This article originally appeared in DataIQ. Author David Reed is the editor of the publication.
Zopa is a peer-to-peer lending company which operates via an online portal. Launched in 2005, it has enabled £526 million to be lent from a pool of 46,000 active savers to 71,000 borrowers. The first company of this kind in the world, its growth has been rapid. The business was also due to come under new regulations introduced by the Financial Conduct Authority to tighten up lending rules. Given its commitment to safe lending, which includes offering loans only to the most credit-worthy and spreading each loan across a minimum of 200 savers, Zopa identified a need to improve its risk management as well as its customer experience by digging deeper into its data.
As Matt McGuire, chief technology officer at Zopa, told DataIQ: “We are a consumer-to-consumer business, so customer insight is very important for product design and converting that audience into customers. Zopa is also a marketplace, so we have the constant challenge of developing relationships between customers. That gives us two different data challenges: the first is to understand the capital flows within the customer base and the second is a more traditional customer insight angle.”
As a small company, Zopa’s resources were relatively constrained. It had a team of four analysts, two of which were delivering against customer insight-type, while the marketing team relied on its own tools. To achieve its industry-beating low rate of bad debt - just 0.19 per cent since 2010 - the business uses credit-scoring software from SAS, but was relying on basic tools for other types of analysis.
“We had been doing things for some time using traditional tools, including Google Analytics, which were a pathway towards recognising that we were missing some of those insights,” says McGuire. With the increasing level of lending activity across the platform, data volumes were putting those resources under pressure at a time when the desire for better insight was increasing.
“A year ago, I saw a lot of talk about analytics and how it was driving product development in companies across a wide range of industries. For me, that highlighted our need,” McGuire recalls. “We had been missing a lot of information, even at the customer level, because we had not been getting the best view. It always helps when something is fashionable - if analytics has been proven to work in other firms, it gives yours an appetite to consider it.”
The first-of-its-kind social lending platform was unlikely to adopt a conventional approach to this problem. McGuire had previous experience of working with software developer Thoughtworks which has its own unique positioning. It describes its purpose as to “revolutionise software creation and delivery, while advocating positive social change."
While McGuire recalls that the developer made an approach and expressed an interest in working with Zopa, client principal at Thoughtworks Jaydeep Korde recalls a more emotional pull: “It is the peer-to-peer lending model which to us was interesting. We’re very interested in how business and technology can disrupt existing business models. I prostrated myself in front of them and said, ‘we love what you are doing, how could we work together?’ There is a real cultural fit between us.”
With a proposal to run a proof-of-concept project, the two partners needed to identify which business challenges should be tackled and the most appropriate data sets to use. “We had an intelligent discussion about how best to make use of their offer. We decided to deploy them against a selection of risk data, rather than marketing as they had the skills already. Most value would be delivered by looking at risk and our parameters,” says McGuire.
“We focused on the customer journey - for example, if you request a personal loan - and the idea that the journey through our marketplace could tell us about your risk profile. It was quite an interactive process at first,” he adds.
Thoughtworks deployed the Clojure programming language to dig into Zopa’s data, validating or rejecting a range of hypotheses about customer interactions and lending times, as well as how customers could be offered the best rates. This is central to the customer experience and to ensuring savers achieve a good return - the average return achieved is 5 per cent.
Says Korde: “What we were looking to see from that data was whether we could draw up any questions that would lead to non-intuitive answers and which would then lead on to further non-intuitive questions, rather than providing confirmation of things they already knew.”
The data being used to support this investigative process lay outside of the sets being run through either Google Analytics or SAS. Most of it was unstructured data from online interactions and had never been considered for use by the risk function before. “It was about opening our eyes and understanding what was possible,” says McGuire.
One of the most significant outcomes from the project was the decision to hire a chief data scientist, Didier Baclin, who joins from a background working at data-rich companies like Amazon. Reporting directly to the CEO, the new hire has made an immediate impact. “I made the case that there would be benefits, even in the early days, but it has turned out to be transformational,” says McGuire.
He notes that the search for insight is all about exploring new opportunities and that it is also part of the agile process which the IT community has adopted. “You can’t predict discovery - you have to reach it. Two months in, our head of data science is still deploying insights which Thoughtworks developed. He is scaling it up for a growing business which is having to innovate and make product changes,” says McGuire.
Zopa is still considering how to develop the project into a solution to support this new analytical focus. “At the moment, it is not a technology, but it may become one,” he says. Korde points out that the nature of the approach which Thoughtworks uses makes it a more difficult pitch compared to traditional IT solutions. That makes Zopa all the more unusual for having adopted it. “Big organisations are often pioneers. But Zopa is in a unique position - it can’t afford not to be innovative,” says Korde.
The project also had benefits for the software developer. Korde says it is surprising how often online businesses struggle with their data sets and the fact that they have to use offline processes to run their operations. Insights from working with Zopa extend into improvements for Thoughtworks itself. Says Korde: “From an analyst’s point of view, having come up with those non-intuitive lines of enquiry, we got better.”
Disclaimer: The statements and opinions expressed in this article are those of the author(s) and do not necessarily reflect the positions of Thoughtworks.