Labour market intelligence: what it is, who sells it, and how to get it (2026)

Labour market intelligence (LMI) is the analysis layer built on top of jobs data: which occupations are in demand and where, which skills are rising and falling, what work pays, how hiring differs by industry and region, and where the supply of people meets or misses it. Governments use it to steer training policy, universities to align programmes with demand, employers to plan workforces and pay, staffing firms to find clients, and product teams to build everything from job boards to career-guidance tools. This page covers what LMI actually consists of, the data it is built from, who sells it, and how the field changed when job postings became measurable at scale.

What labour market intelligence answers

Strip away the vendor language and LMI answers a short list of questions. Demand: which occupations and industries are hiring, at what volume, trending which way. Skills: what abilities and tools employers ask for, and which are emerging or declining inside each occupation. Pay: what work is offered and paid at, by role, seniority and geography. Supply: how many people hold the relevant skills, and where. Transitions: which roles people move between, and what retraining bridges the gaps. Any LMI product - a national dashboard, an enterprise platform, an API - is some packaging of these five.

The data underneath

Three families of sources feed every LMI product, and their trade-offs define the field. Official statistics (labour force surveys, vacancy surveys like the US JOLTS, census and administrative data) are authoritative and comparable across time - and aggregate, slow, and silent on skills. Online job advertisements - the OJA data the EU’s skills agency Cedefop builds on - are the opposite: posting-level, skill-rich, near real-time, but they need heavy deduplication and liveness work and represent advertised demand, not all hiring. Profiles and CVs supply the people side, with known skews toward white-collar, profile-active workers. Serious LMI triangulates: official series for levels, postings for timely detail, profiles for supply.

The classification spine matters as much as the sources. Occupations are compared through taxonomies - SOC in the US, ISCO-08 internationally - industries through ISIC or its regional variants, and skills through skills taxonomies like ESCO and Lightcast Open Skills. Without the spine, a million postings are anecdotes; with it, they join official statistics and become measurement.

Who sells it

The tiers differ less in what they claim than in how you buy them. The enterprise platforms bundle deep history, taxonomy work and dashboards behind procurement (we compare them in talent intelligence platforms and Lightcast alternatives); the free official layer is the baseline every analysis should anchor to; and the API tier exists for teams that want the raw measurements inside their own products and models rather than in a vendor’s dashboard.

Measured, not modeled

The structural shift of the last decade is that labour demand became directly measurable. When postings are collected at the source, deduplicated into vacancies, classified to occupation and industry, and re-checked for liveness, questions that once required a survey cycle - is demand for a skill rising, what does this occupation advertise in pay, how exposed is an industry to AI skills - become queries over live data. That is the layer JobsPipe exposes: postings classified to ISCO-08 and ISIC with extracted skills, and a Labour Market Insights API that serves occupation and industry rankings, advertised-salary percentiles, trending skills and AI-exposure measures computed from the corpus at request time. The honest caveat applies to every OJA-based product including ours: advertised demand is a leading indicator of hiring, not the whole of it - which is why it belongs alongside official statistics, not in place of them.

Getting started

If you are evaluating the field: anchor on the free official sources for your geography, read our nine-source comparison for what each layer measures, and prototype against a self-serve API before committing to an enterprise contract - the questions your product actually asks will tell you which tier you need far better than a sales demo will.