Searchable notices
…
— still open for bidding, — award records held.
Accuracy & coverage
The numbers behind the platform, published live
Procurement software is sold on adjectives. These figures are read straight from the production database every few minutes — including the ones that are not yet flattering.
Coverage
What is in the catalogue right now
Counted from the live catalogue, not a sales deck. Refreshed every fifteen minutes.
Countries covered
…
National, regional, municipal and multilateral buyers, normalised into one schema.
Sources reporting
…
— of configured sources returned data in the last seven days.
Added in 24 hours
…
— new notices in the last seven days.
Freshness
How quickly a published notice reaches you
A tender you find after the deadline is worthless. So we measure the gap between publication at source and availability here, and print it whether or not it flatters us.
Reading live collection timings…
- Median lag
- —
- Slowest nine in ten
- —
Data quality
Duplicates removed before you see them
Aggregators pad their headline counts with the same notice arriving from four feeds. We merge them and show the working.
—
of the — records collected were duplicates of a notice already held, leaving — distinct notices and — that pass the quality rules for search.
- Reference, buyer and title fingerprinting merges cross-portal republications.
- Records missing a buyer, deadline or usable title are withheld from search rather than padded in.
- Award notices are counted separately so open opportunity counts stay honest.
AI accountability
We grade our own scoring against real outcomes
Every win probability we produce is stored with the tender. When you record what happened to the bid, that prediction is scored against reality — for your workspace privately, and in aggregate here.
Reading recorded bid outcomes…
Every AI claim is quoted
Fit assessments and bid drafts must cite a verbatim passage from the tender documents. A sentence that cannot be matched word-for-word against a real document is refused before you ever see it — the model cannot assert a requirement that is not written down.
Duplicates are collapsed, not counted twice
The same notice is republished across national portals, aggregators and funder feeds. We fingerprint and merge them, and we publish the collapse rate rather than inflating the headline count.
Drafts stay drafts
Auto-Pilot produces a compliance matrix and response text for you to review, edit and sign off row by row. Mandatory gaps block a bid from being marked submittable. Nothing is submitted on your behalf.
Original wording is preserved
Notices collected in French, Spanish, Portuguese and other languages are stored in the published original alongside the English rendering, with a one-click switch back to the source text. Translation never overwrites the legal wording.
Method
Exactly how each figure is calculated
So you can challenge them. Every definition below maps to a query against the production database.
- Searchable notices
- Distinct notices your search can return: duplicates merged and withheld or low-quality records excluded. This is the only total we advertise.
- Collection lag
- Hours between the publication timestamp at the source and the moment the notice became searchable here, measured over every notice published in the last seven days.
- Sources reporting
- Configured sources that returned at least one successful collection in the last seven days. Sources paused for repeated failure are counted in the denominator, not hidden.
- Win-score accuracy
- Recorded bid outcomes grouped by the probability we predicted before submission, compared with the share actually won. Published only above a minimum sample size.
Figures refreshed —. Operating state is published on the status page, and security and data handling in the Trust Centre.
Check the numbers against your own market
Search live notices in your sector and country before you pay anything, then take a 14-day trial of alerts, AI fit scoring and bid drafting.