Intelligence suite
Connected views over the same signal plane, relationship record, and teaming evidence.
| Surface | What it answers |
|---|---|
/intelligence | Who's buying, who's winning, and where the market is moving — agency demand, incumbents, per-agency competitive landscape, plus evidence-linked Growing Agency, Emerging Program, Vehicle Dependency, and Contract Concentration Risk signals with confidence and evidence counts. |
/agency/[name] | One agency, fully connected: obligated value, top incumbents with share, open opportunities, protest activity, key contacts, vehicles in play, predicted pipeline. |
/contacts | The tenant relationship system of record: managed government, partner, industry, and internal contacts; interaction history; meeting intelligence; stakeholder coverage; outreach; plus provenance-backed SAM.gov leads. It remains distinct from shared-market Contact Relationship Intelligence, which measures only accepted procurement associations and never acts as a lobbying or influence score. |
/partners | The teaming system of record: managed organizations, company-gap analysis, explainable fit, SAM/SBA discovery, pursuit roles, performance, agreements, JVs, mentor-protege structures, secure rooms, and recommendations. |
/vehicles | Which contract vehicles carry the work — named GWACs / IDIQs / schedules (SEWP, CIO-SP4, Alliant, OASIS+, MAS…), with first-class Pool → Domain → Track → SIN scope and award-holder lineage. The same dashboard shows source-backed IDV ceiling, observed task-order obligations, remaining runway, utilization, vendor-pool usage, and completeness/confidence; missing ceilings remain unknown. |
/protests | GAO bid-protest intelligence. Sustained protests are surfaced as re-opportunities (corrective action / re-solicitation), plus per-agency sustain rates. |
/graph | The governed procurement graph: Agency → Subagency → Office → Program Office → Program → Budget Account; Vehicle → Pool → Domain → Track → SIN → Award Holder; opportunities, solicitations, IDV parents, contracts, task orders, vendors, certifications, contacts, protests, and subawards. Contact graph nodes persist source-backed relationship strength, association count, and last-seen date for procurement-pattern recognition. Canonical subawards retain their source record plus resolved prime, subcontractor, and Contract-or-Task-Order UUIDs. Corporate views preserve parents, subsidiaries, business units, acquisitions, mergers, rebrands, spin-offs, divestitures, JVs, mentor-protégé agreements, and original/current award holders without duplicating one UEI. Vendor nodes retain DSBS capability evidence and renewable certification periods. Contracts retain current/all-options/ceiling values, obligations, remaining potential, modification counts, and separately dated action evidence with a completeness level. Budget nodes preserve Treasury account, requested/enacted amounts, relevance, and source-document evidence. Stable source IDs also preserve replaced-by, related, follow-on, competing, consolidated, and split opportunity lineage. |
Vendor Performance Graph
This shared-market read model answers who is winning, where they perform, and how complete the supporting money evidence is. It is calculated only from accepted procurement-graph projections; it is not a tenant CRM score and does not alter a company's private workspace data.
| Metric | Meaning |
|---|---|
total_awards / prime_award_count / subaward_count | Observed source-backed prime awards plus canonical subaward participations. A Vendor-to-Vendor teaming edge alone never creates a synthetic subaward count. |
total_obligated / average_award_size | Observed obligations and award values from accepted award nodes. Missing source money remains null, never a fabricated $0. |
agency_count / office_count / vehicle_count | Distinct accepted agency, office, and vehicle links reached through the vendor's observed award and IDV relationships. |
recompete_wins | Only completed award-stage opportunities with source-observed follow-on, replacement, or consolidation lineage qualify. It remains null when that proof is absent. |
financial_data_completeness / confidence / calculation_basis | Coverage, relationship confidence, method, source-value basis, warnings, and bounded canonical evidence IDs make each record reproducible. |
Truthful unknowns. A source-backed award can still lack complete financial data. WinLayer shows the known award reach while clearly marking obligations, averages, or recompete wins as unavailable until the necessary evidence is present.
