The Office for National Statistics publishes a monthly count of new online job adverts in the UK, broken down by SOC 2020 occupation code and by region. It is free, it is authoritative, and almost nobody reads it, because it ships as a wide spreadsheet workbook that is painful to work with programmatically.
We pulled out thirteen occupations that UK graduates actually enter, reshaped the whole thing into long format CSV, and put it in a public repository with a DOI. Here is what twelve months of it says.
Change in new online job adverts, UK, Sep 2025 to Aug 2026
| Occupation | Sep-25 | Aug-26 | Change |
|---|---|---|---|
| Management consultants and business analysts | 3,380 | 5,494 | +62.5% |
| IT business analysts, architects and systems designers | 6,417 | 9,861 | +53.7% |
| Graphic and multimedia designers | 1,160 | 1,707 | +47.2% |
| Mechanical engineers | 2,120 | 3,036 | +43.2% |
| Programmers and software development professionals | 9,404 | 11,591 | +23.3% |
| Project support officers | 1,511 | 1,842 | +21.9% |
| Civil engineers | 2,282 | 2,721 | +19.2% |
| Human resources and industrial relations officers | 6,241 | 6,888 | +10.4% |
| Sales accounts and business development managers | 11,776 | 12,837 | +9.0% |
| Data analysts | 1,692 | 1,800 | +6.4% |
| Finance and investment analysts and advisers | 5,854 | 6,205 | +6.0% |
| Advertising and marketing associate professionals | 4,843 | 4,922 | +1.6% |
| Chartered and certified accountants | 1,702 | 1,474 | -13.4% |
Three things that surprised me
Software is not the story people think it is. Given the last two years of coverage you would expect programmer postings to be falling. They are up 23.3%, and they remain the second largest category in the set by absolute volume. The much larger jump is in IT business analysts and systems designers, up 53.7%, which is the specification and integration end of the work rather than the implementation end.
Data analyst is the weakest technical role here. It grew 6.4%, slower than HR officers. It is also the second smallest category in the table at 1,800 adverts. If you are picking between a data analyst track and a software track on the assumption that data is the growth area, this particular measure does not support that.
Accountancy is the only outright decline. Chartered and certified accountants fell 13.4%, the only negative in thirteen occupations, during a period when every other professional category grew.
What these numbers are not
Worth stating plainly, because advert counts get misread constantly.
This counts adverts, not vacancies, not hires and not headcount. One role advertised across three boards can appear more than once. A company that fills a position internally never appears at all. A rising count can mean more demand, or it can mean the same demand being advertised more aggressively or churning faster.
ONS also writes [x] where it did not publish a figure for an occupation in a given month. We carried that through unchanged rather than smoothing it, dropping the row or converting it to zero. If you sum a column containing one, say so.
The occupation selection is ours. The measurement is ONS's, and ONS does not endorse the selection or anything concluded from it.
The data
Two CSVs, long format, UTF-8, with a header row. One is August 2026 by occupation and by all twelve ITL1 regions, the other is UK monthly totals from Sep 2025 to Aug 2026.
- Dataset and DOI: 10.5281/zenodo.23344785
- Original ONS source: Labour demand volumes by Standard Occupation Classification (SOC 2020), UK
- A readable breakdown by region: thecareery.com/uk-job-adverts
Columns are soc_2020, occupation, area, month, new_online_job_adverts. CC BY 4.0, so use it however you like as long as you cite ONS as the source of the measurement.
If you pull it apart and find something we missed, we would like to hear it.
Disclosure: we build free CV and career tools at thecareery.com, which is why we were digging through this data in the first place.
Top comments (3)
The long-format CSV and DOI are strong reproducibility choices. I would also publish the exact SOC-code mapping and a small script that rebuilds the chart from the raw ONS workbook, including the treatment of revised months. That lets readers distinguish a real trend from a classification or extraction change.
Fair points.
The script we have builds the chart from our derived CSV, not the ONS workbook, so the extraction step is the part that isn't checkable. We'll publish that, plus the rule behind the thirteen SOC codes, which isn't documented yet.
Revisions will have to wait until the next release, when there are two editions to compare.
Thanks.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.