Now Reading
CRIF’s Basic Guide To Usage Of Policy Rules: Old Wine In A New Glass

CRIF


Macro-Economy and Micro-Policies

After the NBFC debacle, the Indian economy has seen liquidity drying up at a rapid pace. The lending industry has been affected so severely that credit growth has reduced to half by the end of September 2019 compared to what it was at the beginning of the year, as per the latest RBI data. While some significant non-banking financial companies have witnessed diminished revenues, others have ceased to exist.

To understand how general credit policy rules affect the lending industry, we got in touch with Atrideb Basu, Senior Vice President, Products & Consulting at CRIF India. In an insightful interaction, Basu elaborated how current lending policy rules are doing more harm than benefiting the industry, along with what can be done to improve the situation.



“The concept of policy rules, in general, is very intriguing. It looks at a specific dimension for a customer, and basis that takes a binary decision (approve/decline) essentially to protect the lender from defaulters. So, for example, if the CRIF credit score is less than 700, a lender might decide not to give applicant the loan. Other dimensions that are usually used by lenders are income levels, indebtedness measures, applicant age, past delinquency, number of loans taken in the past, etc. The usage of policies by all the institutions typically rejects around 15%-25% of applicants – and it is not a small number. In a decelerating economy, modifying and managing these rules becomes mandatory” elaborated Atrideb Basu- SVP, Products and Consulting at CRIF India.

The Problem With Current Policy Rules Across Lending Institutions

Experts say that different conditions of regional economies and the significant fluctuation in qualities of individual loans disqualify the foundation of standard lending policies. These issues, nonetheless, should be tended to uniquely in each lending approach. Here, the onus is on an institution’s leadership to decide if the lending policies are reasonable and accommodating the various factors for continuous growth.

Atrideb Basu from CRIF explained, “The usage of policies created are very standard – essentially you create a rule for each dimension of risk – assess the volume impact and possible bad rates – and then use them as the first layer of assessment in the underwriting process. Thus if 100 customers apply for a loan, the policy rules can reject 15-25 customers. If our acquisition cost is INR 10 per customer, then INR 150-INR 250 goes down the drain.”

Also, organisations which are in operation for a long time have so many rules that sometimes it is challenging to assess which version of a policy rule is running, and for which segment and from when.

“The issue here is if you reject a bunch of customers via a rule, you do not have the performance to measure the same. Hence you must partner with a credit bureau to assess if they had taken a loan when you rejected, and how did they do on that particular loan. Once we get this data, this allows us to know if we need to have a policy rule in place,” Basu said.

Atrideb Basu also pointed to the fact that there are so many innovations at hand which can be deployed to maximise lending efficiency. While effective AI models can be built on the massive volume of data generated from the lending industry, there is not enough which is being done on this front as well.

“Hardly I have heard or seen innovations being applied to this area. We talk about fancy things like big data, alternate data, machine learning – but none of it seems to be enhancing this arena of policy rules. Imagine we use something every day, and we do not make it better. I mean the kind of toothbrush we used ten years ago, we do not use the same kind today. So we have to evolve what has been in use forever to make it better in what it does,” Basu added.

The Fine Balance Between Ideal Risk Management And Optimising Customer Value For The Lending Industry

Proper policy rules can help enormous portions of India’s population which have been turned down in the past to have first-time access to credit. In the context of India, the better credit access individuals have, the more and quicker growth the economy can accomplish. The equivalent applies to banks, as they additionally have the advantages of both connecting with smaller fragments of the population through better evaluations, as well as take a cautionary approach for people whose loans are turning non-performing.

The problem does not only pertain to the underwriting side but has multiple use cases on customer management – that is for customers who have existing loans or who are applying for a greater number of loans. Most of the banking systems today also have a bunch of rules for loan amount assignment, limit increases or decreases, collection rules which need to be managed with a newer philosophy, experts tout.

“The question from a business standpoint is – what can we do effectively – that stops this wastage of marketing money and protect the credit risk. There is no alternative than to get a system that manages such rules in production, and the analytical environment capable of mimicking that to evaluate the volume impacts. It is critical to comprehend what is happening in your process. This includes how many rules are running, what is the volume impact for each rule and finally the impact to bad rate – quantity and quality. It seems like an effortless task – but it is not,” said Basu.

Recommendations To Lending Institutions On Policy Rules

There are some things that CRIF recommends. The company already has implemented innovative approaches with players in India and outside to make lending firms more competitive. CRIF combines the following five aspects to define the next level of policy usage – create a structure that has more analytical readiness. Here is what Atrideb Basu says lending institutions can do based on what is wrong with current policy rules.

See Also

First, we must attack the structure of the way these rules are set-up. Most of these rules are hierarchical. To illustrate Rule 1 rejects 10%, then 90% of the population moves to Rule 2. Rule 2 then rejects 5% and then 85% moves to Rule 3, etc and so on. This is an incorrect structure. All the rules should be independent – so that when we look at the monthly impact, we can fathom that 5% population hits five rules, or 7% population hits three rules etc. This allows us to understand the full coverage we get from these rules.

Second, if we think of the 10-15 policy rules as all binary variables and use a decision tree algorithm/logistic scorecard to determine the importance of each rule, it is a direct validation of how important each rule is.

Third, we get a 360-degree feedback on each rule from all departments in the institution using independent view and general best practices followed in the industry. The feedback will give a qualitative approach of the perceived value for each rule, coming from intra-company and inter-company to create necessary measures.

Fourth, the introduction of newer kinds of rules is also necessary. These can form from either new sources of data or extracting more information from existing sources. Whatever be the case, we should introduce a few new rules every year and sunset a few others, depending on performance.

Lastly, the system should allow robust test-control mechanics. CRIF’s decision engine have perfected this – where it can simulate what will be the change impact if one relaxes or tightens an existing rule. Also if you do not do this, you will never really know. If you as a credit manager are sure you should not lend if the customers Fixed Obligation to Income Ratio (FOIR) is more than 50%, unless you open the gates till 60%, you will never know the incremental impact of this 10%.

“I am sure this will make policies sexy again – and the business will gain significant competitive advantages by doing this,” exclaimed Atrideb Basu.


Enjoyed this story? Join our Telegram group. And be part of an engaging community.

Provide your comments below

comments

What's Your Reaction?
Excited
4
Happy
3
In Love
2
Not Sure
0
Silly
0
Scroll To Top