Credit Decision Engine: Do's and Don'ts

If you are reading this article, you are probably looking into whether to build your own credit decision engine or rent one that’s already out there. These decision points however trivial need a closer examination because it could speed up or delay your product launch.

We talked to many banks, fintech partners and financial institutions and in fact most of them don’t need a full blown, expensive and hard to manage decision engine. Especially when they are a stand alone product that requires hiring expensive consultants and monthly fees and additional costs for integration.

Most of the fintech startups and bank partners just need a quick and fast credit decision engine that sucks in data points, runs a few simple rules and integrates into the rest of the enterprise ecosystem and frees up the time and resources for branding and marketing.

We will go through some of the Do’s and Don’ts when considering building or renting a decision engine. But before that, we have to ask some of these tough questions. 

Why do you need a credit decision engine?

So why do you need a credit decision engine? If you are running just a few if-then-else logic as part of your underwriting process, then our suggestion is that you “hard code” some of these rules right into your web pages and spend all of that time and resources on customer service. 

Most of the clients are only dealing with a single source of data. The most relevant dataset for underwriting is probably a national credit bureau. If you’re onboarding clients that have no credit history and your proposition is to build credit for them, then you may not need a full blown $10,000 a month system that runs a handful of rules.

If you are in the business of point of sale financing where a lot of customer interaction is involved with the retail front, you may not need a complex decision engine because most of the verification and assessment are probably made in person. Your website might just need to take in the customer application and send them on their merry ways to a retail store to get the product and services. Spending hundreds of thousands of dollars on a fancy over engineered decision engine will be a complete waste of your resources.

If you are running a complex, multi-staged decision tree and that it also ties into a product offering matrix, you might consider a credit decisioning engine that might be able to help you to accomplish your goals. But there are some features that you must consider before signing up for a decision engine.

Who actually needs a credit decision engine?

In the previous section, we alluded to who actually needs a decision engine. Well, if you are running a pure online business where you don’t get to interact with your customers, we believe you might need to use a decision engine. You will be required to verify the customer’s identity and take more fraud risk than a retail store or a point of sale financing situation. 

When an applicant fills out their application, that personal identifiable information of PII is then sent to a few different places to positively identify this person's identity. From verifying their email address, phone number, address and national ID, to scanning of the customer’s face and other out-of-wallet information to verify the identity, only a decent decision engine can keep track of all of this information and inform you whether or not to proceed with the applicant. 

If you are thinking about using multiple data sources such as banking transaction data and credit bureau data to make a comprehensive decision, you might want to consider a credit decision engine. It will be way too complicated to hard code some of these third party integrations and run complex rules across different data assets and having to normalize the data structure across a variety of third party vendors.

Lastly, if you are thinking about risk based pricing (skip to the next section if you don’t want this is), then you probably would like to build your product matrix into a product decision engine attached to a credit decision engine to configure risk based product assignment rules into a credit decisioning suite to keep track of it all.

By all means the above are the only examples and scenarios that might make you consider using a fully built out credit decision engine. There are many other reasons to use a credit decision engine to keep track of all of the decisions made, for compliance and legal reasons.

Credit Decision Engine - Do’s

Whether you are building, designing or renting a credit decision engine, you must consider the following. 

• Third party integration: to start writing rules or run your custom credit score, you must have all of the ingredients from third party data providers ready to go. Sometimes these third party data providers are difficult to work with both from a financial perspective and from a technical perspective. Accessing sensitive data such as credit bureau data requires you to have licenses, insurance and a tech infrastructure to be allowed in. You might also consider the time and engineering resources needed to connect with these third parties, it could be months before you get the integration done and it’s not the end of it, it’s just the beginning.

• Third party data assets: Sometimes, you spend months focusing on tunneling through the third party’s API gateway and only to find yourself docked onto the wrong product. JSON or XML responses (don’t worry about what I just said, read on) don’t return the attributes, scores that were sold to you by the salesperson. It’s maddening and requires weeks and if not months to get the right product linked to your account. By now you may have missed a few timelines. If you have all your ducks in a row, you have to create a database architecture that stores all of this information and makes it available for the fun part, rule writing. 

• Rules editor: If you are an engineer, you might think to yourself why do I need a rules editor? I can write all of these requirements in code. Unless you want to be constantly bugged by product managers and risk analysts to change rules and product settings, you should help yourself now and build a graphical interface for these folks to program these rules themselves. A rules editor should have a decent variable library so your risk folks can go in and write rules, adjust parameters. You must also have a way to version these rules, test these rules and so on. This rules editor module might take a year to build and you might have the time to plan ahead. If you revert back to your original thinking of hard coding, you will be blamed for any errors. It’s not a good feeling. If you are reading this as a product owner or a risk manager, you must demand a rules editor where you can lay out multiple layers of if/then/else logic, use your favorite scores and set parameters to approve and decline folks at a minimum. As a product owner, you must be able to program all of the product rules such as price, amount, duration, fees and other product parameters so an approved individual or a small business owner can see these product parameters as part of their digital journey.

