Every property
on the map.
Already scored.
PinIQ does not make you open listings one at a time. Every property in view is already underwritten and graded — cap rate, cash flow, DSCR, ROI — so you pan the map and read the answers off the pins. Rank the whole set by viability, screen it on the numbers, then open the one that earned it. And PinIQ says what each score is made of: a component with no real source is excluded, never filled in.
Sourcing is solved.
Underwriting is not.
PinIQ sits on the seam between finding a property and knowing what it's worth to you. One engine computes a property's monthly economics, and every card, tab and report in the product reads that same engine — so two screens can never disagree about the same house.
The whole market, not one listing
Every property in the result set is on the map at once, each pin carrying its own letter grade and score. Turn on Search Area and the ranked list re-narrows to whatever you can see, sorted by viability, ROI, cash flow, cap rate or price. Screen the set on minimum score, DSCR or price per unit before you open anything — then open one listing into a full underwrite: mortgage-inclusive cash flow, the 1% and 50% rules, break-even rent.
Dual-mode workspace
Flip the workspace to commercial for a pro-forma, IRR, a GP/LP waterfall and a rate-stress sensitivity grid. An AI parser reads a T-12 or rent roll so nobody retypes it.
Comps beside the AVM
A one-click comp report puts weighted comps — similarity × recency × distance, all disclosed — next to the raw automated value, with the investor overlay: ARV, cap rate, 70% MAO.
Pro-forma vs. actuals
The budget you underwrote is snapshotted and frozen, then measured against what actually happened each month — favorable in teal, off-plan in amber, per property and per portfolio.
It tells you what it is made of
Where a rent is inferred from asking price rather than measured, PinIQ names the share of rows that applies to and the metrics it moves. A property whose cap rate or price per unit cannot be verified is excluded from a filter rather than passed through, one with no financial anchor scores N/A instead of a confident guess, and a capped search prints 15,000 of N matching rather than averaging a silently truncated set.
Teardown, priced
For the spec developer the question is residual land value, not cash flow. PinIQ prices the teardown-and-rebuild on each parcel, and a Teardown band filter surfaces the ones that clear it.
The properties the models cover
One tab lists every property PinIQ's own models can speak to, each with its rent basis and its distress band, so you see where the evidence is strongest before you underwrite.
The champions price the corpus
Where a rent used to be inferred from asking price, the regional rent model now replaces it on refresh and stamps the row modelled, so the basis you see is the basis you get.
Commercial paper, property by property
CMBS ABS-EE filings put 7,049 properties across 68 trusts and 55 states on the map, and CFPB mortgage-performance data gives a delinquency climate for 507 counties — 76.6% of the US population.
Every screen on this page is drawn procedurally in your browser — no screenshots, no stock imagery. The model figures on them — MAPE, R², lift, corpus and parcel counts — are PinIQ's own held-out test results. The properties, prices, grades and result counts are sample data.
Underwrite one house.
This is the real shape of PinIQ's underwriting engine: the same default assumptions (20% down, 30-year amortization, 5% vacancy, 5% maintenance, 8% management, 3% closing costs) and the same Viability weighting — Financial 55 · Location 20 · Market 15 · Risk 10. A house you invented has no real location, market or risk signal, so those three are excluded here and the weight renormalizes onto financial alone — exactly what the product does with a missing component. Move anything.
Rate-stress surface · DSCR by rate and rent
Taxes and insurance here are modelled at 1.6% and 0.5% of price per year, and cash invested adds the product's default 3% buy-side closing cost. In the product taxes and insurance come from the property record, and anything without a real source is excluded from the score with the remaining weights renormalized — the score always reports its own basis.
The models PinIQ
actually owns.
Anyone can resell a data feed. These are trained in-house on a licensed listings corpus whose terms permit training, versioned in a registry, and demoted automatically when they drift.
PinIQ Rent Model
A gradient-boosted rent estimate served by a round-trip-exact pure-TypeScript scorer — no ONNX runtime in the serving container. Champions are qualified per region, so a property routes to the model trained on its states: tri-state 15.9%, West 14.1%, Texas 14.0% test MAPE. A property outside every covered region gets an honest refusal, never a guess.
tri-state band ±22%
Predictive distress
The forward moat: a percentile ranking of which parcels are most likely to hit a distress filing in the next twelve months. Three metros serve — 584,920 Allegheny County parcels on a true mortgage-foreclosure-filing label, 316,497 Prince George's County parcels on a foreclosure-registry label, and 858,283 NYC lots on a disclosed composite proxy, because true filings are not in NYC Open Data — surfaced as a percentile only, never a claimed probability. Pittsburgh's own composite-label v1.0 failed the pre-registered bar and was killed rather than shipped soft.
top-decile lift · NYC · Allegheny · PG
Viability Score
One 0–100 composite on a single shared scale, with an A+ to F grade table used everywhere. Rebuilt to use real data only: a component with no real source is dropped and the remaining weights renormalize, so the number can't be inflated by a placeholder.
fin · loc · mkt · risk
Decay monitor
The serving path writes one prediction in twenty into a backtest table as it returns it. At the six-month mark the monitor re-resolves those rent outcomes against a fresh independent valuation pull and flags any region whose rolling error crosses 20% — a flagged champion is demoted, which alone removes it from serving.
auto-demote
Comp report
Weighted comps shown beside the raw automated value rather than instead of it, with the weighting disclosed. Includes a manual-subject mode for an address that isn't listed anywhere. Carries a USPAP disclaimer and is never called an appraisal.
quota-guarded
Free public data,
maximized.
Licensed feeds are rented; public-domain data is owned — no licensor can revoke it, re-price it, or forbid training on it. So PinIQ ingests the free layer hard, keeps its history, and tells you per metric where each number came from.
What PinIQ refuses to do: no scraped competitor content, ever · no raw-data export — aggregated output only · restricted paid feeds and share-alike data physically siloed away from the trainable corpus · skip-trace personal data firewalled out of the lake · no fabricated metric, no invented delta, and a dash instead of a number when history doesn't exist yet.