How lender matching works

Every lender publishes guidelines: minimum revenue, time in business, credit, maximum positions, restricted industries and states. A shop with twenty lenders cannot keep all of that in its head, so deals go to lenders who decline them on sight, and good lenders get missed. Matching does the lookup for you and shows its working.

Matching starts from a snapshot of the deal built from the merchant, the owners, the bank statement figures, the existing positions and the request. The metrics we evaluate today:

Each lender can have one or more programs (an A-paper program, a renewals program, a sole-prop program). Each program holds rules. A rule is a metric, an operator and a value, plus a severity:

Operators cover the usual cases: at least, at most, equals, in a list, not in a list, between, and exists. Lists cover states, industries and entity types.

Each rule returns one of three results, and the third one matters more than people expect:

• PASS: the deal meets the rule.
• FAIL: the deal does not.
• UNKNOWN: the deal is missing the data (no credit score yet, no statements entered). Unknown is not a pass; it tells you what to collect.

Each result keeps the actual and the required value, so the explanation reads like an underwriter's note: "Requires $20,000+ revenue; this merchant has $17,450."

Ineligible            = any hard rule FAILS
Needs data            = no hard fail, but a hard rule is UNKNOWN
Eligible with warnings = no hard fail or unknown, but a soft rule FAILS
Eligible              = everything passes

Lenders are then ranked by tier first and score second. The score is the share of weighted soft rules that pass, so two eligible lenders can still be told apart.

Lender guidelines are often softer than they look. A deal that misses a hard rule by a small margin (a FICO ten points short, revenue a few percent under) is worth a call to your rep to ask for an exception. Matching flags those near misses instead of hiding the lender.

A restaurant in New York asks for $60,000. Statements show $82,000 average monthly true revenue, an average daily balance of $9,800, one NSF in the worst month, two negative days, one open position paying $6,300 a month, 30 months in business and an owner FICO of 640.

Lender B gets the submission. Lender A is not a near miss here (40 points is a lot), but if the owner's FICO came back at 675 it would be. You could still submit to Lender A anyway; a good system lets you and records that you did.

Before picking an amount, check affordability: the new payment plus $6,300 a month against $82,000 of revenue. The stacking calculator does that sum.

• Record where each rule came from (portal, rep email, observed) and when you last checked it.
• Compare decline reasons with predictions: a decline for a reason the rules said passed means the stored criteria are wrong.
• Re-run matching whenever the statements, positions or the request change.
• Keep matching rule-based. Learned approval odds can help rank lenders later, but they should never override a hard rule.

See how this works in the product on the lender matching page.