Summary

• Credibly removes a time-consuming and variable process by automating bank statement reviews
• Automation allowed Credibly to 4x the number of applications reviewed with a reduction in review time

‍

Credibly has positioned itself as a leader in the small business financing space by leveraging innovative approaches to risk assessment. Unlike many other financing providers, Credibly factors in data points like business cash flow to approve businesses with lower FICO scores. For example, a younger business may not have the strongest FICO score, but business performance can be a good indicator as to whether they are succeeding and able to make remittances on their financing.

A key part of gauging business performance is looking at their recent bank statements to understand whether they’ve been able to maintain healthy and consistent cash flow. However, bank statements can have hundreds if not thousands of transactions on them, which can be time consuming and inconsistent to analyze manually.

In 2023, to support the growing number of applications for financing from Credibly and to improve the speed to decision, Credibly built a model to automatically analyze the information in bank statements and help the underwriting team understand the merchant’s cash flow.

How it works

Automated transaction classification

The model first looks to classify transactions in a bank statement by looking at the transaction descriptions or memos. This allows Credibly to extract information and decline any files that do not qualify at the beginning of the process– For example, if a merchant has too many instances of insufficient funds, this often indicates that they don’t have enough revenue going into their accounts to qualify.

The model then uses a set of predetermined criteria to find the true business cash flow. For example, cash deposits can also include transfers between the owner’s accounts or cash back offers, which would not be true business revenue.

Identifying existing financing obligations

If the file is eligible to continue on in the process, the model then looks for any indicators that a business might have existing financing such as a merchant cash advance, which is not typically reported elsewhere. The model compares the transaction descriptions to a list of known providers to flag any matches.

One advantage of completing this process manually was that underwriters could learn about new MCA providers that appeared on statements. To maintain this advantage without losing the benefits of automations, Credibly’s Data Science team trained the model to identify transaction patterns that are typical when a business has an MCA. This could look like a lump sum deposit followed by a regular cadence of withdrawals. After running a couple verifications, the model will flag any potential MCAs to a human reviewer who can then verify and add the new provider to the existing list. This allows Credibly to continuously update the model without significant human interactions.

Finally, any transactions that are unable to be confidently classified by the model are sent to  underwriters for manual review.

Scaling with automation

Automating the analysis of bank statements has allowed Credibly to scale up operations while also reducing the time from submission to offer. Since implementing this automation, Credibly has been able to 4x the number of applications processed while reducing average submission to offer time to 4 business hours.

Expanding automation for more

Bank statements and the transactions in them can vary widely and Credibly is continuing to look for ways to reduce the number of files that require manual review so more businesses can experience faster decisions on their applications.