Good Job

The Missing Bridge to Work: What Tunisia’s FBR@WORK Programme Is Revealing About Skills and Jobs That Usually Stay Hidden 

Amel Karboul

CEO, the Education Outcomes Fund

We launched FBR@WORK in Tunisia with an ambitious hypothesis: could a results-based model help unemployed young people move into stable work at scale? In this first phase, the programme aims to support at least 2,000 young people aged 18 to 35 who have been out of work for at least one year into sustainable employment, with a minimum of 50% women, specifically into formal open-ended contracts (CDI) with social security rather than short-term or subsidised jobs. The premise sits inside the programme’s own name. FBR, or financement basé sur les résultats which admittedly sounds slightly more sophisticated in French, was meant to test whether tying financing to actual employment outcomes would reveal something different from the usual skill-gap training logic.   

What the first months of implementation suggest is not that skills do not matter. It is that the conventional diagnosis may be too narrow. If the dominant story is “there is a skills gap, therefore we need more training,” the early lessons from FBR@WORK suggest a more uncomfortable possibility: for many unemployed young people, the real bottleneck may be less the absence of training in the abstract and more the need for a dedicated, personalised support capable of sourcing, screening, preparing, matching, and supporting them into actual jobs. Public employment agencies play a central role in this but as evidence suggests, they increasingly need complementary partners to extend their reach and deliver the kind of tailored, employer-specific pathways that standardized public service provision cannot easily accommodate¹. 

That is a very different problem, and it points to a very different way of designing skills policy. These are first signals from four months of live implementation, the kind of early operational evidence that rarely makes it into policy debates, but that is worth surfacing precisely because it challenges assumptions before they harden into conventional wisdom. 

And this matters well beyond one programme in Tunisia. Governments everywhere are facing a harsher equation: tighter fiscal space, growing pressure to create jobs for young people, and labour markets being reshaped by economic disruption and new technologies, including artificial intelligence. The International Labour Organization (ILO) estimates that in 2025 around 262 million young people worldwide were neither employed nor in education or training. 

Against that backdrop, the real elephant in the room may be bigger than the design of standalone skills programmes. It may be the much larger sums we already spend on higher education, TVET, and training systems that are still too often judged by enrolment, completion, or credentials rather than by whether they help young people make a credible transition into work. The uncomfortable question is whether, in too many countries, we are still financing education systems as if credentials were the goal, when what young people increasingly need is a credible bridge into work. Six early lessons from FBR@WORK Tunisia suggest what that could mean in practice. 

1. The first lesson is not that training is irrelevant. It is that training as usually designed is often misaligned with labour market reality. 

One of the clearest signals in the first months concerns training length. One implementing partner initially designed a training model of around 320 hours, later reduced to 250 hours. Even then, employers still considered 150 completed hours too long. Ultimately, the programme had to adapt further and allow participants to begin placement after 40 hours of training, with the remainder continuing post-placement.   

That is not a minor operational tweak. It is a strategic finding. 

For years, much of the skills world has assumed that longer training signals seriousness, depth, and quality. But the FBR@WORK experience suggests that in labour markets where employers are hiring against real and urgent vacancies, long front-loaded pre-employment training can become a liability. It slows entry into work, mismatches employer urgency, and ignores the financial realities of young people who cannot afford to sit in unpaid training for months. 

The provocation here is not “shorter is always better.” It is more precise than that: training length should be determined by actual labour market function, not by provider habit or pedagogical convention. If employers need people quickly, and if participants need income quickly, then systems built around long pre-placement training may be solving the wrong problem in the wrong order. 

2. The real value of a provider may lie less in being a trainer and more in being a complementary workforce integration actor functioning as a support partner to public employment agencies such as ANETI. 

One of the most interesting features of the report is that the more promising models did not look like classic classroom-based training providers. They looked much more like specialised workforce integration and placement partners. 

In the first months implementing partners mapped employers, identified live vacancies, engaged human resources teams from the outset, co-designed training content, pre-selected candidates, involved employers early in interviews, and adapted training to actual recruitment cycles and workplace needs. In one case, the training with one company was explicitly co-designed around their hiring needs and included both technical and soft skills aligned with the role.   

This matters because it changes how we think about what governments and funders are actually buying. 

If the real function of a good provider is not merely to “deliver skill gap training” but to bridge the distance between disadvantaged youth and employer demand, then the provider is doing much more than education. It is helping employers outsource the hard work they often do not have the time, systems, patience, or incentives to do themselves: finding the right people among the group of unemployed youth, preparing them for a specific job, reducing recruitment risk, and supporting onboarding and early retention. 

That said, this model does not operate in the same way across all market contexts. It is far more scalable and cost-efficient where employers recruit in cohorts, making joint upfront selection feasible and efficient. Where employers tend to hire only one or two profiles at a time, the same approach becomes harder to operationalise, more resource-intensive, and less able to generate volume quickly. 

The gap being bridged here is not just a skills gap. It is also a recruitment gap, a trust gap, an onboarding gap, and in some cases even a motivation and signalling gap. FBR@WORK suggests that the most effective skills interventions may be those that quietly function as an extended HR and workforce integration arm for employers, supporting public employment agencies in fulfilling their intermediary mission. 

3. The sequence “train, then place, then retain” may be one of the sector’s most overrated assumptions. 

The original sequential design of train → place → retain proved too rigid and did not respond well enough to labour market needs. Allowing placement to begin before full training completion improved alignment with employer demand and made the model more responsive to urgent recruitment needs. 

This matters because so much of the sector still assumes that employability must be built first and only then tested in the labour market. But real labour markets do not behave that neatly. Employers often hire for partial readiness and train on the job. Young people often learn faster once employment is real rather than hypothetical. And motivation changes when a pathway leads to actual income rather than the promise of eventual opportunity. 

What FBR@WORK points toward instead is a more integrated model: some preparation before placement, but not too much; then structured learning that continues into employment; with enough flexibility to adapt to employer demand without abandoning quality. Shorter, targeted training formats and models closely integrated with employment, including apprenticeship-style approaches or phased training that continues after placement, are beginning to look more effective.   

This should challenge ministries and funders to rethink not only content, but sequencing. Perhaps the right unit of analysis is not the course, but the transition into work. 

4. Signaling matters more than the sector often admits. 

Another gold nugget is deceptively simple: EOF helped one partner improve sourcing by ensuring that job descriptions and the names of recruiting companies were communicated upfront, whereas this information had previously been withheld. The result was better filtering of unsuitable applicants and a more streamlined selection process.   

That is a small operational detail with large strategic implications. 

Much of the employability space still markets programmes in vague language: training opportunities, capacity building, empowerment, readiness, exposure. But for young people, especially those who have been unemployed for a year or more, vagueness is often not attractive. It creates mistrust, low commitment, and weak self-selection. What they need to know is: what is the job, who is the employer, where is it, what does it pay, and why should I believe this will lead somewhere real? 

This is not just communications. It is part of programme design. 

When opportunities are concrete, candidate behaviour changes. When employers are visible, provider credibility rises. When requirements are explicit, unsuitable applications fall. In other words, better signalling is not cosmetic. It is part of labour-market efficiency. 

Additional examples from implementation reinforce this point. In some cases, candidates only discovered after training that jobs involved late-night shifts or Saturday work and then declined the offers. In both cases, the placement failed not because of a skills gap, but because of an information gap. One partner highlighted radical transparency from the first contact as a key success factor: young people are told upfront about working hours, employer expectations, the non-accredited nature of the training certificate, and the explicit objective of the programme—training for employment, not for credentials. 

5. If governments want sustainable reform, the learning agenda cannot be a side activity. It has to serve a policy adoption pathway. 

One of the strongest institutional lessons from the first months is about the learning agenda. A key early lesson was the importance of first understanding what government expects in terms of legacy and sustainability, and only then designing the learning agenda to generate actionable knowledge in support of that vision. 

This is a much deeper point than it looks. 

Too many programmes treat learning as something parallel: a set of studies, evaluations, dashboards, and lessons captured because donors expect it. FBR@WORK suggests a better discipline. The first question is not “what would be interesting to learn?” The first question is “what would government need to know in order to adopt, adapt, fund, regulate, or scale this differently?” 

Once that becomes the anchor, the learning agenda changes shape. It becomes about comparative effectiveness versus existing programmes, legal frameworks for outcomes-based contracting, institutional readiness, ecosystem capacity, and the practical conditions under which a ministry or public agency could genuinely own the model. 

But that sequencing is only the starting point. Learning priorities do not get fixed once and for all at the beginning of a programme; they continue to evolve throughout implementation as the context shifts, political priorities change, and new operational questions emerge. Even in the first months, government counterparts have already surfaced additional areas of interest, from youth mobility between place of residence and job location to more detailed information on training content, including the balance between hard and soft skills and the place of digital and language skills. That early alignment with government systems is a positive signal: it strengthens the relevance of programme learning in real time while also increasing the chances that the knowledge generated will support long-term sustainability and institutional ownership. 

That is the right sequence, and the right mindset: first anchor learning in the government’s long-term goals, then keep refining it as implementation generates new questions and priorities. It is a discipline many funders still miss. 

6. Verification, public-private coordination, and data systems are not back-office details. They are part of the model. 

FBR@WORK also highlights something the skills world often underestimates: if you want to fund for outcomes rather than activities, you need serious operating infrastructure. 

Key has been the independent verification, CNSS (the national social security fund), backed eligibility checks, alternative manual verification, memorandums of understanding with ANETI (the national public employment agency), communications protocols, and a shared data platform that supports not only evaluation but also sourcing and operational decision-making. Government stakeholders were interested enough in the platform that they asked whether it could be leveraged for other programmes as well.   

That matters because many skills systems still treat data, verification, and coordination as administrative overhead. FBR@WORK suggests the opposite. These are not peripheral. They are part of what makes a labour-market intervention credible, scalable, and governable. 

