Credolab turns to smartphones as a proxy for credit scoring

  • Raises US$3.2 million in total funding; sees 38% quarterly revenue growth
  • Has access to users’ anonymised data, uses it solely for credit scoring


The Credolab team (from left) Global Sales head Arun Kalra; co-founder & CEO Peter Barcak; and chief product officer Michele Tucci

FOUNDED in 2016, Singapore-based Credolab is in the B2B business of selling its credit scoring tool that uses mobile phone metadata to help those in finance institutions with the underwriting of loans and credit cards.

In a prior interview with the chief executive officer of GoBear, Adrian Chng, he shed light on the company's alternative credit scoring tool, Easy Apply, developed in partnership with Credolab.

Wearing many hats, Chng is the co-founder of Credolab as well as the founder and chairman of Fintonia Group, a venture capital and private equity firm that invested US$800,000 (RM3.35 million) in Credolab. To date, Credolab has raised US$3.2 million (RM13.39 million) in total funding and has 49 financial institutions as clients spanning 14 countries.

Credolab has processed over 10 million loan applicants to date, a 2,000 times growth from when it first launched. The company closed April 2019 with an average rate of return (ARR) of US$2 million (RM8.37 million) and reports a 38% revenue growth quarter-on-quarter.

To further understand Credolab’s offering, DNA spoke to the chief product officer, Michele Tucci about how the company’s tool differs and whether traditional credit-scoring methods will soon be replaced. While the banks still need to carry out the ordinary process of income and identity verification, Credolab augments the credit-decisioning process through its tool.

Measuring willingness vs ability to repay

Among the underbanked, the absence of transactional data such as bill or credit card payments prove to be a challenge when it comes to credit-scoring. “On top of the bank’s process, we provide a complementary assessment of the user and this is typically done on a behavioural level rather than on a compliance or transactional level,” said Tucci.

He claims that past attempts to use alternative sources of data to credit score the underbanked have failed because of the lack of predictiveness from the data sources. “The predictiveness of a scorecard is measure by the Gini coefficient from 0 to 1 with 0 being equal to flipping a coin in terms of chances of getting it right and 1 being absolute predictiveness.”

"Using social media or psychometric data was leading to very low Gini scores which gave very little room to banks and lenders to approve unbanked customers confidently,” explained Tucci. Credolab, instead, uses mobile device metadata or “data about other data.”

In the case of GoBear’s Easy Apply, some examples of anonymised data captured are the size of files in e-mails, the number of contacts and mobile apps, and the amount of data the mobile user consumes. The propriety algorithms extracts and analyses these data points to be used as predictive scorecards.

Since the there is no transactional data to indicate the ability of underbanked users to make loan repayments, Credolab measures the users’ willingness to repay instead with their mobile phones as proxy. Interestingly, the mobile penetration rate among this group is at almost 70% as reported by banking clients. “Being unbanked doesn’t mean they are not sophisticated or are completely tapped out of the digital world.”

Assuring privacy and security

Admittedly, Credolab accesses private information on users’ mobile phones but it stresses that personal information is not used in assessments. "The idea is to preserve the privacy of users. Yes, we do access information that are technically private because it's coming from the mobile phone of an individual. But we don't do so to sell advertising or monetise data. We do it for the sole purpose of assessing their credit- worthiness,” Tucci clarifies.

In fact, the data Credolab collects is not shared with its clients and the company assures all local regulations are complied by. "We store this metadata for up to three years which, besides being a requirement of the law, is a way to improve our algorithms over time. Machine learning algorithms are quite greedy in terms of data analysed and so the more data we can feed the machine, the more accurate is becomes over time.”

But with concerns over the regulatory environment not evolving fast enough to keep up with technological changes, the use of Credolab’s tool ultimately boils down to consumers trusting their banks. "Our app is downloaded only as part of a particular application process of a particular credit card or personal loan. If the user trusts the bank, then they trust Credolab. In the app we explain very clearly what data we collect, why and how we process it. By being transparent, we try to gain their trust."

Credolab’s banking clients, however, have declined to be named. As far as security measures go, the app also uses the one-time-password (OTP) as a unique identifier “to make sure the customer is who he claims he is and the mobile phone belongs to him”.

On the horizon, Credolab is in the process of launching a mobile platform with anti-fraud and e-know-your-customer (eKYC) capabilities for banks to outsource its digital onboarding. In the first quarter of 2020, the company aims to offer analytics-as-a-service with credit profiling and intent detection.


Related Stories :

Keyword(s) :
Author Name :
Download Digerati50 2020-2021 PDF

Digerati50 2020-2021

Get and download a digital copy of Digerati50 2020-2021