evidoria

← Back to browse

Good practice Imported

Big Data for Labour Market Intelligence in Tunisia — Machine-Classified Job Vacancies for VET Planning

Tunisia · Tunis · See the Tunisia profile · See the Tunis profile

Evidence: Descriptive / self-reported Top 56% 47/100 · Ask Evidence Copilot about this practice

The EU's European Training Foundation scrapes Tunisian online job ads and classifies them with machine learning into ISCO, ESCO, NACE and ISCED codes, feeding a public dashboard meant to steer VET curricula. The methodology and its biases are published; the effects are not.

37,535
Tunisian job vacancies analysed in the initial working paper (1 Apr-30 Sep 2020)
52% >19,000 ads
Share of vacancies from TANITJOBS
15% ~5,500 ads
Share of vacancies from JORA
13% ~4,800 ads
Share of vacancies from KEEJOB
6% ~2,200 ads
Share of vacancies from EMPLOINAT
26%
Advertisements not stating a contract type
42.4%
Permanent contracts at 'Professionals' skill level
31.5%
Permanent contracts at lower skill levels
23,000,000
Total vacancies collected programme-wide (since 2019, as of June 2025)
10,000,000
Vacancies retained after deduplication (as of June 2025)
6 Egypt, Morocco, Tunisia, Kenya, Ukraine, Georgia
Countries covered
5
Capacity-building programmes delivered (2019-2024)
45
Training videos produced
Big Data for Labour Market Intelligence in Tunisia — Machine-Classified Job Vacancies for VET Planning

Details

Maturity
Established
Promoter
European Training Foundation (ETF)
Period
2019–ongoing
Keywords
labour-market intelligence, machine learning, natural language processing, VET planning, skills anticipation, online job vacancies

Context

Big Data for Labour Market Intelligence is run by the European Training Foundation (ETF, an EU agency headquartered in Turin) for Tunisia's Ministère de la Formation Professionnelle et de l'Emploi, scraping online job vacancies and classifying them via NLP/machine learning into ISCO-08, ESCO 1.1.1, NACE and ISCED-2011 codes to feed public dashboards for skills anticipation and VET planning. Tunisia was one of the first two full-scale pilots, live since 2020.

Objectives

Provide skills-anticipation data to steer vocational education and training provision, mirroring Cedefop's Skills-OVATE methodology so results are comparable across countries.

Activities

Technical partners Lightcast and CRISP (Università degli Studi di Milano-Bicocca) built the classification pipeline. The ETF's Tunisia working paper (16 November 2020) analyses 37,535 vacancies collected between 1 April and 30 September 2020 (about 52% from TANITJOBS, 15% from JORA, 13% from KEEJOB and 6% from EMPLOINAT). By June 2025 the wider programme had expanded to 23 million vacancies collected (10 million after deduplication) across six countries — Egypt, Morocco, Tunisia, Kenya, Ukraine and Georgia — with five capacity-building programmes delivered between 2019 and 2024 and 45 training videos produced.

Results

26% of advertisements did not state a contract type; permanent contracts accounted for 42.4% of advertisements at the 'Professionals' skill level versus 31.5% at lower skill levels. Nothing published shows how many Tunisian VET programmes, qualifications or guidance services were actually changed because of the dashboards.

Conclusions

This is a well-documented, honestly self-caveated data infrastructure — the ETF discloses its own sampling bias, the exclusion of self-employment, and its dependence on specialist expertise — but its downstream educational effect remains unmeasured, and operational ownership inside Tunisia could not be confirmed, risking the system remaining donor-run rather than institutionalised.

Implementation
Indicative cost
High (€500k–€5M) — Run by an EU agency (ETF) with specialist technical partners (Lightcast, CRISP/Milano-Bicocca) across six countries; likely a high-cost specialist data-science infrastructure, though no line-item budget is published.
Time to results
Long (> 3 years) — Live in Tunisia since 2020 with monthly data collection continuing as of 2025; the wider six-country programme has run since 2019 with capacity-building through 2024.
Staffing & skills
European Training Foundation (ETF) — programme ownership, Lightcast (technical partner), CRISP, Università degli Studi di Milano-Bicocca (technical partner), Tunisia's Ministère de la Formation Professionnelle et de l'Emploi (national counterpart)

Conditions for success

  • Specialist NLP/machine-learning expertise and technical partners to build and maintain the classification pipeline
  • Alignment with an established methodology (Cedefop Skills-OVATE) for cross-country comparability
  • Use of published, standard classification schemes (ISCO-08, ESCO 1.1.1, NACE, ISCED-2011) for interoperability
  • Capacity-building programmes and training videos for national analysts and ministry staff

Common failure modes

  • Online-vacancy sampling bias — some sectors are over-represented, others entirely absent
  • Web advertisements are not representative of self-employment, a serious gap in a labour market with large informal/self-employed segments
  • Operational ownership inside Tunisia could not be confirmed, risking the system remaining donor-run rather than institutionalised
  • No published evidence of how many VET programmes or qualifications were actually revised as a result of the dashboards

Replication kit

Reusable artefacts from this practice — as published by their sources.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful