Find the solicitations nobody sent you.
Public buyers publish. They do not distribute. RFP discovery watches the places they publish — federal notices, fifty state systems, thousands of local portals and the buyer websites that carry a PDF and nothing else — and turns what it finds into one record per opportunity, with the original posting attached.
- Produces
- Opportunity records
- Runs
- On each listener’s own schedule
- Needs
- A company profile and one source
The opportunity you lost was published where you were not looking.
There is no single place public procurement appears. A federal notice is in one system, a state solicitation is in one of fifty, and a county that spends real money posts a PDF on a page with no feed, no alert and no search that reaches it. The buyer has met their publication obligation either way.
So teams do it by hand: a person with twelve browser tabs on a Monday morning, a saved search that stopped working when a portal changed its markup, and an aggregator subscription that arrives a day late with the buyer redacted. It works well enough to find the solicitations you were already expecting, which are the ones you least need help finding.
The cost is not the searching. It is that the response window starts when the buyer publishes, not when you notice — and the four days you spent noticing are four days the incumbent did not spend.
- Late startsA thirty-day window found on day nine is a twenty-one-day window, and it is the drafting end that gets compressed.
- The same bid, counted three timesOne solicitation republished by an aggregator and mirrored on the buyer’s site reads as three opportunities, and somebody qualifies it three times.
- Whole tiers unwatchedLocal and district buyers are the most fragmented tier and the least competed. They are usually the first thing a manual process drops.
From a source you name to a record on somebody’s desk.
Five steps, each of which leaves something behind that the next one uses. Nothing is inferred twice, and nothing arrives without saying where it came from.
Describe what you want watched
A listener is a source, a filter and a schedule: this portal, these categories and jurisdictions, this often.
A listener you can read, edit and switch off — not a black box that decides for you.
Fetch and store the original
Each run collects the posting and its attachments as published, with the fetch time recorded.
A stored original that a record can always be checked against, even after the portal removes it.
Normalize the fields
Buyer, solicitation number, category, closing date and time zone are mapped from whatever the source called them.
Records from fifty formats that can be sorted, filtered and compared as one list.
Deduplicate against everything held
Postings matching on identifier, document hash or buyer-and-closing-date collapse into one record.
One opportunity, with every posting that produced it kept and openable underneath.
Score and route
The record is matched against every capability profile in the organization and queued for whoever owns that profile.
A ranked queue with the reason each opportunity surfaced, and an alert for the ones that clear your threshold.
A sweep, and the three postings it collapsed into one.
The left side is a run: what each listener fetched, what was new, and what was thrown away as a duplicate. The right side is one of those duplicates — the same solicitation published three times, under three titles, merged into a single record that still carries all three.
- Federal publications
- State portals
- Local and district portals
- Buyer websites
- Aggregators you subscribe to
Every row is a listener you configured: a source, a filter and a schedule. Nothing is discarded silently — a duplicate is attached to the record it duplicates, and can be opened.
- verdict: Kept as the record
Enterprise records management system — RFP
— First posting, and the only one carrying the attachment package.
- verdict: Duplicate
RFP: records management modernization (Phase II)
— Same solicitation number and closing date; a different title and a phase number the portal does not use.
- verdict: Duplicate
Records Mgmt System RFP — State of (redacted)
— Same document hash as the portal attachment, republished a day later with the buyer masked.
Enterprise records management system
- Buyer
- State administrative agency
- Closes
- 8 Sep 2026, 19:00 UTC
- Postings merged
- 3
- Attachments
- 11 files, 214 pages
What normalization did to the three postings
- “Bid No.” / “Ref” / “Sol #”STA-2026-2210
- “Due 9/8/26 2 PM CT” / “08-Sep-2026”2026-09-08 19:00 UTC
- portal taxonomy, free-textSoftware · records management
- absent on two of threeNone stated
Counts are from this fixture run. They are not a statement about how many sources are monitored or how much is published anywhere.
