140+ signals.
One model. No black box.
The engine isn't magic — it's discipline at scale. Every signal it reads is named, every weight is documented, and every output traces back to the inputs that produced it. This is the layer that makes “explainable” not a marketing word.
Five categories. Different weights at different moments.
The same 140+ signals don't matter equally on every deal. The engine routes weighting by asset class, deal size, and market regime — so a multifamily file in a hot market reads on different dials than a hospitality file in a tight credit window.
Sponsor
Who the borrower is — the first dimension every lender reads. The engine pulls verified prior projects, post-close liquidity, FICO, credit events, net worth, and operating-partner depth. Heaviest weight on smaller deals and value-add files.
Asset
What the asset is, where it is, and what state it's in. The engine reads class, geography, market tier, occupancy (current and stabilized), in-place vs. projected NOI, rent roll concentration, lease maturities, and value-add scope.
Market
The conditions the asset operates in. Live cap rates and their 12-month trend, TTM rent growth, vacancy trajectory, supply pipeline depth, absorption, and the local refi market — CMBS vs. agency vs. balance sheet.
Structure
How the deal is built. LTV, LTC, LTARV, DSCR (in-place and stabilized), debt yield, reserves for interest and cap-ex, IO period, amortization, recourse carve-outs, mezz/pref position, and exit strategy realism.
Context
The macro overlay. Fed funds rate and trajectory, 5-year forward curve, current agency appetite for this profile, refi window depth, and capital-market stress signals that shift desk behavior without showing up in any individual deal's numbers.
How the model stays honest.
Predictive systems decay when they stop being checked against the world. The engine is rebuilt against live closing data quarterly — signals are re-weighted, new ones are added, dead ones are retired.
Re-weighted Quarterly
Every quarter, the model is re-trained against the prior 24 months of closed-vs-declined deal outcomes. Weights move; signals get added or retired based on predictive value.
Every Score Traceable
Every forecast comes with the per-signal breakdown that produced it. If a deal scores a 78, you can see which signals contributed positively, which dragged, and by how much.
Fix Impact Modeled
The recommendation engine doesn't guess — it runs a fresh forecast with the proposed fix applied, then returns the delta. That's how it knows fix.01 lifts FPI by 7 and not 3.
What It Won't Do
The engine doesn't price the loan, name the lender, or commit capital. It scores the file and recommends the move. The funding decision stays with capital sources.
No Credit Pulls
Sponsor signals use stated and verified inputs, not bureau queries. Nothing the engine reads triggers a hard inquiry on your or your borrower's profile.
Same-Day Reads
Median runtime is under four minutes from intake to score sheet. Complex files (large commercial, unusual asset classes) extend to the same business day.
Get your funding score in minutes — no credit pull, no commitment.
See where your deal stands and exactly what to fix before you ever submit it to capital.