Contact Relationship Intelligence
This shared-market calculation recognizes recurring procurement patterns around a Contact, Office, Program, opportunity, and vendor. It does not score lobbying activity, personal influence, or a tenant's private CRM relationship.
| Field | Meaning |
|---|---|
relationship_strength | A bounded 0–100 deterministic strength derived from accepted Contact → Contract or Task Order associations, distinct award records, and recency. It is zero when no accepted procurement association exists. |
interaction_count | The number of active, accepted Contact-to-award association edges. It represents observed procurement appearances, not unverified meetings, outreach, or lobbying interactions. |
last_seen_date | The most recent source-backed Contact-to-award observation used by the graph calculation; a Contact without an association retains its source record date but is not published in the intelligence list. |
relationship_confidence / relationship_evidence_edge_keys | Average accepted-edge confidence plus bounded graph-edge evidence identifiers make every metric reviewable and reproducible. |
Evidence boundary. One source-backed appearance may establish a Contact's procurement context, but it cannot establish vendor preference. The separate
FREQUENTLY_AWARDS_TO relationship appears only after repeated accepted award associations and retires if the source evidence becomes inactive.Teaming Intelligence Graph
This shared-market view distinguishes observed teaming history from a potential partner recommendation. A relationship appears only when accepted canonical subaward evidence connects a prime vendor to a subcontractor; a loose Vendor-to-Vendor edge or a name similarity cannot create one.
| Field | Meaning |
|---|---|
relationship_type | subcontractor for a source-observed prime-to-subcontractor relationship. Potential teaming recommendations remain a separate, clearly labeled product output. |
shared_contract_count / shared_contract_entity_ids | Distinct canonical Contract or Task Order entities that carry accepted subaward evidence for this prime/subcontractor pair. |
shared_agencies / shared_agency_entity_ids | Agencies linked to those same contracts. The API returns display names and their canonical graph IDs. |
total_relationship_value / relationship_value_completeness | The sum of observed subaward amounts plus the share of canonical subaward records with a sourced amount. Missing money remains null instead of becoming $0. |
observed_subaward_count / evidence_subaward_entity_ids | The bounded canonical Subaward records supporting the aggregate. They make the relationship reproducible and reviewable. |
Evidence boundary. WinLayer never represents a recommendation, generic partnership label, or unverified company-name match as past performance. Historic teaming is published only from canonical, accepted subaward evidence and is retired when its supporting graph evidence is no longer active.
Program Matching
Programs can use different names in budget PDFs, agency forecasts, SAM notices, and award descriptions. WinLayer resolves them with a governed profile: structured procurement scope first, then an optional pgvector semantic comparison when an embedding is available.
| Rule | How it protects the graph |
|---|---|
Agency → Office → Program Office → Program | Agency, office, program-office, program code, budget account, NAICS, and PSC fields create the scoped matching context. A program code links directly only after its agency scope is checked. |
Exact / structured lexical | A normalized program-name match must remain in compatible procurement scope. If semantic coverage is unavailable, automated matching requires strong structured agreement plus a near-exact name; otherwise a separate Program remains. |
Structured semantic vector | A pgvector cosine comparison can support a differently named program only when it is at least 92%, shares compatible scope, has at least 50% structural agreement, clears the combined score gate, and exceeds the next candidate by 5 points. |
Cross-agency boundary | Different agency codes are a hard exclusion. A high vector score never overrides that boundary, and ambiguous candidates are not auto-linked. |
Profile freshness and provenance | Canonical program text, aliases, structured context, source systems, embedding model/hash/time, and the bounded semantic-resolution event are stored in PostgreSQL. Raw vectors and raw source-document text never leave the authenticated API. |
Truthful semantic state. When no embedding provider is configured or it is temporarily unavailable, WinLayer labels the profile as structured-only. It does not imply that semantic matching ran, and a later bounded profile refresh can add coverage without changing the original source evidence.
NAICS ↔ PSC crosswalk
Contracting officers can select a broad NAICS even when the Product and Service Code gives a stronger capability clue. WinLayer therefore maintains a governed many-to-many mapping learned from repeated USAspending, SAM.gov, and historical-award pairings and refined by Platform Administrators.
| Field | Meaning |
|---|---|
signal.naics_original | The source-record NAICS. WinLayer preserves it and never relabels an inference as source truth. |
signal.naics_inferred | Additional six-digit capability codes supported by the sourced PSC and active crosswalk evidence. |
signal.crosswalk_confidence | The strongest active mapping confidence, 0–100, retained in relevance explanations and graph-edge evidence. |
Evidence gate. One pairing cannot activate a learned mapping. Active learned rows require repeated observations and at least 50% confidence; inferred NAICS receives confidence-weighted relevance credit, never the full credit reserved for a source NAICS.
Everything links. Agency names across the product click through to their profile; managed contacts connect tenant relationship history and partner evidence to the same provenance-tracked signals, graph entities, opportunities, programs, and documents.