Last month I told you August was up 10.9% on July. I was wrong. The real number is flat, and how I got it wrong is the useful part.
Flint reads about 130 job sources twice a day. For this post I kept the eight that ran the same fetch, every day, all summer: NHS Jobs, Civil Service Jobs, Greenhouse (156 company boards), Lever (32), Ashby, Michael Page, TES, and nine trade-press boards on Madgex (Law Gazette, The Lawyer, The Grocer, Nature Careers and friends). Eight sources, about 200 underlying boards. "New" means an ad id we hadn't seen before, so this counts the ads we observed appearing, not every vacancy in the country. Between them, around 4,000 a week. Here's every week since 20 July.
That's a ruler. Nobody went to the beach.
So where does the famous August dip come from? The calendar. The raw monthly totals do fall, 18,108 in July to 16,326 in August, and every one of the eight sources falls with them. But July had 23 working days, August had 20, and almost nobody posts a job at the weekend (0.6% of NHS ads, 1.6% of Teaching Vacancies, 2.1% of Greenhouse). Divide by working days and the dip is gone.
And it isn't one big source hiding movement in the others. Per working day, every source posted within 10% of its July rate in August.
The civil service, the NHS and a few hundred tech and professional firms all kept posting at the same daily rate through the school holidays. I'd assumed the opposite for years. I think most people have.
So what was the 10.9%?
Plumbing. In September I summed "new ads" across every source we read, and that sum moves whenever the set of sources does. Totaljobs joined the pool in July, and its first weeks were a backlog of existing ads rather than new hiring. Changes in which searches we run against Reed and Adzuna change how many ads they return. One shift like that swings a monthly total by more than the thing you're trying to measure. Eight sources I understand beat 130 I don't..
The one real summer collapse we could see
Teaching. State schools post on Teaching Vacancies, the DfE's own board. It isn't one of the eight above (we changed how we read it in October), but from May to September the fetch was constant, and the whole school year is visible in the weekly line.
Teachers who want to leave in the summer must give notice by 31 May. Schools know their September gaps in the first week of June, post for them, and then post almost nothing for eight weeks: 160 to 190 ads a week through August, an 88% fall from the May peak. First week back, 456. Four weeks later, over 1,000 again, hiring for January.
So the August hiring slowdown wasn't visible on the professional boards we could measure reliably. In teaching, it was enormous.
What to do with this
If you're job hunting, don't wait for September. On these sources the roles arrived in August at the same daily rate as July, and an application in a quiet-feeling month is still an application. If you teach, your two windows are May to June and September to October, and late July is dead.
One piece of housekeeping. I also promised to rerun September's senior-to-junior count. It moved, from twenty to one to seven to one in software, but the pool it's measured on nearly doubled in the meantime, so I don't trust the comparison. From tonight Flint records a nightly snapshot of the pool with a fixed definition of what counts, by discipline and by source, so the next version of that number is measured against the same thing.
Next check: 12 November.
Data: Flint UK new job ads by week, July to October 2026. Published under CC BY 4.0. Quote any number here with a link back. Query set on request: chris@getflint.work.