How Process Automation is evolving with Rule Engine in Fintech?

“Know the rules well so that you can effectively break them” – Dalai Lama

Let’s begin by talking about Rule Engine first:?

What is a Rule Engine??

A rule engine is a system that takes data and rules as input. It will apply those rules to the data and will give us an output based on the rule definition

Rules are easy to understand as compared to other business logic or code. Rules and in their more recent incarnations like (DMN, CMMN, and BPMN) create a bridge between Business Analysts and Developers to understand and implement the business logic.

Automation Challenge in Fintech:?

The business team has identified a market opportunity and wants to make a new product live ASAP.

The only bottleneck is the traditional IT development, testing, release cycle and legal compliance vetting process may run into weeks if not months, especially if it is a financial product.

The message is clear for businesses - embrace automation today or go out of business tomorrow.

Wouldn’t it be great if? IT systems have a nimble rule engine that can take their business rule(s) live in the shortest amount of time while providing businesses with the means to backtest or run simulations on how these rule changes will affect the business in driving growth and helping drive revenue growth?

The answer to all above is a “rule engine”. Rule engines have been around for a while just like car engines and they have been steadily evolving over time. It is for the technology leaders who design the systems to anticipate and put a rule engine in place that makes the business happy by reducing time to market and even better giving them the control to make those changes.

Traditional IT systems and the changes/business rules to be implemented involve a team of developers and testers who work hand in hand with business teams to make things happen.

Ok, What “Rules” are you talking about?

Let’s check out some simple examples from the fintech and e-commerce domains.

• It's the holiday season and anyone who spends more than a certain amount will be eligible for rewards and or free packaging. This is to be triggered only during the period leading to the holiday and to be triggered beyond a certain purchase (order) amount.
• You cannot apply for a loan if are below 18 or 21 years of age.?
• If the customer’s occupation type or industry is in a “red-line” category, he should not be allowed to proceed with applying for a loan.
• Based on the volume, and value of digital transactions (payments on purchases by customers) in a certain time period for a merchant - his qualification and access to a line of credit get changed.

Great, let's detail it some more – let’s pick a domain that’s close to my heart – Digital lending

In the Digital Lending domain “business rules” can be broadly classified as

• Knock Out rules – These are rules to only let a qualified customer move forward as part of the customer journey, these rules may be triggered at various points of the customer/applicant journey.

Examples

• Is the Application rejected in the last ‘x’ days or months? Show appropriate message to the customer and ask him to check back after ‘x’ days/months
• Dedupe rules – Reject or notify a customer if the customer had already registered before and is already present in the system
• If the bureau score below ‘x’ for a certain product reject the customer
• Deviation Rules

These are rules where it is not a clear go (green) vs reject (red), these may need further scrutiny and these cases may go through a higher level of approval workflow

Examples

• The number of unsecured dues is greater than ‘x’ but still less than the knockout value of ‘y’.
• There is a DPD (days past due) on a certain running loan but that is below the rejection criteria.
• Score Carding rules

These are rules where scores are calculated based on customer data that has been provided or retrieved via third parties like bureau reports, bank data, and structured and unstructured data, coupled with weights for the parameters fed by the AI / ML system.

Example?

• For MSMEs: measures of cash-flow based creditworthiness?
• A number of loan enquiries made in the last ‘x’ months.
• Current outstanding loans (secured and unsecured ) along with due amounts and DPDs (days past due)
• Customer demographics like geography, occupation, vintage, customer repayment behaviour on existing loans, customer type (business vs retail, referred by, preapproved customer) etc
• Other business rules

These can be any rules that don’t fit the above criteria

Example?

• send the customer for re-KYC if a certain KYC was done more than 1 year ago
• For certain higher loan amounts, there are additional data needed from the customer and the customer journey has to accommodate capturing these additional data points
• Identify cross-sell opportunities and suggest to customers some value-added services

Provide the customer with offers and reward points (calculation and processing of reward points) based on his interactions and transactions on the platform

How? AI / ML systems complement Rule Engines

• AI / ML systems work hand in hand with rule engines by translating the available data of the customers, predicting certain yet unknown data points of the customer by extrapolating structured and unstructured data gathered about the customer (e.g. similarities with other existing customers bucketed together by demographics).
• AI / ML systems provide actionable insights based on real-time and historical process data, customer vintage and behaviour on your platform, transaction data, social behaviour, insights into smartphone/app usage etc ) gauge future behaviour and come up with tailored financial indicators which are further passed onto the prediction models to come up with weights for various parameters used by rule engines.?

Conclusion

Businesses that can respond faster to emerging scenarios and adapt in the shortest amount of time are the ones that will thrive in the current age and will become “anti-fragile”.?

A good rule engine coupled with dynamic data points and weights provided by your AI/ML system can act as guardrails to protect your business from fraud, bring in the right type of customer and lay the track for you to run your business nimbly.

About the Author

Srinivas Nidumolu (Srini) currently works as Chief Technology Officer – Digital Lending at RapiPay and is upbeat about digital banking, payments, lending, insurance and e-commerce. He has 20 + years of experience spanning companies in the US and India. He is at heart just a curious student.