We all know that our phones are gathering data to show us targeted ads and increase our chances of making a purchase. But what if this data was used for something more beneficial, like giving you access to loans that were previously unattainable?
With a background in risk management, Peter Barcak founded behavioral data analytics platform credolab, which aims to use smartphone and web metadata to assess a borrower’s creditworthiness, even if they have limited or no credit history.
With clients across nine industries in over 40 markets, credolab is looking to leverage its expertise in mining behavioral insights and expand to other verticals, including broader risk, fraud, and marketing-focused solutions.
KrASIA recently spoke with Peter to discuss how his platform works, and his plans for the future.
The following interview has been edited and consolidated for brevity and clarity.
KrASIA (Kr): Credolab analyzes over 10 million behavioral features. Are you able to share with us some of these behavioral features and their implications on digital credit scorecards?
Peter Barcak (PB): Our products are all about analyzing web and smartphone metadata. We look at a constellation of data points, such as the number of apps on a person’s phone, how often they schedule a calendar event or take selfies, battery usage, and more. Our machine learning algorithms identify which of these micro-behavioural patterns are predictive and stable over time, and we recommend our clients add them to their scorecards. By combining this behavioural assessment with traditional creditworthiness evaluation, our clients can approve more good customers, reject more bad customers, and reduce their risk and fraud costs.
Kr: How does analyzing clients’ mobile phone behaviours result in a better predictor of the clients’ credit scorecards compared to the traditional credit score?
PB: I believe that analyzing behaviors on a smartphone is a complementary way to assess creditworthiness, rather than a better one. While traditional data may indicate that two applicants are both creditworthy, analyzing their smartphone behavior may reveal a different story. For example, an applicant may have the ability to repay a loan but may not have the willingness to do so, and analyzing their smartphone behavior can help identify that. Credolab helps identify a high correlation between confirmed risky or fraudulent applicants and an applicant’s intent not to repay a loan. As a result, our clients have experienced benefits including up to a 40% increase in predictivity, up to a 22% decrease in fraud costs, and up to a 32% increase in approval rates.
Kr: In your opinion, what role does credolab play in promoting greater financial inclusion?
PB: I found that many people are financially excluded because credit reporting agencies (CRAs) can only provide scores for those with an existing credit history in the middle to upper-income groups. Credolab helps lenders reach more people by both assessing the past and predicting future behaviour. We also work with credit bureaus as clients and see a growing desire for inclusivity and improved scoring throughout the industry.
Kr: How does credolab differentiate itself from its competitors?
PB: While there are other companies that offer behavioral data for credit scoring or fraud detection, we realized about a year and a half ago that the raw data we were collecting could be monetized as a data enrichment package, not just as a score. By grouping different features together, we could solve specific problems, like detecting fraud or enriching data for marketing purposes. Our features provide a snapshot of the user, which can be fed into our client’s existing segmentation.
We’ve developed four modules for anti-fraud, account takeover, marketing segmentation, and data enrichment. These help us to perform different functions like detecting loan stacking or anomalous behaviors, device fingerprinting for Android and iOS devices, and optimizing marketing campaigns based on customer micro-behavioral patterns.
These modules have been released on our website and were built on information we have been collecting for seven years.
Kr: How have you begun to expand to non-financial institutions?
PB: Our solutions are not limited to consumer finance and digital lenders as we are now serving clients globally across different segments such as ride-hailing apps, BNPL and EWA players, crypto and student lenders, and insurance companies. However, not all businesses claiming to use alternative data actually do, and some face the challenge of inactive customers across the different services they offer. Credolab’s machine learning algorithms can identify opportunities within such user bases and help clients detect user intent from the first interaction, even with limited information.
Kr: Any other partnerships or upcoming plans you can share with us?
PB: I’m excited to announce that we’re testing a new marketing product that predicts personality types based on smartphone usage data. By applying the OCEAN Personality Framework to smartphone metadata, we can help mobile app publishers identify the personality traits of their users and customize communications accordingly, without compromising user privacy.
Our technology can recommend communication styles, images, and action verbs based on user behavior patterns, attitudes, and motivations. This enables marketing teams to have a deeper understanding of their users and personalize messaging according to their users’ traits.
We’ve already received interest from major online platforms in Latin America and Asian digital banks, and we hope to work with them in the future.