Bad sources? The risks of alternative data

Bad sources? The risks of alternative data

New ways to capture and package previously inaccessible data have given financial institutions (FIs) a diverse set of methods with which to assess the creditworthiness of corporate and retail customers. Despite the appeal, however, deploying this data does not guarantee clearer credit risk assessments – instead, it may muddy already murky waters.

Alex Davies ([email protected])

Bad Sources Image

Valuable data everywhere…

Lenders have several new non-traditional sources of information to help them assess their potential clients. FIs offering corporate loans, for example, can use commercial satellite imagery, updated daily, to gain fresh insight into a potential borrower’s supply chain and inventory. Using satellite imagery, credit analysts can track factory shipments and mine operations as proxies for determining a debtor’s cashflow. A persistently empty shipment lot or static mining machinery suggest low production and concurrently low revenue. Knowing that a current or prospective debtor is struggling to sell and ship goods should prompt a creditor to manage its risk – by demanding more collateral, for example, or refusing to extend a facility.

Equally, social media and online activity offer retail lenders a similar opportunity to infer a borrower’s creditworthiness. Someone’s lifestyle – posted online – could indicate their likelihood of default. Do posts of them at expensive restaurants reflect their stated income? And with access to a person’s location, tracked via their phone, a lender could verify how often they turn up for work. Most people would consider such information sacrosanct – but an investigation by Vice’s technology arm Motherboard found that some carriers operating in the US (AT&T, T-Mobile and Sprint) have been selling customers’ real-time location data. Meanwhile, The New York Times found multitudes of apps selling location data, in a market that hit $21 billion in ad sales in 2018.

Lenders could examine these data streams to estimate a borrower’s spending habits. By building this dataset over time, they could carry out additional analysis to determine correlations between location patterns and delinquency/default rates that may not be readily apparent.

Hide and seek

As has been much discussed in contemporary media, holding and using such data on individual customers has triggered serious privacy concerns. And FIs should always examine the accuracy of any alternative data they employ to ascertain that it is an adequate proxy for their purpose. Relying on alternative data could in fact muddle credit risk assessments, because of issues around its availability and openness to manipulation (see Table 1).

Table 1: Two main issues with alternative data – availability and manipulation

Two main issues with alternative data - availability and manipulation

Source: Chartis Research

Of course, manipulation of the type described in Table 1 requires intent, capability and a certain moral laxness that few borrowers possess. Nevertheless, the obstacles outlined above all stem from the assumption that observed data reflects real behavior. Trucks are assumed to be carrying shipments; borrowers are assumed to act in similar ways.

Seek and verify

If FIs are incorporating alternative data into credit risk assessments and lending decisions, they must subject it to rigorous validation that examines the completeness and accuracy of datasets. Importantly, FIs should test modeling inputs derived from alternative data against financial data that states variables such as cashflow. Doing so will prove (or disprove) that a given set of alternative data is an accurate proxy for variables of interest in credit risk analysis.

FIs should take alternative data from multiple sources, using and validating them against each other. They should also interrogate the chain of inferences by which they translate observations into analysis. In short, however much alternative data FIs use, they should use it to augment traditional data, not to replace it. 

Further reading

Technology Solutions for Credit Risk 2.0

(May 2018, Chartis Research)

Technology Solutions for Credit Risk 2.0: Vendor Landscape

(December 2019, Chartis Research)

Why more is not always better in the world of credit

(July 2019, Chartis Research)

Points of View are short articles in which members of the Chartis team express their opinions on relevant topics in the risk technology marketplace. Chartis is a trading name of Infopro Digital Services Limited, whose branded publications consist of the opinions of its research analysts and should not be construed as advice.

If you have any comments or queries on Chartis Points of View, you can email the individual author, or email Chartis at [email protected].