Blog article
See all stories »

Is credit scoring the next part of the lending value chain for disruption?


Well it’s definitely ready and the data and organisations to support you are established, so here’s what we know…

Traditional Scoring Methods/Data

The UK market typically engages with 1 or more of the 3 stand-alone independent bureaux, supplying historic voter’s role and credit payment data for inclusion within credit decisioning scorecards. Data is supplied by participating lenders, usually through monthly updates to the credit bureaux. Real time updates were initially mandated within the payday lending market, but most lenders rely on information that was provided up to 30 days ago. 

Bureaux have invested in additional data sources to extend their coverage of validated application data and undertaken cleansing exercises in order to offer improved accuracy for users. The general standard of coverage and quality of data is high and the bureaux will continue to be the main source of data for credit applications for the foreseeable future. The type of data that they can provide has also increased as lenders request more varied data points to include during a credit assessment. These include current account turnover information that allows lenders to see the typical income and deliver confidence levels against the applicant’s declared income. Further confidence can be achieved when validating identification by presenting applicants with questions related to credit bureau information. Applicants are required to pass a defined number of questions and this is increasingly the method for online ID validation. Whilst this process for ID verification is widely used, we can only estimate that the volume of drop outs from this process could be improved with a fully automated process that doesn’t require applicants to input information.

So the credit bureaux have operated in the UK for in excess of 30 years and produced a strong and stable system to facilitate credit application processing. And whilst this is an effective solution, THERE HAS BEEN LITTLE, IF ANY, SIGNIFICANT ADVANCEMENT OF THE DATA SETS OR DRIVE TO LEVERAGE NEW DATA SOURCES. This isn’t down to inertia on the part of the credit bureaux as most of the developing data sources are known to them and some are in dialogue. The point is that adoption is slow on the part of the lenders, but we think that this is about to change.

Structured Data Sources

Available through numerous third party vendors such as GB Group and Trulioo, who have utilised the data for ID and verification. Structured data sources are supplied by utility and mobile phone suppliers and are used by government organisations such as the Police and HMRC. At the present time this data is not utilised in credit applications, but the sources could be used to enhance traditional assessments. Data such as mobile phone number reverse look ups, to confirm addresses and type of contract, can demonstrate stability and extend the potential assessment of applicants without previous credit history or new to country. A mobile number can also be ‘pinged’ in order to ascertain where the phone is currently located – an inexpensive basic check, but also a very strong fraud indicator.

Additional data services such as insurance policy at the applicant’s address may also demonstrate stability and whilst not sufficient in isolation, they could represent new data that could be included within a scorecard.

Examples of other structured data sources that demonstrate stability or habitation include:

  • utility meter numbers which are personalised to specific users,
  • email – highlighting the type of email account such as webmail or domain. Address type such as enquiry, work or personalised email

All lenders will tell you that one of the best assessments of an applicant’s financial health is through the current account. The ability to view not only credits, but outbound transaction volume/velocity and recipient will deliver a good insight into the current risk associated with the applicant’s ability to make scheduled payments and PSD2 has opended up access to this data. The credit bureaux and new third parties such as Yodlee, MiiCard, Kontomatik and CreditKudos have jumped on the ability to extract data directly from an applicant’s online banking view, subject to authorised access from the applicant themselves. The crucial element is the ability to deconstruct the online view of bank statements, segregating credits and transactions into data that can be used to accurately build income and expenditure. MCC codes are typically used for debit transactions but a high level views of expense types and there isn’t a common view on categories that are more granular and that can be consumed by scorecards. So this is a solution that still has to mature…

Unstructured Data Sources

Hello Soda, Big Data Scoring and Friendly Score are pioneers of using social media data to analyse online profiles to produce a supplementary view of a customer’s financial stability. For example, use of words that suggest poor financial health such as “skint” are combined with other extracted key words in order to build a profile of recent activities. Whilst this data is unlikely to be used in isolation to make a credit judgement, it does supplement a thin credit file to help lenders make a decision. The level of analytics and also the fact that applicants have to volunteer access to each social media account again limits the current performance, but social media behaviour will no doubt become more integrated into credit assessments over the coming years.

New Data Sources

One company has taken, what is with hindsight, a very logical view to providing more data to lenders. Aire Labs have built a supplementary proposal that asks applicants their attitude towards jobs, finance and most importantly how they would react to a pressurised debt situation. These questions are presented in a non-linear way, according to the previous answer and build up a strong profile of the applicant which can be used to supplement a credit application. The relevance of this data is that it’s substantiated by analytical behaviour models and more importantly the input is in effect a real time view of the applicant’s current attitude.

What is also interesting is that applicants are willing to provide additional data in order to support a lender making a credit decision, with over 85% inputting further data, when initially referred. Acceptance rates have increased by over 10% in trials, which are customer that previously would have been turned down, usually because they had no credit history or were new to the UK. So this product ticks all the boxes for inclusiveness and increasing revenues. Again though as with all new solutions, the results of the delinquency performance will make or break this type of solution.


Like most lending technology businesses, Nostrum has the ability to integrate any third party solutions into our digital on boarding process. The skill is inserting the data capture at the right point in the process. Some of the vendors described above could be used at the front end of an application prior to passing data to a Bureaux and thus reducing costs by eliminating unnecessary calls to a Bureaux where its proven that an applicant would be declined. Others would be better placed post Bureaux call, in order to supplement the CRA data to make a better informed decision. For all of the solutions, they would be integrated into the process through API’s, passing data in real-time, in order to provide a quick decision, whichever way that ultimately goes.

