Engagement: our own build, from a specification written by our data lead
HAM role: specification, build, daily running
Timing: September 2026, still running
4,312
public contract notices read in the first 45 days
41
passed as worth a bid, under 1 in 100
Daily
run every day since 11 September 2026
02
The situation
UK public bodies publish thousands of contract notices across two government services. We were missing work we could have won, because nobody had the hours to read them all and most are not relevant to a firm our size.
03
The move
We built a scanner to do the reading.
It pulls every open notice from Find a Tender and Contracts Finder.
It scores each one against six disciplines we actually deliver, within a contract value we can realistically take on.
It produces a daily shortlist, and the team works from one shared board.
In the first 45 days it read 4,312 notices: 3,792 from Find a Tender and 520 from Contracts Finder. It passed 41.
04
The rigour
The first scoring rule averaged a notice across all six disciplines, which meant a straightforward single-discipline job always failed. We changed it to score on the best two, and kept the original figure alongside so the difference can be checked.
The government service limits how fast it can be queried. The scanner waits when told to and stores what it has already read, so a daily run only fetches what is new.
The scanner does the reading. A person reads the shortlist and decides what is worth bidding for.
05
Where it stands
Running every day since 11 September 2026
4,312 notices read in the first 45 days, 41 passed
The shortlist is reviewed by a person before any bid decision
06
What we built
The scanner, the scoring rules, the daily report and the shared board the team works from.
07
What's next
More sources, and the same method pointed at grants.
We get into the system, build it from nothing to working, and stay on to grow it.