Nothing above is a live query or another customer’s pipeline. Every record in it was written for this page, and no buying organization named in it is real.
What discovery does that a saved search does not.
These are properties of the pipeline rather than a list of portals: a connector list printed on a marketing page is out of date the week it is printed.
Listeners you configure
Each listener names its source, its filters and its schedule, and reports its own last run. A listener that stops returning results says so instead of going quiet.
Provenance on every record
Source, fetch time and a link back to the original posting. A record you cannot trace back to a published document is a rumour.
Deduplication that keeps the evidence
Duplicates are attached to the record they duplicate rather than deleted, so a posting that carried an extra attachment is not lost with it.
Attachments pulled with the notice
The package is fetched with the posting, so the document work can start without anybody logging into a portal to download it.
Amendment tracking
A revised posting updates the record it belongs to and flags what changed, rather than arriving as a second opportunity with a similar name.
Closing dates in one time zone
A due date is stored with its published time zone and shown in yours, because “due 2 PM” has cost somebody a bid in every state.
Three teams, three reasons the same feature matters.
Kinds of company, not customers. Nothing here is anybody’s account.
A regional services firm
- Bids in four states and three counties, and finds most work through relationships that only cover half the territory.
- Listeners on each state system and on the twenty local portals inside the service area, filtered to two categories.
- The pipeline stops being a function of who happened to hear about something at a conference.
A national prime
- Watches federal notices closely and has no view of the state and local work that uses the same delivery teams.
- Federal listeners alongside state and cooperative sources, with records routed to the practice that owns each category.
- One pipeline across tiers, so a capacity conversation is about all the work rather than the federal half of it.
A small specialist
- Two people, no capture team, and a category that is bid in a hundred jurisdictions a year.
- A narrow filter and a high score threshold, so the queue is short and everything in it is worth reading.
- The morning starts with a handful of opportunities that match, instead of an hour of portals.
Discovery is on every plan. What changes is how much you can watch.
Every plan includes opportunity discovery across public sources, with provenance and deduplication. The ladder is about scale: how many listeners you can run, how many saved searches you can keep, and how many matches a month the plan covers.
- Starter includes discovery, a small number of listeners and saved searches, and one company profile.
- Professional and above raise the listener and saved-search allowances and the monthly match volume.
- Agency adds unlimited company profiles under a fair-use ceiling, for a firm watching on behalf of many clients.
- Private and custom source connectors — a credentialed portal of your own — are an Agency capability.
Plan names and allowances are mirrored from the billing catalog. Amounts are on the pricing page, which reads them from Stripe rather than from a number typed into a marketing page.
What people ask about discovery.
Which sources does it actually watch?
Seven categories: federal publications, state procurement systems, local and special-district portals, buyer websites reached by crawl, news, public databases such as award registers, and public social posts. A named list of every portal would be out of date the week it was published, so what is stated is the category, and every record says which source it came from.
How quickly does something appear after it is published?
On the schedule you set for that listener. A listener that runs hourly finds a posting within the hour of its next run; one that runs daily finds it the next morning. Nothing here promises a number, because the honest answer depends on the source and on how often you have chosen to ask it.
What happens when a portal changes its layout?
The listener reports a failed or empty run rather than quietly returning nothing. Silent failure is the specific way saved searches and scrapers let teams down, so a listener that stops finding what it used to find is surfaced as a problem to fix.
Can it watch a portal we have credentials for?
Private and custom connectors are an Agency-plan capability. The credentials belong to your organization, are stored encrypted, and are used only by your listeners.
Is this a database of open bids we can browse?
Discovery is a pipeline into your workspace, scored against your profile. A public searchable directory is a separate thing, and it is listed under Opportunities in the navigation rather than being what this feature is.
Point it at one source and see what you have been missing.
Set up a single listener on a portal you already watch by hand, and compare a week of what it returns against a week of your own tabs.