So what?

Continue with your proven scorecards and data sources and you will continue to accept the same types of profiles and produce the same delinquency performance. But the population is changing; we see that every day, with the proliferation of smart phone usage and communication through social media. The world has changed significantly in the last 30 years and the vast amount of data that is available for your personal profile is your IP. Personally I invest time and effort in LinkedIN for my current role, but if I look at the roles I’ve had, the time served in each job and the progression would provide more information than you would get from me, if I applied for a loan. So why isn’t it used? The same goes for attitudinal analysis. If you have a customer that is 21 years old, they may have absolutely no credit history, but be prolific users of social media, from which data can be extracted to add substance to the scorecard. In this example the 21 year old may well have been declined for credit because of a lack of data to validate, so these new sources deliver improved ability to offer credit, in-line with the FCA’s drive on inclusiveness. The same applies to migrants entering the UK, where little data exists on their current situation. However, most will take out a mobile phone contract on arrival, and the capability to validate the registered address could potentially be crucial in assessing them for credit.In summary, we’d say that this wave of new data is here now. Some lenders are trialling already with substantial improvements in accepts rates and as first movers, they will steal a march on their competitors. You’d predict that the main retail banks would be slow to react to this new data; that said they are all regular attendees at fintech events, so they may prove the market wrong.  Suffice to say that the market needs to shift – improved digitisation of credit applications and automation will only partially work in the medium to long term without integration into these new data providers. I foresee short term consolidation of these providers by the existing credit bureaux (although they do have form for acquisition that will aid their dominance), because each provider has different data capability which may be more relevant for specific credit types or sectors.  

The data sources are also evolving and for the foreseeable future. One thing is for sure, that education and awareness regarding individuals’ on-line profiles should be provided by schools. Personal data that can be publicly consumed will make up your profile, so exercise caution when posting comments that could be taken out of context, as it may affect your future credit application.

This blog was written by Richard Sunman, who is Head of Commercial at Nostrum Group. Nostrum Group builds digital lending solutions, and was named in the Deloitte Technology Fast 500 & Sunday Times Tech Track 100.  




Comments: (6)

Matt Schofield
Matt Schofield - Credit Kudos - London 30 March, 2016, 15:001 like 1 like

For us, the key missing ingredient in credit scoring is consumer involvement. By engaging a consumer in their outcomes, you provide the necessary transparency demanded by modern borrowers.

Konstantin Rabin
Konstantin Rabin - Kontomatik - Warsaw 31 March, 2016, 11:051 like 1 like

Banking data is a jet fuel for any company that wants to issue loans. Such data can be used for verification of the provided information as well as indepth analysis of client's financial health. The thing is, for years banking data was nearly impossible to obtain. Today Kontomatik supplies it within 12 seconds or less. 

Gerard Hergenroeder
Gerard Hergenroeder - Payments Shark - Millersvile 04 April, 2016, 14:02Be the first to give this comment the thumbs up 0 likes

Credit scoring can be improved by adding unstructured data from social media and telephone records along with sophisticated algorithms to add another dimension to traditional scoring techniques. Think about building an "integrity index" to complement a FICO score and standard credit reports. This takes us back to the days when bankers looked at applicants straight in their eyes. Technology can bring back the human element back to credit scoring.

Ketharaman Swaminathan
Ketharaman Swaminathan - GTM360 Marketing Solutions - Pune 04 April, 2016, 19:27Be the first to give this comment the thumbs up 0 likes

It's a misconception that banks use only credit scores to decide whether to grant a loan or not and that there's a whole set of other data that they're not using today. For example, they already use income statements and tax returns. I agree that they can use even more data in the lending process but the question is, why fix it if it ain't broken? What's the compelling reason / ROI for doing that? It's not as though the lending process is badly broken today - if it were, banks wouldn't the most profitable industry in FORTUNE 500.

A Finextra member
A Finextra member 05 April, 2016, 06:54Be the first to give this comment the thumbs up 0 likes The issue is that the usual mass population is changing. Applicants are perhaps migrant workers, ex military and just folks without any credit history. Currently they are declined because mainstream lenders using only standard bureau have no ability. So whilst the process isn't broken, it must evolve to deal with changes in application profile and improved data sources. Inertia will eventually catch up and open the door for new entrants with sophisticated solutions to grab market share.
Ketharaman Swaminathan
Ketharaman Swaminathan - GTM360 Marketing Solutions - Pune 05 April, 2016, 10:19Be the first to give this comment the thumbs up 0 likes

ICYMI, in the run up to the GFC, banks found a way to make money from lending to  the "thin file" and "no file" segments that you're talking about. And they went ahead by taking a business decision to lend to people with 540 FICO instead of 750+ FICO. Their existing lending systems didn't hamstring them in any way. They didn't need unconventional data sources like social graph. 

If banks are declining loans to these segments today, it's perhaps because they don't find them profitable, which is again a business decision and not a technology constraint, as I'd highlighted in Are Banks Losing Customers Or Shedding Customers?  

Maybe it's only me but finsurgents exhibit greater inertia by mixing up business decisions for technology constraints, whether it was regarding branch, cash and plastic in the past or lending now.

PS: I hear that the so-called sophisticated solutions you're talking about are actually selling their loans to investment banks, who're packaging them as bonds and selling them ahead for fat fees. So, maybe banks have already found a way to profit from these segments, just not by lending to them directly. Again, they didn't need unconventional data sources like social graph.