Alternative data for credit assessment has been all the rage over the last 5 years. With globally 3 billion individuals without a credit record, lenders are looking at alternative data to penetrate this untapped market. Alternative data is also being used to complement bureau scores to improve credit decisions particularly in data rich markets meaning that alternative data is particularly useful for both credit aware and unaware. Here is a summary of some of the alternative data sources out there:

1. Demographic score – these are your traditional scorecards looks solely at demographic information of the customer and is based on information captured at application. Whilst generic demographic scores can be weak (bureaus sometimes hold a small number of demographic data fields), custom-built scores are stronger.
2. Geo-spatial score – this is a score related to the average credit risk profile of people living in a particular region. You are like your neighbour. These scores are only available in regions with an effective postal-code system or where extensive geo-coding has taken place.
3. Network scoring – A similar concept exists when scoring a network – i.e. through phone records, or home affairs data – linking individuals through family relationships and acquaintances – You are like your connections (homophily).
4. Mobile data score – this score has grown in popularity over the last few years. Consumers load an app on their phone. This app tracks mobile phone behaviour of the customer and those data are used to generate a score. There are two challenges here: data laws (invasion of privacy) and you can only score customers who have installed your app on their smart phone.
5. Online data score – these scores utilise the online behaviour of customers. This can include analysing the websites visited and the time spent on certain pages. In addition, in thin file (where there is no credit data on a consumer) markets, lenders have also utilised an applicant’s social media data (e.g. language used in Instagram, Facebook and Twitter posts, one’s education levels, the social network, products purchased on eBay, etc.). This has been used successfully in markets such as the Philippines, Russia, Mexico and Brazil. However recently introduced regulatory hurdles have meant the ability to use this data has diminished in many markets.
6. Incidental credit data – this includes utility data (available in the US), rental data (property, vehicle and appliances), income information (available in countries like Norway), telco data (incorporating voice and data usage, information about the asset used and airtime recharge data – this has been utilised in countries like Kenya).
7. Psychometric scoring – this incorporates an online questionnaire that an applicant fills in. The questions measure elements of the applicant’s character. These elements are used within a credit scorecard. There are many advantages to this score including the following:
   1. It can be used for consumer, microfinance and SME
   2. It has been shown to be predictive in many markets both established and developing
   3. Unlike many of the other scores, a psychometric score can be kept and utilised to make future decisions on the customer (e.g. marketing and collections). This is due to the principle that a person’s character is unlikely to change.
   4. It has been shown to be consistently predictive
   5. It is universal – meaning everyone can be tested

  Now not all alternative data is created equal. Two useful areas to consider is both the predictiveness of the data and the availability of the data (i.e. roughly what percentage of customers have this data available). Below I’ve created a graph mapping out both indicative availability and indicative predictability. These are just indicative as values do vary for different segments of the population. In terms of alternative data, the most promising is psychometrics with the highest availability (i.e. everyone can be scored with psychometrics if the person takes a test) and with very reasonable Ginis for alternative data scores.

  So which alternative data score is right for you? This depends on multiple things.

1. Where do you want to use the score?
2. Do your country’s privacy laws give you access to the data needed for this particular score?
3. How easy will it be to deploy this score within your given process?
4. Does your country have an established geo-coding system?
5. Are there alternative credit scores available (utilising, e.g. incidental credit data)?
6. Will I be using the score for thin file, or making margin calls on thick file, or both?
7. Do I want a score that I might be able to use multiple times?

We use/have used a variety of alternative data at Principa. If you would like to discuss alternative data, feel free to drop me a line so we can have a discussion. To find out more about Principa’s offerings including our psychometric score, click here.