Workforce Reskilling Needs Employer Demand Data

Reskilling programmes are judged by enrolment and completion. Employers hire on evidence of capability. The two datasets rarely meet.

The short answer: build the demand side first, from employer hiring and task data, and only then size the supply of courses.

Evidence note: The ILO and the World Bank both treat skills systems as labour-market infrastructure. Their programme evidence shows the same pattern: training without an employer demand signal produces certificates, not hires.

Data layer	Question it answers	Common failure
Vacancy text	Which tasks employers actually buy	Pasted boilerplate job ads
Hiring outcomes	Who gets hired from which pipeline	Tracking enrolment instead of placement
Task change	Which skills are replacing which	Assuming job titles are stable
Employer commitments	Who will interview the graduates	Policy pledges without interview dates

Enrolment is a supply number

Most reskilling dashboards count what is easy: seats filled, modules completed, satisfaction scores. Those are supply-side measures. They say nothing about whether the trained skill was demanded in the same geography and quarter.

**A completion certificate is inventory.** It only becomes a market outcome when an employer hires against it. Until placement is measured, the programme is a cost centre with good reviews.

The fix is unglamorous: build the demand file before the curriculum. Vacancy text, task lists, and named employers willing to interview are the raw material.

Mining demand from hiring behaviour

Job adverts are noisy but honest about tasks. Ten adverts for the same title often describe three different jobs. Grouping adverts by tasks rather than titles gives a cleaner picture of what is actually purchased.

Hiring outcomes close the loop. Track which training pipelines produced hires, at what wage, and how long the placement lasted. A pipeline that places 80 percent into short contracts is different from one that places 40 percent into permanent roles.

Publish both. Programmes that hide placement mix are usually hiding dilution.

Task change beats title change

Job titles are sticky; tasks are not. A maintenance technician role gains diagnostics and loses mechanical fitting over years while the title never moves. Skills demand built on titles misses this drift entirely.

The practical method is a task census: list the recurring tasks in a role, mark which are growing, stable or shrinking, and train against the growing list. Employer interviews validate the ranking.

This is also how to detect fake demand. If nobody can name the growing tasks with examples from last month, the skill is a talking point, not a market.

What employers owe the system

Employers ask for job-ready graduates and rarely commit to anything verifiable. The credible unit of commitment is an interview slot with a date, not a memorandum of understanding.

Where employers co-design assessments, placement rises because the test matches the work. Where they only attend launch events, the training drifts toward generic content.

Market analysts should score programmes by the share of curriculum mapped to named employer tasks. That single ratio predicts placement better than most accreditation badges.

Sizing the reskilling market without wishful numbers

The addressable market is the gap between tasks employers are hiring for and tasks current workers can evidence, priced at what employers actually pay for bridging. Anything larger is aspiration.

Separate compliance-driven training, which renews on schedule, from capability-driven training, which follows task change. They have different cycles and very different price elasticity.

Date the estimate. Task demand shifts with technology adoption, so a reskilling market size without a demand vintage is a marketing number.

Teams that need a consistent cross-market view, rather than one clip of data at a time, often pair this kind of desk check with independent market intelligence so every conclusion carries its source and date. The point is not another report. It is a method that survives the next quarter.

What the data cannot tell you

Hiring data under-represents internal mobility. When a company reskills an existing employee, no vacancy is posted and no market signal is emitted, yet that is exactly the behaviour employers say they prefer. Treat external hiring data as a floor on demand, not the total.

Wage data needs the same caution. Posted wages are offers, not outcomes, and in tight segments the advert is stale before the salary moves. Pair posted wages with placement salaries from programme records whenever both exist.

And beware the pilot trap. A small demonstration cohort with generous support can place beautifully and prove nothing about scale. Ask what the placement rate looks like at the third, unspecial cohort, where the employer relationship is routine and the hand-holding is gone.

Who this analysis does not help

It will not help anyone selecting a personal career. Individual choice turns on aptitude, interest and local openings that no aggregate demand file captures.

It is also not an accreditation review. Curriculum quality, instruction and student support are separate disciplines, and a demand-aligned course taught badly still fails its learners.

A quarterly desk routine that works

Choose the role families you serve and keep them fixed for a year, so cohort and demand comparisons survive longer than a quarter's enthusiasm.

Each quarter, refresh the task map from new vacancy postings, mark tasks that appeared or disappeared, and compare against the curriculum's task coverage. Every gap is either a course update or a decision not to train it, and both should be written down.

Pull placement outcomes for cohorts finishing in the quarter and split them by employer: new relationships, repeat relationships, and conversions from interviews to hires. The repeat-relationship share is the healthiest single number in the file.

Close with one paragraph per role family naming what changed in demand and what the programme will do about it. Vague paragraphs here are how three-year drift becomes permanent.

Rule of thumb: track the share of curriculum mapped to named employer tasks. That one ratio predicts placement better than any accreditation badge on the wall.

Frequently asked questions

Why do reskilling programmes fail to place graduates?

Usually because the curriculum was built from course availability, not from employer task demand with named interviewers.

What is the best demand indicator?

Grouped vacancy text plus hiring outcomes from the same geography. Adverts show what is bought; placements show what worked.

How often does task demand change?

Faster than titles. Review task maps annually for technical roles and after every major technology adoption.

Should government fund demand data first?

Yes. A shared, current task-demand dataset is cheaper than repeated failed cohorts and benefits every provider.

Can online course enrolment data substitute for employer demand?

No. Enrolment shows what learners believe is worth buying. Vacancy tasks and hiring outcomes show what employers actually pay for, and the two diverge regularly.

Is a skills taxonomy necessary before starting?

A light one, yes. Even a shared list of twenty tasks per role family beats comparing course titles, which is how most regional skills conversations stall.

How large must a demand sample be before acting on it?

Large enough to be stable quarter to quarter. If the task ranking reshuffles every refresh, the grouping is wrong, not the market.

Do remote-first employers change the geography rule?

They widen the hiring pool but not the working conditions. Track where the work is performed, because relocation and time-zone constraints still bind most roles.

Sources and method

This article uses the following public sources. Figures retain the source definition and date. It is market analysis, not investment, legal or medical advice.