Some of the most valuable lead data in real estate never touches a spreadsheet. It lives in PDFs: probate filings, court dockets, foreclosure notices, recorded documents. Scanned, stamped, sometimes crooked, occasionally handwritten in the margins.
Counties publish these because they're required to, not because they want them used. And for most teams, that's where the data dies. Someone downloads the PDF, means to read it, and it joins four hundred others in a folder called "to process."
Why a human can't keep up
Reading a filing and typing the estate name, case number, filing date, and address into a spreadsheet takes a few minutes. That's fine for ten documents. It's a full time job for a busy county's weekly output, and it's error prone in a way that hurts later: one transposed digit in a case number and you can never match that record again.
What extraction actually involves
Turning documents into data is a pipeline, not a trick:
- Text extraction. Digital PDFs give up their text easily. Scans need OCR, and county scans test OCR's patience. Quality checking matters here.
- Structure. Every county formats its filings differently, so the parser has to know what a case number, a decedent name, or a property reference looks like in that county's documents specifically.
- Validation. Dates should be dates, case numbers should match the county's pattern, and anything that doesn't parse cleanly gets flagged for a human instead of silently corrupting the dataset.
- Enrichment. A probate filing rarely includes the property address. Matching the estate against parcel and assessor data is what turns a court record into a mailable, callable lead.
The payoff
We've built formatters that do this for probate and court records, and the before and after is stark: a folder of unread PDFs becomes a clean list with names, addresses, and dates, updated on schedule, ready for skip tracing. Data that was technically public but practically invisible becomes a lead source your competitors aren't touching, precisely because it was this annoying to get.
Sitting on a pile of documents you know contains leads? Send us a sample and we'll tell you what's extractable.