Can you beat the credit scoring engines through your digital behaviour?

Credit scores are predictive indicators of credit worthiness of a consumer or a business looking for credit. Historically credit bureaus such as Experian, Equifax etc., have had proprietory credit models to predict behaviour of customers and their probability of default. However, with Fintech and social media, surely there are ways to augment these models?

They stole your identity, but after seeing your credit score, they gave it back.

Recently, a research by a team from the National Bureau of Economic Research covered over 270,000 purchases between October 2015 and December 2016. The purchases were from a furniture store that offered credit, and also used digital foot print as a parameter to make credit decisions. There were some interesting findings from the research.

Whether we like it or not, all of us who are on the web, either from a mobile or a laptop, leave a digital footprint. In the process we inadvertently give away far too much information to the algorithms that judge if we are credit worthy or not. Digital footprints are often considered to be proxy for character. They are often used alongside scores from credit bureaus to assess credit worthiness of a customer.

The good news is that, many variables that these algorithms use are very much under the control of the customer. A brief summary of the findings from the research are as follows.

Device used to access the internet

When we access the website of an ecommerce firm to make a purchase, they can identify the kind of device we use to access the site. As per the research, customers who use iOS devices are considered to be in the top quartile of credit worthiness.

ISP used to access the internet

The credit worthiness of a demography can be a function of the ISP they use to access the internet. In Germany, where the research was held, customers who used T-online are more affluent and considered credit worthy.


Image Source

EMail provider used

Using a free and largely outdated email service such as Yahoo or Hotmail points to higher default rates. On the other hand, a paid email service like T-Online indicates better credit worthiness.

EMail address used

The email address used revealed the credit worthiness of a customer. If a customer had Forename and Surname on the email address, it indicated higher credit worthiness. Abstractly named email addresses and those that do not have any connection to attributes of a customer generally indicated higher default rates.

This could also be extended to SMEs, where the company was named after the founder, the default rates were lower.

Channel used to visit an Ecommerce site

The mode of reaching the website where the transaction happened highlighted the self-control of the customer. A customer who used a pay per click ad or an affiliate link to get to the site was found to have higher credit risk. Instead typing the address of the site into the browser or through an unpaid search engine indicated better credit worthiness.

Time of the day

The time of the day when a customer accessed an ecommerce site and applied for credit informed the algorithm about the customer. Customers accessing the site during the day (12-6 PM) were almost half likely to default as those who got online between 12 AM and 6 AM.

These were the variables that the research identified as core to the algorithms that used digital footprints. However, there are other interesting variables too. For example, new age credit scoring firms use fonts on a customer’s machine as a variable to decide credit worthiness. Customers who had fonts that were used on gambling websites installed on their laptops were considered higher risk.

These are interesting times where there is more data out there about all of us, than we would like. But a good understanding of how they affect our social and economic standing, helps us beat them.

Arunkumar Krishnakumar is a Fintech thought leader and an investor.

Get fresh daily insights from an amazing team of Fintech thought leaders around the world. Ride the Fintech wave by reading us daily in your email.



  1. This is an interesting post – the use of digital footprint for underwriting consumer loans is an interesting new area of research. But your headline highlights one of the challenges. If credit scoring using online behavior can in fact be “beat”, then it is no use in underwriting loans. The use of any data in credit scoring is only of use if it is truly indicative of the credit risk involved in lending to the borrower. As soon as the data can be gamed to the borrowers advantage, the data is no longer indicative of the underlying credit risk, and thus of no value in underwriting.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.