Finding the seller before the listing exists
Off-market is shifting from relationships to records. Here's the whole engine.
Last issue I promised the off-market sourcing engine. Here's the whole thing: how the shops that consistently buy before the listing exists actually find their sellers, and why that game just stopped belonging to whoever's bought the most broker lunches.
- Sasha Deneux
The Take: off-market is shifting from relationships to records, and the ceiling was always geography
Every acquisitions person eventually says some version of the same sentence: the good ones never hit the market. By the time an OM lands in your inbox, it has usually been priced by a broker whose job is to price it fully and shown to the buyers closer to the deal first. What's left on-market is, by construction, what everyone else passed on or bid up.
The standard answer is relationships. Know every broker in your market and the first look comes to you. That works, and it'll keep working. But it has a ceiling, and the ceiling is geography. Your network is deep where you've done deals. Cross a state line and you're the out-of-town buyer who gets the third call, not the first. If your buy box is something like manufactured housing communities spread across several states, there's no version of lunch that scales to that.
So the operators who consistently buy off-market run a colder play. Instead of waiting for a seller to raise their hand, they build a list of every owner who fits the buy box and ask a better question: which of these owners is most likely to say yes if I call this year? Most of the answer is sitting in public county records, readable by anyone.
Here's what changed. That reading used to swallow an analyst whole. County data is a mess of inconsistent formats, scanned instruments, and entity-name chaos, and that extraction and normalization grind is the work that kept the play stuck at one county. The strategy didn't change. The throughput did. And throughput is what turns a one-county play into a multi-state one.
The Teardown: owner-propensity scoring, end to end
The system is called owner-propensity scoring: score every owner in your buy box 0 to 100 on how likely they are to sell, then work the top of the list. Six signals, each traceable to a public record:
Ownership tenure. The date on the last arm's-length deed at the county recorder. Long holds change the seller's math: depreciation run down, loan near payoff, an owner twenty years older than when they bought. You're hunting the unforced exits.
Loan maturity. Recorded mortgages and deeds of trust. A maturing loan is a forced decision point: refinance, inject equity, or sell. Maturity isn't always stated on the instrument, so estimate a window from the origination date and typical term structures.
Absentee ownership. The assessor's two addresses, property versus tax mailing. When they don't match, the owner is absentee. Distance is management drag, and it makes selling easier to say yes to.
Entity and portfolio structure. The Secretary of State's registry tells you what "Sunrise Holdings LLC" actually is, when it was formed, and who's behind it. A decade-old closed-end fund vehicle is designed to sell. Resolving the entity to a human is what turns a score into a phone call.
Life events and distress. Estate and inter-family transfers, liens, tax delinquency, code violations. Willingness to sell is often fatigue plus a trigger, and this is where triggers surface in public records.
Asset-specific signals. For parks, that's sub-institutional size, private well and septic, no management footprint. For office, business-license churn and permit inactivity.
Weight them into the score. As an example starting point (an opening bid to tune per market, not settled truth): tenure 20 points, maturity window 20, distress flags 20, absentee 15, entity profile 15, asset fit 10. Then band it and attach an action to each band: 80-100 gets a human call this month, 60-79 gets mail and a watch, 40-59 just gets re-scored, under 40 sits in the database until a flag moves it up on its own.
What keeps it alive is a monthly refresh. Re-pull the county layers, re-resolve entities (the same principal can hide behind half a dozen differently-named LLCs across four counties), re-score everyone, and diff against last month. The diff is the product: "these owners crossed into the call tier, and here's the flag that moved each one" is a Monday-morning artifact an acquisitions team will actually use.
Two honest caveats. Every one of these signals also fires on owners who will never sell. The score won't predict any individual; what it does is re-sort a list of thousands so your calls go where the signals stack, and that re-sort is the entire economics of an outreach program. And build compliance in from day one: lawful sources only, scrub numbers against the National Do Not Call Registry, treat TCPA as a hard boundary, and a human approves every single contact. No auto-contact, ever. That's decision-support, not legal advice, so run your program past counsel.
Signal
A family office bought a Los Angeles office building after running the read with AI instead of a broker. (GlobeSt, Mar 11.) Why it matters: the "call my broker" reflex is the exact step this issue is about routing around. When the read on a deal can come from records, the first look stops being something you have to be owed.
AI pilots at CRE firms jumped from 5% to 92% in three years, but only 5% say they've hit most of their program goals. (JLL survey of 1,000+ pros, reported by Bisnow.) Why it matters: buying the tool was never the play. The gap between 92% and 5% is the whole difference between a license and a loop that actually runs every month.
AI has outrun governance in multifamily: only a small share of firms testing or scaling it are ready for real scrutiny. (Grant Thornton, reported by GlobeSt, Jul 8.) Why it matters: an outreach engine without compliance wired in from day one isn't an asset, it's a liability. Build the DNC scrub and the human-approves-every-contact rule in first, not bolted on later.
About 40% of firms over $100M in revenue saw AI cut costs by 10% or less, and only 4% beat 30%. (Bain & Co.) Why it matters: "the technology worked, the value didn't arrive" is what happens when AI gets sprinkled across the org. Point it at one bottleneck, the county-data grind, and the math changes.
From NextAutomation
This scoring engine is exactly the kind of system we build with acquisitions teams inside the AI Team Program: weekly working sessions in your stack, arguing over the weight table with your people until they run the whole loop without us. Want the starting kit first? Reply PROGRAM and we'll send you the AI-Native Team Playbook, our free pack for standing up an AI-native team, or book a call and bring your buy box.
Book a callThe listing is the last step of a sale, not the first. Many of the signals that precede a sale sit in public records long before a listing exists, and anyone can read them.
Next week: the inbound side. Classifying every OM that hits your inbox HOT, WARM, or PASS against your buy box, with the reasoning attached and the broker reply drafted.
- NextAutomation Team