p/heron-data We automate document-heavy workflows using AI

Start new thread

Bank Statement Parsing & Fraud Detection - Instantaneous bank statement PDF parsing and fraud detection

Jamie Parker

Heron Data

•

2yr ago

Turn PDF bank statements into parsed transactions + insights in seconds. No human-in-the-loop. Fraud detection. Join 100+ B2B Lenders and Fintechs turning PDFs into cash-based P&Ls with Heron Data.

Replies

Best

Jamie Parker

Heron Data

Maker

Hey Hunters! Jamie here from Heron Data. Heron Data supports over 100 B2B Fintechs and Lenders to power money experiences with bank data. For the last year or so this has been our most common feature request so excited to 🚀 PDF parsing and fraud identification to the community today! Really excited to see what you guys will build. What's new? You can now: 📜 Upload PDFs in our web app / post them to our APIs in addition to JSON data from Plaid or other aggregators 🤓 We'll parse transactions in seconds (no Human-in-the-Loop) for over 90% of bank statements 🕵️ We'll also flag signs of fraud in case anyone has been tampering with the PDF before sending it your way ✅ Once we've confirmed that we have parsed the PDF correctly and fraud isn't present, transactions will be categorised and we'll calculate the spreads you need to underwrite 🤖 If you use Rules we'll tell you in real time whether applicants pass your screening criteria so you can focus 100% of your time on the right cases Heron customers like Torpago, Ramp, and Pipe have increased Revenue by 50%, 25x'd underwriting speed, and 10x'd limits with Heron. If you want to supercharge your underwriters and automate document workflows - book a call on our website or email sales@herondata.io Demo here: shorturl.at/lUY48

Report

2yr ago

Hasan Diwan

Do you use any sort of OCR on the PDF statements? If so, are you planning on providing or do you provide the ML models you use for parsing? If not, is this on your roadmap and can we work on it together?

Report

2yr ago

Jamie Parker

Heron Data

Maker

@hasan_diwan Yep! We use OCR on the PDF statements. This is a mix of in-house and third party solutions. We don't provide the ML models we use for parsing, you just send us the PDFs and we return the JSON data in addition to our categorisation and analytics. Happy to chat ML anytime - it's a core part of our business :)

Report

2yr ago

Roland M

Safely Finance

@jamie_parker this is super interesting. What size business do you think this serves best, or rather how do you like to think about the “SMB” box?

Report

2yr ago

Jamie Parker

Heron Data

Maker

@rolanddeep great question! My thoughts are: - When folks have less than ~$1m annual revenue, bank data is the only reliable data source. Typically these folks don't have high-quality up-to-date accounts. - That being said when your loan size is less than around $10k, you can often underwrite the business owner or follow a more consumer-style underwriting flow. For a loan size of $10-$100k bank data is by far the best (and often the only reliably available) data type. - When you are looking at larger loans - $100k+, you will typically collect additional data (e.g. Commerce, Accounting, etc.) and then bank data is useful for monitoring (easier to stay connected to a bank account than an accounting system) and verifying other data types (accounting data is easy to manipulate; bank data can't be). We make that process 100x easier as you can compare our revenue with accounting revenue. For a median company our revenue estimates match up with around 95% accuracy to a cash-based P&L. So - to answer your question bank data is indispensable to build automation into $10k-$100k unsecured loans but is really useful for any loan size to validate against other data sources.

Report

2yr ago