Every weeknight, DealHound sweeps every parcel in NYC, cross-references the public record for signs of distress, and surfaces the few hundred where a motivated owner is statistically hiding. Off-market by definition. The phone work is the deal — the screen finds who to call.
01 · Universe
~33k parcelsEvery lot in the target zips, pulled monthly from NYC PLUTO.
02 · Buy-box
~28k filteredOnly buildable building-class codes in the buyer's buy-box. The rest are noise.
03 · Signals
100k+ eventsDOB + ECB + HPD violations, tax-lien candidates, permit silence, FAR slack.
04 · Worth a call
~117 flaggedScore ≥ 35 and modeled spread ≥ $100K. The working board.
Who it's for
Not a SaaS. A private engine wired between the operator who finds the deals and the closer who calls the owners.
Works the phones. Walks the houses. Decides what's real. Has been finding distressed NYC deals for years on instinct — DealHound just sharpens where the phone gets pointed.
"The score finds motivated owner candidates. The phone converts them."
Pays per closed deal. The board lives or dies by pursued / dead / closed feedback — that's the loop that re-weights the score over time.
Owns the ingest, the scoring, the cockpit. Watches the feedback loop and tunes the weights when reality disagrees with the spreadsheet.
Not the closer's analyst — DealHound is. The operator's job is to make the analyst better at its job every week.
The score
Each signal is a public-record fact that, on its own, hints the owner has checked out, run out of money, or been forced into a corner. Stacked, they're a statistical bet on who picks up the phone.
Every parcel gets a modeled ARV — zip × building-class median $/sqft from the last 9 months of public rolling sales. From that, subtract honest assumptions: ~62% of ARV as acquisition (until the property's actually listed or negotiated) and a per-sqft renovation budget that varies by gut/medium/light.
What's left is the modeled spread. It's the floor George needs to see before he's willing to spend a phone call. It is not the deal — the deal is found at the kitchen table.
Every number is "est." Acquisition tightens once George has a real conversation.
What a lead looks like
The detail page collapses PLUTO, comps, violations, lien status, FAR slack, and the score breakdown into one screen — then puts approve / snooze / kill on the bottom.
The cockpit
The board is the closer's day. New on the left, closed on the right. The dead column isn't waste — it's the feedback that tells DealHound what to weight less next month.
Five tracks
The score is universal — the way leads are surfaced is split by asset class, ordered by how warmly each lane fits the closer's actual phone time today.
1-3 family. The warmest lane. Highest call-to-close conversion in the data.
Commercial ground floor, residential above. Slower cycle, real upside on rents.
Yield product. Pricier acquisition; FAR slack and violations matter more here.
Largest pool, coldest call. Tax-lien signal carries most of the weight in this lane.
FAR-driven. The score is mostly "how much could you legally put here."
Under the hood
The data is all NYC Open Data — public records, free, verified IDs. The work is in cross-referencing them honestly.
Every dataset below is a public NYC Open Data ID, filtered to the boroughs we work.
What it doesn't do
DealHound finds candidates. Calling these limits out is how the score stays trusted as the board grows.
Zip × class median $/sqft. Good enough to surface candidates, never good enough to underwrite.
62% of ARV is a crude rule until a property is listed or actually negotiated. Real numbers come from George's call.
The score finds motivated owner candidates. The conversion still happens at the kitchen table.
Updates around the annual lien sale. Strongest in spring; quietest mid-fall.
What ships next
The single largest predictor in the model isn't live yet. Once it is, the rest of Phase 1 is about turning a flagged parcel into a sent letter without leaving the board.
Each item is queued, scoped, and waiting on the prior one to clear.