The next generation of Credit Bureaus

Anyone working in the credit industry in London is well aware of the dominance that the three major companies have in the market. Joining their services and working with them has always, and still is, a gruelling experience. That’s not to say that the people working there have bad intentions, or deliberately make things hard, on the contrary, they are usually great people trying to do their jobs supporting old legacy systems and way of doing business.

Already we can see a few startups that are entering the market, trying to provide an alternative for the traditional credit reference agencies. From Aire Labs, that uses personality questionnaires of sort to evaluate a person’s willingness to pay debt, to Credit Kudos who uses bank transactions to evaluate a person’s affordability.

 However interesting these new entrants are, to the best of my knowledge, they are still not used as a primary credit reference for credit providers. Instead they are used as a secondary reference in cases where the primary reference failed to help make the final decision.

I thought I would share my thoughts on what would make me choose a challenger credit reference agency as a primary source over the incumbents. Some of these are quite simple while other are more complicated to implement.

 I divided these requirements to two sections. The first is a set of requirements from data perspective and the second is a set of technical/operational requirements.

From data availability perspective:

Using bank transactions

The current implementation of credit information revolves around credit providers reporting to the credit bureaus about credit accounts they own, and high-level bank account turnover (though not all bank accounts).

 The problems with the self-reporting from credit providers are in the accuracy and time delay of these reports. More than once I heard from employees of credit providers about how they are running “projects” to fix their credit reporting processes. While the fixes might be minor in most cases, it still shows the weakness of relying on that data.

 The other issue is that these reports are updated once a month, not always on time. This means that when one credit provider processes an application, they might not be aware that the applicant has open credit accounts that weren’t reported yet.

 Using bank transaction data solves most of these issues. If the loan repayments are correctly identified from the transactions, it can reveal a lot about the applicant:

• Monthly repayment amount

• Missed payments

• Newly acquired loans

But that’s not all we can learn from bank transactions. 

 Instead of using the account turnover data that the bureaus use, the credit provider can discern between fixed expenses and variant expenses, or between different types of income sources.

 From the applicant income, the employer can be identified as well, which can be used for assumptions around income stability, company or industry stability. These can be factored in the risk assessment.

 Discerning between fixed and variant expenses is also very helpful when trying to assess how much debt the applicant can actually serve.

Show the funnel of a customer applications

This is where things can get really interesting. 

When a credit applicant is rejected by a credit provider, they usually continue applying with other providers, without being able to estimate the chances for successful application.

Now imagine if the future credit bureau could monitor the application funnel and monitor where applicants apply after the rejection and where do they actually get their credit at the end.

In this scenario, they could perhaps create an ecosystem, where if the applicant is refused with one credit provider is then directed to another, where the chances for approval are higher. Then, if indeed the application is accepted, the bureau and the original credit provider can share the referral fees from the accepting credit provider.

This eco-system can improve the customer experience by reducing the customer frustration and providing a more suitable credit product. It can also reduce acquisition costs for all lenders.

Automated data acquisition

I will demonstrate it by an example:

A while ago I tried to open a business account, for a business I was involved in. For that purpose I had to provide the personal details of the directors.

That information was used by the bank for KYC with a leading credit bureau, and the KYC failed because the name of one of the directors was misspelled in the credit bureau database.

I filed a complaint with the bureau (don’t get me started on how long it took me to figure out how to do that) and after several months, they got back to me saying that the error was caused by someone manually entering the director’s name into the system.

Do I need to say more??

From technical perspective

Simple on boarding

My experience with credit bureaus is that the on-boarding process is always long and expensive, having many people involved in the process.

The new generation of credit bureaus, using more modern technologies, should be able to spin up new environment faster and more efficiently than the incumbents. 

This means that new lenders could on-board without paying high upfront costs, and likely much cheaper prices per search. This technical efficiency could also mean shorter approval process for new lenders.

Sandbox environment with real anonymised data

 Today it’s nearly impossible for a new lender to compare between the different credit bureaus. In order for a company to get access to a testing environment, they would need to go through the entire on-boarding process, which is usually very long. This means that in most cases they would prefer to choose one to start the process with, instead of testing several.

The technical ability to spin up a testing environment for new lenders will help them make a better, more informed decision.

Machine learning

Machine learning can be used to help find patterns in the data and better refine the data that can be provided to lenders.

In conclusion, the credit referencing sector is crying for innovation, and I’m full of hope that the challengers in the market will be able to dominate in the future.

With Open Banking, new technologies and machine learning, the opportunity to innovate is greater than ever.

I’ll be happy to hear other ideas and thoughts about how this industry could improve.