How Landbay is democratizing data with dbt and Looker
Landbay’s vision is to be the UK’s leading buy-to-let lending platform–they want to be the go-to lender for brokers and the partner of choice for investors. And by all accounts, they are well on their way to achieving this vision. In 2018, Landbay’s 1500% growth over a four year horizon earned it a spot on Deloitte’s “fast 50” awards recognizing the UKs fastest growing companies. In 2019, the 90-person team earned 10 “buy-to-let lender of the year” awards. This is the story of how Landbay is democratizing data to achieve their vision.
Landbay is a buy-to-let mortgage marketplace lender. Borrowers pay interest, and Landbay’s platform passes that interest along to its institutional investors. The team completed its first mortgage in July 2014, began deploying institutional capital in July 2017, and signed a landmark £1billion funding agreement in partnership with an established global financial institution in July 2019. To date, Landbay has completed over £400 million of specialist buy-to-let mortgages, and has had zero defaults–a particular point of pride for a team that pays close attention to underwriting quality.
“The key difference with Landbay is our marketplace funding model,” said Julian Cork, COO at Landbay. “The marketplace has two sides, the lending and funding. Our vision is to be the go-to lender in this space and a default option for brokers, as well as the partner of choice for those that want to invest in buy-to-let mortgages.”
Data is central to this vision.
The most pressing need for access to data arose after Landbay onboarded its first institutional investor. The UK requires a critical report for all mortgage lenders called the “Bank of England report”. This report contains a mortgage lender’s entire portfolio of loans–where the money was loaned, repayment timelines, and portfolio performance. It’s a massive collection of data. In order for Landbay to work with a greater number of institutional investors, it needed to automate the production of the Bank of England report.
Landbay is built on an AWS microservice architecture that includes 79 databases. The business also uses a variety of SaaS tools–Netsuite ERP, Segment, and Google Analytics, just to name a few–to power core business operations. The infrastructure they adopted for the Bank of England report needed to extend to a variety of future use cases as well.
This is when Landbay began evaluating Looker. The deeper they went in the buying process with Looker, the more they began to hear about dbt. “I had heard about dbt before,” said Chris Burrell, CTO at Landbay. “And it was mentioned on nearly every customer reference call we did with Looker customers.” By separating the analytics engineering workflow that happens in dbt from the data analysis environment, data teams can ensure that all data served up in the business intelligence tools is accurate, tested, and modeled in a way that is easy to analyze–even if you’re not an analyst.
With this setup, the Landbay team was able to democratize data in three key ways:
1. Deliver accurate, tested data to Looker users
With dbt, Landbay can deliver the Bank of England report quickly and, even more importantly, with absolute confidence that every data point is correct. “Before we rolled out the new setup, we did a massive quality assurance check on every piece of our data transformation logic,” said Bruno Murino, Data Analyst at Landbay. “We sat down with brokers, lenders, and product managers and asked: ‘What needs to be true at all times?’ And then we wrote dbt tests to catch those errors.”
With data models built in dbt, anyone running any report in Looker can be confident that the data is 100% accurate. “If an engineer makes a semantic change, dbt tests catch that,” said Chris. “And we stop the bad data from flowing into Looker: Looker users might see old data, but they’ll never see wrong data.”
2. Add new data sources quickly
Landbay aims to provide a better-than-great customer experience and achieving this requires data. For example, one customer experience improvement they’ve made recently is to speed up the loan approval process by gathering all the data underwriters needed in one place. “Using a Glue Crawler we can automatically create tables with external data in Redshift, then we clean the data and join it on the postcode using dbt, and finally serve it up to end users in Looker.” Bruno said. “We have already incorporated land registry data, police data, fire brigade data and geographical data using that method.” Instead of conducting each of these searches individually, underwriters can access the data in Looker and quickly determine how sustainable and affordable a particular loan is.
Opportunities like this exist in every part of the mortgage process:
3. Build a data-driven culture without a massive data team
With dbt, Landbay’s two-person data team is able to meet the data demands of a fast-growing, 90-person company. This efficiency comes from spending their time providing clean, tested data sets rather than tracking down data quality issues or building dashboard after dashboard in Looker. “With dbt, we can empower users without technology team involvement, and the outcomes for the business are huge,” Julian said. “All of our data–broker payment automation, sales order value reporting, broker performance, marketing campaigns–is served up by dbt and accessible in Looker. As a result, people feel empowered to get the best outcomes for Landbay.”