When you’re launching a new product, business line, or starting up a business, you’ve got to move fast and break things. This means taking a minimum viable product (MVP) approach, where you’ve got to sacrifice scalability by implementing manual processes to support the early-stage business. Commonly, a manual process will be in place for credit applications and approvals – pulling the credit report, reviewing the data against a scorecard or policy, and then making the decision. Since this likely takes a day — or often longer — the process decreases your customer’s experience, and can hurt your ability to scale and grow revenue the longer you wait to automate.
To grow the business and take it to the next level, you need to migrate away from the paper-pushing approach. The next step is to move toward an automated solution that integrates credit decisions with the back office, such as an ERP, CRM, or another custom system, employing APIs.
Using an Application Programming Interface (API) to Connect to Your Decision Engine
An API, or Application Programming Interface, is many things. It’s a set of instructions and technical documentation for developers. It’s a collection of services that allow you to interact with a product or service. And it’s a way for businesses to open up and allow for new kinds of innovation – allowing for new business models and application development that wouldn’t be possible without APIs.
In the last decade, APIs have become system agnostic, meaning they plug-and-play into nearly any system because they are standardized and popular amongst the development community.
Because of this popularity, APIs make it easier for the business to get buy-in from the IT department, which is essential to automating the credit decisioning process. Without an API, the IT department must devote significant resources to the project because more infrastructure to host large database will be required. APIs allow you to pull data in real-time only when you need it, reducing system complexity and decreasing application development costs. Reduced complexity also means less risk because you are more assured that your IT department will be successful with the integration. Often, when IT departments are presented with information about the API, their response is “No problem, this is standard. We have integrated with a very similar API before. We can do this.”
How does your decision engine interact with APIs? You can use APIs to get the raw data elements your credit policy or model needs to render a decision, no matter if the data is internal to your business or provided by third parties.
Taking Decisions to the Next Level with Machine Learning
According to a recent Harvard Business Review project, the key to successfully utilizing machine learning isn’t to get caught up in new and exotic algorithms but to make the deployment of machine learning easier. There are many use cases where machine learning can be employed, but use cases where data-driven decisions are being made, as in the credit approval process, are archetypical.
During the early stages of the machine learning process, you train the model by feeding it data from past applications. Then, as you use the engine for real-time processing, the engine learns from past decisions. If the engine was originally approving applications with a borderline credit score but found that these applications often ended up being poor risks, the model would then begin turning down these applications.
The key ingredient in making machine learning start to work for your credit department is to have domain experts, credit managers, help the IT department focus on the key variables that can help the machine learning model to predict key outcomes – credit losses, bankruptcies, and business failures, and to put the models through many rounds of testing and validation before putting them into real-life practice.
Now is the time to move your manual processes online using an API and machine learning. According to Mary Meeker’s Annual Internet Trend Report, 60 percent of customers pay digitally compared to 40 percent in the store. And it’s likely that the gap will continue to grow. The longer you wait, the further ahead your competitors will be in digitizing the customer experience — and the harder it will be to regain your footing and catch up.
