Scoring model

Inspect the factors behind a fit score and preliminary win-probability estimate before making a pursuit decision.

Beta evaluation status. The score shows its contributing factors in the workspace. Predictive performance, calibration, and partner usefulness are still under independent evaluation. Missing or stale source evidence can lower confidence or leave a field unavailable.

Profile × evidence → six separate decision measures → eligibility gate → governed recommendation.

Six separate decision measures

WinLayer keeps capability fit separate from legal and contractual eligibility, deal quality, competitive posture, win likelihood, and evidence quality. Each measure shows its inputs and explanation in Radar and the opportunity workspace.

Measure	Question	Governed inputs
`Company fit`	Can and should we perform it?	NAICS and scope similarity, relevant past awards, contract-size fit against the company revenue tier, vehicle access, geography when stated, agency and buying-office history, and company capabilities.
`Eligibility`	May we pursue it as proposed?	Authoritative set-aside and size evidence, certification status, required vehicle access, and recorded organizational-conflict constraints. This is a hard gate, not points added to another score.
`Strategic attractiveness`	Is it worth pursuing?	Known or bounded value, contract type, expected duration, growth alignment, timing, and the cost of closing known gaps.
`Competitive position`	How strong is our position?	Buying-office and customer history, incumbent strength, competitive density, partner posture, and supported vulnerability signals.
`PWin`	What is the preliminary likelihood?	Eligible company fit, competitive position, evidence strength, acquisition context, and confidence decay. It remains a preliminary heuristic until the historical benchmark and confirmed outcomes establish calibration.
`Evidence confidence`	How much should we trust the other measures?	Source authority, coverage, freshness, agreement, citations, unresolved fields, and conflicts. Missing evidence lowers confidence rather than becoming a guessed value.

Recommendation rule. An authoritative eligibility failure produces no bid. A partner-closable vehicle, certification, or capability gap produces needs partner. Incomplete material evidence produces needs more info. Otherwise company fit, strategic attractiveness, competitive position, preliminary PWin, and evidence confidence determine pursue, monitor, or defer. Radar shows the six inputs beside that result.

Fit score

Factor	Points	How it matches
`Source NAICS code`	up to +40	A source-record NAICS match receives exact or bounded affinity credit across the six-, four-, three-, and two-digit hierarchy.
`PSC-inferred NAICS`	up to +34	A governed PSC crosswalk can contribute confidence-weighted NAICS credit; mappings below 50% confidence contribute nothing.
`Target agency`	+25	The issuing agency matches one of the company profile's target agencies.
`Set-aside held`	+20	The opportunity's set-aside matches a certification recorded in the company profile.
`Set-aside locked out`	-15	A restricted set-aside does not match a certification recorded in the company profile.
`Capability keyword`	+15 to +27	One or more company keywords appear in the title or description; additional matches add bounded credit.

Win probability

PWin is computed only after the eligibility gate, using fit and supported competition context: full-and-open vs set-aside, incumbent presence, requirement age, and signal completeness. The result is a band used by the recommendation rule:

Band	Recommendation	Read it as
`strong`	pursue	Well-positioned; commit capture effort after review.
`competitive`	monitor	Credible position; monitor and validate gaps before committing.
`long_shot`	defer	Do not commit now; revisit only if the evidence or position improves.
`disqualified`	no bid	A hard eligibility gate failed.

PWin is a preliminary, heuristic pre-bid estimate. Missing evidence or a partner-closable gap can override the default band recommendation with needs more info or needs partner. Treat the result as a review prompt until enough confirmed outcomes support calibration.