• Compliance document store: A decent credit decision engine must have some interaction with the digital journey other than sucking in third party datasets. Some of these interactions may include uploaded documentation, consent documents and e-signed offer agreements. These uploaded documents should be scanned through OCR (Optical Character Recognition) and the important information off of these scanned documents should be used to compare applicant data that was supplied during the customer onboarding process. I know it’s a lot, but these are the minimum set of actions a decent credit decision engine should have to operate an online finance business. Whether you are a bank of financial services company, tying your credit decision engine with your application flow is a must.

There are so many other Do’s that you must consider but the aforementioned few pointers might help you to design your own decision engine or ask for these features when you are shopping around for a credit decision engine. Now, let’s talk about some Don’ts when building or picking a decision engine.

Credit Decision Engine - Dont’s

There are a ton of offerings out there, some claim to have A.I. and some claims to have automation. But you have to ask yourself the original question posed in the article, do you need all of those bells and whistles, assuming that they are real to begin with. It could very well be marketing schemes to hike up prices or sell you features that are irrelevant to launch your first product. So here we go with some of those Don’ts.

• Decision Engine Price: Any software, not just decision engines that costs more than $1,000 a month needs some careful examination. There are many good offerings out there that’s trustworthy and backed by teams that’ve been working in the industry for decades. Some of these teams have put together a very robust platform and offer their solutions at a very competitive price tag. However the old adage still applies, you get what you pay for. The best way to try these things out is to ask for a 30 day free trial. Play around with all the items mentioned in the Do’s section and ask questions. At a very minimum, ask for video recordings of their demo and ask for a live demo. This will save a lot of time and effort when shopping for this type of software. Lastly, if they are still stiffing you for a $10,000 a month bill and you really like what you see, ask to talk to a few clients of theirs and see what are some of the pitfalls you might encounter. Lastly, if they are asking for some sort of onboarding fee, run. Don’t look back.

• Support staff: Most of the decision engines available today are not plug and play and require a massive amount of integration from you. If you don’t have resources or the knowledge to deal with messy integrations to your website, it could be very costly. Some of these firms will have a support model where the software is costly and the support model (hourly rates) is even worse. Your entire budget for the product will get sucked into dealing with a standalone decision engine.

• Enterprise connectivity: One of the biggest don’ts is don’t sign up for a platform or build anything of your own that doesn’t have an open API architecture. The credit decision engine is the heartbeat of your entire finance tech stack and it must have the ability to connect to the rest of the enterprise system. For instance, if the application is becoming unverifiable and needing additional documentation, the credit decision engine must be able to talk to your communications system to send out emails and sms to alert the end client on what to do next and alert the back office staff to be prepared to examine document uploaded through the client portal. The decision engine also need to be able to call other loan, credit card or bank core systems to properly onboard a client when their application has been approved. From time to time, if the bank core and credit card management system produces additional documents or statements, the decision engine should be able to route that information to the appropriate client portal and let the clients know that there’s additional information that needs to be reviewed.

• Basic functionalities: Their credit decision engine also must be able to keep track of all the underwriting rule versions ever produced by your risk management team. Form time to time, these credit risk rules may need to be modified in accordance with the business model and older applications must have the older underwriting rules attached. And newer applications will use the newly edited rules moving forward. Besides compliance and legal reasons for keeping different versions of the rules with the right application, your data analytics team and risk team will thank you for doing that. When they conduct analysis with the right version of the underwriting rules along with the right application is the only way to tell whether the new set of underwriting is working as desired. If you confuse or override the new rules with the old credit underwriting rules, no one can properly assess whether the new underwriting is working as designed.

A few parting words when choosing a Credit Decision Engine

When it comes to picking and choosing an approach of building or implementing a new credit decision engine, take your time and balance features with pricing or cost of development. If you have extensive banking knowledge and your product doesn’t require a tremendous amount of rules, perhaps building something in house is the right choice.

However, if you have a bank partner, or working in a highly regulated environment where compliance, underwriting processes and procedures are utmost important, we suggest you shop around and compare your needs and features that comes with some of the off the shelf offerings. The most important aspect of picking the right partner is the partners themselves. You want to work with someone that has had prior experience in banking, finance, technology and areas of enterprise software development space.

For now, these are some of the Do’s and Don’ts we collected over the years on some of the fundamental items to look for in a credit decision engine. We will leave you with some examples of graphical interfaces for credit decision engines below.