How the picture is built.
VectorBrief assembles federal + state-and-local procurement data into a single consistent picture and re-ranks it for your firm continuously. What follows is the 30,000-foot view — enough to answer "is this trustworthy" without exposing the formula.
What we cover
A multi-tier source mix: federal contract opportunities, federal awards (bulk + delta), federal grants, regulatory notices, state procurement portals, and cooperative purchasing vehicles. The /data page shows the live counts; the cadence table beneath it shows how often each tier refreshes.
We don't publish the supplier list — every public-procurement vendor uses a similar mix of authoritative US-government feeds, and the differentiation is in resolution, freshness, and ranking, not in which feed someone subscribes to.
How fresh is it
Per-tier cadences are published on /data. Federal contract opportunities and federal awards refresh every few minutes. Federal grants refresh daily. Regulatory notices refresh hourly. State and cooperative tiers refresh every few hours.
Every record carries provenance — source label, upstream id, first seen, last refreshed, observation count — so a buyer can verify the lineage upstream without trusting our internal timestamp.
How we resolve agencies
Federal agencies appear under different names across feeds — full name, common abbreviation, sub-bureau, parent department. We maintain an alias-graph that collapses these into a canonical agency identity, then attach every opportunity, award, and regulation to that canonical identity.
The result: one search for "Department of Defense" pulls every record across every source, regardless of how that source spells it. The agency-coverage card on the search page shows how complete the resolution was for the prior 24 hours.
How we resolve vendors
Vendor identity follows UEI (Unique Entity Identifier) and CAGE precedence with name-based fuzzy matching as a fallback for pre-UEI historical records. Same opportunity, same vendor across every source — even when the upstream feed mis-spells, abbreviates, or uses an outdated DUNS.
The Resolved Companies tile on /data lights up when this layer ships in production. Until then it carries a zero — we're honest about coverage gaps.
How we rank opportunities
Two layers, run in tandem:
- Deterministic eligibility filters — set-aside qualifications, cooperative-vehicle access, agency exclusions, state-or-region constraints. These are facts about your firm; we apply them as hard gates rather than soft signals.
- Semantic similarity to your past wins, your stated capabilities, and the agencies you've competed for. We combine the two layers into a single score per opportunity per subscriber, refreshed every time either side changes.
Each opportunity in your feed carries a one-sentence rationale that names the specific evidence behind the score — the past contract, the matching code, the certification or capacity that unlocked the ranking — so you don't have to take the number on faith.
How we measure ourselves
Calibration plot: predicted win rate vs empirical win rate, refreshed every time enough new pursuit outcomes have landed to recompute. The goal is monotonic alignment — opportunities ranked higher should win more often than opportunities ranked lower, on average.
We publish the plot internally and re-run it on every model change. External publication of the plot is a follow-up.
What we don't do
- We don't write proposals for you.
- We don't lobby on your behalf.
- We don't share your watchlist, profile, or pipeline with any other subscriber, prospect, or third party. Ever.
- We don't sell ads, anonymized data, or aggregate firm-level patterns to third parties.
For a downloadable PDF of this page, contact hello@vectorbrief.io.