How a microlending giant banks on data-driven decisions

Toni Sekinah, research analyst and features editor, DataIQ

Kiva is a non-profit microlending platform that has been using data to make better decisions. It is also making its data more accessible throughout the organisation and to academic researchers in a bid to deepen the body of knowledge around the impact of microlending on poverty alleviation. Kiva’s chief technology officer, Kevin O’Brien, who has been with the organisation for the last seven-and-a-half years, explained how.

Small business ownerThe organisation works by connecting lenders with those who want to borrow small amounts of money. Lenders can offer as little as $25 through the San Francisco-based platform which borrowers can use to start or grow a business, finance their studies or fund other activities that will help to better their situation. All the money from lenders goes towards funding loans, while Kiva covers its costs through optional donations, grants and sponsors. Kiva has funded loans to the tune of $1.17 billion, with a 97% repayment rate.

During the years that O’Brien has been with Kiva, he has progressed from software engineer to engineering manager to CTO, and has also seen an evolution in the way that the organisation uses data.

"Data-driven culture has reduced the operating budget by 25-30%."

He said that, when he first joined the organisation, "Kiva was going gangbusters in terms of growth” and so there wasn’t much need to deep dive into data because it was just trying to keep things afloat. However, when some of the hype around microfinance died down, Kiva started leveraging data to figure out how best to attract lenders and deploy their capital. He said: “We got better about using data for figuring out the way we can run things more efficiently and have the most impact.”

When O’Brien headed up Kiva’s engineering team, one of his first tasks in that role was to bring down the cost of operating. By bringing in a data-driven culture, he was successfully able to lower the overall operating budget by 25-30% over the course of a year-and-a-half.

He said that having a data-driven culture means that, instead of building out a lot of expensive projects, small tasks are completed to figure out if a hypothesis is worth exploring and, if so, throw more resources at it.

“That allowed us to have a leaner team and gave us a chance to step back and think about what Kiva does well and what we should focus our efforts on,” he said. Using an external provider for data warehousing was financially beneficial and so Kiva switched to Snowflake. O’Brien said this provided a significant saving of around 50% compared to running in-house.

"I was able to save donor dollars rather than overspending on software services."

“I’m happy to say that I was able to save donor dollars and put them into places where they could be more impactful, rather than overspending on software services or building things out that we didn’t need to build,” O’Brien said.

Kevin O'Brien, chief technology officer at KivaAnother important aspect of a data-driven culture is access to data for those who need it. The way O'Brien did that was to centralise data processing. “We took all the BigQuery analysis data pieces that we were doing in various places and pushed that into Snowflake. We transform it, we load it and are able to run some pretty heavy data analysis and we’re able to pull it into Looker and surface it to most of the people at Kiva,” he said.

However, it is not only people within Kiva who would like to see its data. O’Brien said that a lot of researchers request to see Kiva’s data to be able “to write some pretty interesting papers about the effect of Kiva dollars on poverty levels and social impact.”

The external data warehouse tool is used to build new snapshots of sanitised data so that academic researchers can undertake PhD-level research. The process for this is that once a Kiva has verified that a researcher is from an academic institution, it will look for the exact data that has been requested and make sure there is nothing personally identifiable within it. It then sets up a query, runs it and exports that data out to the academic. Ideally, O’Brien would like to set up a programme that would enable researchers to query the data directly once it was known not to contain any PII.

"Academics could teach us things that we could be doing better."

Moving forward, O’Brien wants there to be a data analyst on every team at Kiva across all departments. He is increasing the number of data engineers “to expand our ability to fulfil all the requests and do all the analysis and transformation to data we can.”

Although almost anybody in the organisation is able to access any data they need to make a decision, Kiva is still not leveraging the outside community as much as it could.

Academic researchers want to help Kiva with the skill and time they have to analyse its data. “We’ll be really well-served to leverage that talent that is available and have them teach us things that we could be doing better, and part of that is being able to get the data in front of them so they can do that analysis and come up with that insight,” said O’Brien.



Knowledge-based content manager, DataIQ
Toni is the senior features editor responsible for the origination of DataIQ's interviews, articles and blogs.