Labour market data

Skills not met

/v1/labour/skills-unmet ranks the skills adverts ask for by how often they are not met. Each row has the share of adverts asking for the skill and the level they usually ask, the residents below that level, the people turned down for it, and the share of hired against turned down applicants who had it at the level the role asked. A big positive difference_pct_points says the skill is part of why people get these jobs.

score is adverts_share_pct times people_below_share_pct divided by 100: of every 100 pairings of an advert and a person, about this many are an advert asking for the skill and a person who does not yet meet it. Filter by place, sector, soc, level, source and window (3, 6, 12 or 24 months), or pass skill and by to break one skill down by place, sector, SOC 2020, level or month.

Adverts are counted only from roles posted on the platform, employers' own applicant tracking system boards and Find an Apprenticeship. Listings from job boards and aggregators, and listings imported from them, are left out, because their terms do not allow vacancy counts to be passed on.

curl -s "https://opportunityplatform.co.uk/api/v1/labour/skills-unmet?place=oxford&window=12" \
  -H "Authorization: Bearer $OP_API_KEY"

The endpoints

• Endpoint

  GET /v1/labour/skills-unmet

  What it answers

  Which skills adverts ask for that people do not yet meet, ranked, with the people turned down for each and what hired applicants had. By place, sector, SOC 2020, level, source and month.

  Paged

  Yes, by rank

• Endpoint

  GET /v1/labour/skills-gaps

  What it answers

  Where skills gaps are forecast, by occupation, with how many independent published sources agree.

  Paged

  No

• Endpoint

  GET /v1/labour/course-demand

  What it answers

  What people are training for: apprenticeship starts, further education enrolments and Skills Bootcamps, with year on year change.

  Paged

  No

• Endpoint

  GET /v1/labour/evidence

  What it answers

  The published evidence rows behind the figures, each with its source, licence, link and date.

  Paged

  Yes, newest first

• Endpoint

  GET /v1/labour/sources

  What it answers

  Every source you can receive, with its licence, cadence and freshness.

  Paged

  No

• Endpoint

  GET /v1/labour/regions

  What it answers

  UK regions by Travel to Work Area (the default), council, LSIP area or ITL region, with official ONS codes.

  Paged

  No

Course demand, skills gaps and evidence take country (ISO 3166-1 alpha-2, default GB); regions cover the United Kingdom. The full parameters and response schemas are in the API reference.

One response shape

Every labour response has the same four parts. data holds the figures. sources names where each figure came from, with its licence, link and the attribution_statement to show with it. attribution_notice is the rule below, in every response so it travels with the data. meta.as_of is the date the data describes, and meta.generated is when the response was built.

curl -s "https://opportunityplatform.co.uk/api/v1/labour/course-demand?region=E12000007&limit=1" \
  -H "Authorization: Bearer $OP_API_KEY"

{
  "data": [
    {
      "source_key": "dfe_apprenticeships",
      "metric": "apprenticeship_starts",
      "region_code": "E12000007",
      "subject_area": "Construction, Planning and the Built Environment",
      "period_end": "2026-07-31",
      "value": 3120,
      "year_on_year": { "previous_value": 2890, "change_pct": 8 }
    }
  ],
  "sources": [
    {
      "key": "dfe_apprenticeships",
      "publisher": "Department for Education",
      "licence": "Open Government Licence v3.0 (Crown copyright)",
      "attribution_statement": "Contains public sector information licensed under the Open Government Licence v3.0."
    }
  ],
  "attribution_notice": "If you show, republish or pass on any figure from this response, ...",
  "meta": { "as_of": "2026-07-31", "returned": 1, "limit": 1, "next_cursor": null }
}

Attribution: show it and pass it on

Wherever you show, republish or pass on a figure, show the attribution_statement of each source it came from, word for word. If you give the data to someone else, they take on the same rule. The statements are:

Small numbers and missing figures

Counts of people are never shown below 5: 1 to 4 is the string "<5", never 0 and never a number. Expect a string where you expect a number and handle it. A published statistic with no figure (the publisher suppressed it or did not publish it) is null. It is never a stand-in for zero.

sources_agreeing on skills gaps counts distinct published sources pointing to a gap for an occupation. A news headline is never counted.

Regions

/v1/labour/regions lists official ONS geographies under the Open Government Licence v3.0, searchable by name. Travel to Work Areas are the default because they are drawn from where people actually commute, so work is read across council lines the way people travel. Nothing about individual vacancies is returned.

Later

Built, and not open yet. Each waits on one thing:

To hear when one opens, get in touch.

Paging

/v1/labour/skills-unmet and /v1/labour/evidence page with meta.next_cursor, exactly as jobs do: send it back unchanged as cursor with the same filters. Skills not met pages by rank. The other endpoints return one bounded list, and their next_cursor is always null.

Questions