How To Lend To Businesses Using Bank Data From Plaid
We started Heron Data because we learned while building our own lending FinTech that while less than 10% of companies have high quality, up-to-date accounting data, every single business has up-to-date bank data. Bank transaction data is also the single source of truth about what happens in a company. It's great for joining other datasets to (like Commerce data) and even for verifying accounting data because it can't be easily doctored.
What's more, it's easy to pull bank data via API with Plaid. But how do you go from the raw transactions to a cash-based P&L statement that you can use to assess an applicant's financial health?
You could start by building a set of categories that match up to the sections of a cashflow statement, labelling a a dataset of tens of millions of transactions with the relevant category, then training ML models to categorise in real time with a high degree of accuracy and coverage. You would probably also want to ship a dashboard to visualise companies that you want to dig into, recategorise transactions, and then continue to add training data and retrain your ML to avoid model drift and keep up with new vendors.
You could do all that, or you could use Heron Data. Book a call to hear more.