet with the chief risk officer, chief technology officer, and bank CEO to update them on the latest developments, risks, and opportunities. This helps everyone contribute to and have a stake in our data quality“ — Casper, Director of Data.
Measuring and monitoring critical data assets
Critical data elements are the most important metrics at Lunar and are used for key processes or decisions. This includes customer, loan, and sign-up metrics. These must be held to a higher standard, with clear ownership and comprehensive monitoring.
“With 1,500 data models, 6,000+ dbt tests, and hundreds of dashboards, it’s inevitable that some tests will fail at any given point. Knowing if a critical data element is impacted used to be nearly impossible“ — Sine, Head of Data Warehouse
The team at Lunar used Synq to define their ten most important data elements as Data Products. The Data Products dashboard is used daily to get an overview of the product’s health and if there are issues that stakeholders should be notified about.
Critical Data Elements defined as Data Products
“Although we still have issues, we’re now aware of them and know if they need to be addressed. And we’re more often than not the first to notice” – Bjørn, Data Governance Manager
Improving test coverage through automated freshness and volume monitors.
The team at Lunar knew they had a weakness in monitoring the sources flowing into the data lake and data warehouse. With more than 800 sources and data arriving at different intervals, some conditioned on business seasonality, maintaining thresholds for freshness and expected volume in dbt tests was not scalable.
“As a regulated bank, our KPIs must be accurate, reliable, and updated on time. Anomaly monitors are an essential complement to our dbt tests to complete our data governance strategy” – Casper, Director of Data.
With Synq, the data team deployed rules to automatically add freshness and volume monitors to new sources, ensuring they’re always aware of stale or missing data.
An anomaly monitor was triggered due to a sudden drop in the business customer row count.
Activating roles and responsibilities through unified ownership
Lunar built an internal framework for identifying data owner roles for most data assets. Service Data Owner (SDO) owns stewardship practices on data. The Data Lake Owner (DLO) owns the raw data collection. The Data Model Owner (DMO) works with and governs data within the data warehouse, and the Business Metric Owner (BMO) defines and monitors key business metrics.
Ownership used to live in a static document, which meant that it was most often an afterthought. To address this, Lunar created 12 owner groups, including Financial Crime Prevention (FCP), Advanced Modelling, and Users & Onboarding in Synq. Issues on data assets are automatically tagging the relevant owner in Slack, creating an increased sense of ownership, reducing triaging time, and improving time to resolution.
Lunar’s ownership model in action – issues are routed to relevant Slack channels, and owners are tagged
“The ownership engine in Synq lets us mix up defining ownership through code in dbt yml files with selecting folders or assets in the Synq UI. This ensures that the ownership workflow fits technical and less technical users,” says Bjørn.
Defining ownership in Synq using the dbt yml owner definition
The data governance team also uses the ownership model to track and improve data quality across domains.
“Relevant owners are automatically tagged in Slack, and alerts are sent to relevant channels to ensure they don’t get lost in the noise” – Nicolai, Senior Data Analyst
Reporting on data quality KPIs to the C-level
“We knew our data quality was improving. And we knew that some areas were weaker than others. But we had no way of tracking and communicating progress to our senior stakeholders,” says Bjørn.
The team worked on a customized data quality report using metadata collected by Synq. The report can be broken down into relevant dimensions, such as critical data assets and ownership domain, so the team can systematically address gaps. For example, it may be okay to have low coverage of data quality controls for newer domains, whereas domains such as Finance need much tighter controls.
Outline from the March data quality report–coverage and uptime are broken down by criticality and tracked over time
“Every three months, we have our executive data committee with the C-level. We run them over the data quality metrics to show where we’ve improved and be transparent about where we have gaps” — Bjørn, Data Governance Manager.
“The power of our data governance framework is that it all works together,” says Bjørn. A good example of this is:
- Automated monitors combined with dbt tests help us be the first to learn about issues.
- Highlighting alerts on critical data elements expedites resolution priority.
- Having owners assigned ensures that issues are brought in front of the right person.
- Measuring data quality across several dimensions helps us systematically improve.
We’ve defined critical elements, ownership, and monitoring across hundreds of key tables. Next up for us is starting to expand our use cases for the data warehouse so it can directly power critical reports such as external customer reporting and regulatory reporting – Sine, Head of Data Warehouse