About our data | BeforeYouFindAJob

Where the data comes from

Every figure is published by the Office for National Statistics under the Open Government Licence v3.0. This site loads four ONS releases and one classification:

Online job advert volumes by occupation: Advert counts by SOC 2020 occupation, month and UK region.
Skills, competencies and job requirements: The share of new adverts mentioning each requirement.
Online job advert salaries: Advertised salary statistics, where the publisher reports them.
Vacancy Survey (VACS02): The official vacancy estimate, shown only as context.
SOC 2020 coding index: The official job-title-to-occupation classification.

This site adds nothing to those releases beyond arranging them. It does not survey anyone, collect adverts, or buy data.

An online job advert is a posting collected from a job board or recruitment page. Textkernel, a commercial data provider, collects adverts from roughly 90,000 such sites. ONS removes non-UK and non-English adverts, deduplicates, assigns geography, classifies each advert into an occupation, and publishes the result.

An advert is not a vacancy and not a job. One advert can be several positions, a speculative opportunity, a training route, or a role also advertised elsewhere. In August 2024 ONS recorded about 1.43 million deduplicated adverts against 860,000 vacancies in the official Vacancy Survey for the same period, roughly 66 per cent more adverts than vacancies. Long-term movements were broadly similar; the levels were not.

That is why this site never states how many jobs are available. It does not know, and neither does the source.

The ONS and Textkernel boundary

Textkernel collects the adverts and derives fields including location, salary, skills and industry. ONS then applies its own processing and publishes the statistics. The published output is ONS's; the underlying collection is Textkernel's.

This site uses only the ONS-published outputs. It has no access to the underlying Textkernel corpus and cannot inspect, re-derive or correct it. Where ONS labels its output as official statistics in development, that status carries through to here.

How deduplication is performed, upstream

Textkernel identifies related postings across sources, and ONS retains the earliest when duplicates overlap. In ONS's 2017 to 2024 analysis, each unique new advert had on average around two further adverts identified as duplicates.

ONS acknowledges that this is imperfect in both directions: some duplicates remain, and some distinct adverts may have been classified as duplicates. This site inherits ONS's deduplicated counts and does not attempt to reconstruct the corpus or estimate a "true" number of postings behind them.

How occupations are classified, and where that fails

Occupations follow SOC 2020, the UK's standard classification: 9 major groups, 26 sub-major, 104 minor and 412 unit groups. A job title is matched to an occupation through the official coding index, which holds 32,376 titles.

The limitations are substantial and are not hidden here:

ONS manually labelled 1,000 adverts and measured 65 per cent precision when coding advert titles to four-digit SOC automatically, and 74 per cent at two-digit. Roughly one in three detailed classifications is wrong.
2,439 index titles code into more than one occupation. "Analyst" appears under five; "Manager" under fifty-one. Where the index supplies a qualifier, this site shows it, because that qualifier is what makes the entry true.
231 titles, Director, Engineer, Supervisor, Teacher, Trainee and similar, ONS marks as impossible to code from the title alone. This site excludes them rather than guessing.
Because a search term can be genuinely ambiguous, this site offers every plausible occupation and asks you to choose. It does not pick one for you.

How geography is assigned

Adverts are counted against the location stated in the advert, which is not necessarily where the work happens: a remote or hybrid role advertised from a head office counts to that office's region.

This site reports the twelve UK regions, which is the finest grain the monthly series publishes by occupation. ONS also publishes local-authority figures, but on a quarterly grain, showing those beside a monthly advert count would put two different periods on one page, so they are not used.

Regional figures do not sum to the UK figure. They are published separately rather than derived from it. Across the 264 occupations in the latest month the difference reaches 4.5 per cent, and for one occupation the regions total more than the UK figure. Both published totals are shown so the gap is visible rather than reconciled away.

ONS also publishes adverts it could not assign to an occupation, recorded as "Unknown". These are excluded from occupation pages because they belong to no occupation, not because they do not exist.

Why comparisons are refused more often than they are made

Two published numbers can differ for two entirely different reasons: the labour market moved, or the way the number was produced changed. Presenting the second as the first is the most damaging mistake this product could make, so a comparison must be established rather than assumed.

ONS documents changes in which sources it collects from. Three sit inside this series:

July 2023, a major source (CareerBuilder) largely disappeared from the collection.
January 2024, a source was reinstated.
May 2024, a large source was added.

A change in the number of adverts across one of those dates may be a change in what was collected rather than in the market. Separately, the 2020 to 2021 period required expiry-date imputation after a source-system issue.

Any comparison spanning a documented break or the imputed period is reported as not comparable. There are no waiting periods and no minimum record counts, the question is whether two figures are the same measurement, not whether they are large ones. And two observations can, at most, show a difference: a trend is a claim about a series, and no amount of comparing two points produces one.

Why each measure has a different latest date

The four releases are published on different schedules, so they do not describe the same moment. At the time of writing, advert counts run to June 2026 while advertised-salary evidence ends in May 2025, a gap of thirteen months, on every occupation.

This site therefore never shows a single page-level "updated" date. Each measure carries its own evidence period, and where salary evidence is older than the advert count beside it, the page says so. One shared date would let the most recent measure make the oldest look current.

Where ONS publishes a revised edition, the revision is treated as a reason not to compare across it rather than as a silent correction.

How skills and salary are represented

Skills and requirements

These are shares of adverts, not counts of jobs: the proportion of new online adverts for an occupation that mentioned a requirement. The taxonomy is Textkernel's SCO, which is not SOC and does not map onto it.

ONS breaks requirements down by occupation for the latest year only. Every earlier year in the same release covers the whole economy and carries no occupation, so there is no earlier figure for any occupation to compare against, which is why this site reports that comparison as unavailable rather than making one.

Advertised salary

Advertised salary is not pay. It is what adverts stated, and adverts that state a salary are a self-selecting subset of all adverts. It says nothing about what anyone is actually paid, what was negotiated, or what the role is worth.

What this site derives itself

Almost nothing, and deliberately. The published figures are shown as published. What this site adds is arrangement and refusal:

Selecting the latest published period for each measure and stating it beside the figure.
Expressing each region's share of the regional total, with both published totals shown.
Counting how many occupations each index title codes into, so shared titles can be marked.
Deciding whether a comparison may be made at all, and refusing where it may not.
Attaching a limitation to every number, in the same place as the number.

No figure on this site is modelled, forecast, smoothed, seasonally adjusted by us, or imputed by us.

What this site deliberately does not claim

How many jobs are available. These are advert counts, and adverts are not vacancies.
What anyone is paid. Advertised salary is not pay.
That an occupation is growing, shrinking, secure or threatened. Those are claims about a market, and no count of adverts establishes one.
What employers currently call a role. The coding index is a classification tool, not evidence of usage or frequency.
Which employers are advertising, or anything about an individual advert.
Whether an occupation suits you. That is not a statistical question.

Where the evidence cannot answer something, this site says so rather than filling the gap. Search an occupation

Licence

Contains public sector information licensed under the Open Government Licence v3.0.