Another key success factor is that the programme also creates the right incentives for the State. ANETI is highly cooperative, especially at the local level, in part because the programme helps it deliver against results that are themselves linked to wider state performance incentives. 

The implication is important for funders in particular. If you want demand-led, outcomes-oriented, government-embedded skills systems, you cannot fund only delivery. You also have to fund the architecture: data systems, verification tools, coordination mechanisms, government interfaces, and enough structured governance to let the model function without collapsing into improvisation. 

What this means for the future of skills policy 

The first implication is that ministries of employment and ministries of finance may need to ask a more disciplined question when looking at youth employment spending: are we paying for hours of instruction, or are we paying for transitions into work? 

Those are not the same thing. 

And this is not a marginal issue. In fiscally constrained systems, the question is no longer only how to improve the design of a few labour-market programmes at the edge. It is whether much larger public investments in higher education, TVET, and workforce systems are being organised around the wrong outcomes altogether. 

FBR@WORK suggests that once the target population is genuinely difficult, in this case young people unemployed for at least a year, the old assumption that more training hours will solve the problem starts to break down. The more relevant questions become: how tightly is the intervention linked to real employer demand? How quickly can it respond to vacancies? How credible is the matching process? How much support exists during onboarding? And what kind of public-private coordination makes all of this function at speed without sacrificing integrity? 

But there is another layer here that labour-market programmes often overlook: even when a young person is ready to move into work, the wider social policy system can create incentives not to do so. In Tunisia, some young people have declined formal job offers because low-paid employment would have caused them to lose benefits linked to inactivity, such as free healthcare or other forms of support, leaving them little better off, or sometimes worse off, despite being employed. I was reminded of a conversation with a partner government in another context, where young people with disabilities risked losing disability-related benefits once they entered work, even though employment did not reduce their need for additional transport and other support. The policy was eventually adjusted, but it illustrates a broader point: social protection systems can unintentionally penalise entry into work if they are not designed carefully.  

A separate but related issue concerns the incentives built into different contract types. In Tunisia, some trained youth have used CIVP, the state-subsidised employment scheme, as a bargaining tool, because the additional allowance could leave them better off in net terms than a CDI at the same gross salary, even if the CDI would offer stronger long-term protection through social security contributions.  

In other words, even when we build the bridge into work, the wider system may still stop young people from crossing it. That is why employment policy and social protection policy cannot be designed separately. 

This does not argue for less focus on skills. It argues for a different organising logic: one centred on pathways into work, not just training provision. The bolder conclusion is that many governments may still be overinvesting in training as a standalone product, while underinvesting in the labour-market intermediation and system design needed to turn learning into employment. 

What funders should stop funding, start funding, and fund differently 

Stop funding long, supply-driven training programmes whose link to real hiring demand is weak, delayed, or mostly rhetorical. A certificate is not automatically better than nothing. If training does not lead to a credible pathway into work, it can leave young people more frustrated, more distrustful, and sometimes financially worse off than before, especially where participation means giving up informal income or other opportunities. The bar for what counts as beneficial training should therefore be much higher: we should back it when there is a real chance it will lead to meaningful improvements in people’s lives, and recognise that when it does not, it may create more harm than good. 

Start funding models that combine targeted preparation with employer engagement, live vacancy matching, recruitment support, onboarding, and retention. 

Fund differently by recognising that the most valuable part of a provider’s work may not be the classroom at all. It may be the bridge: identifying suitable candidates, screening for fit, motivating participation, adapting preparation to employer needs, coordinating with public systems, and staying close enough to the employment relationship to reduce drop-off in the early months. 

And fund the infrastructure around that bridge. Verification. Data systems. Public-private protocols. Government capacity. Learning agendas that are designed to influence adoption decisions. 

If that sounds less glamorous than another fancy skills programme, it is. It is certainly less satisfying for funders or governments who would rather report thousands trained than a smaller number of young people meaningfully placed into jobs, alongside the data and system infrastructure needed to make those transitions work. It may nevertheless be far more useful. 

Conclusion 

So the lesson from FBR@WORK Tunisia is not a lazy provocation like “training is dead.” That would miss the point. 

The more precise lesson is harder, and more interesting. For young people who have already been locked out of work, the main barrier may not be the absence of skills alone. It may be the absence of a system capable of converting potential into employment through real employer demand, faster preparation, better matching, and more intelligent support. 

And that question is only becoming more urgent. The ILO has noted that younger workers are often more exposed to occupations at greater risk of automation and AI-driven restructuring, making the quality of the transition into first work even more important.   

Perhaps the real policy failure is not that we underfund skills, but that we still overfund systems that produce credentials without credible transitions into work. In an age of fiscal constraint, youth frustration, and technological disruption, the question is no longer whether to reform skills programmes at the margins. It is whether we are willing to rethink the much larger education-to-employment bargain altogether. 

¹ Gomez, S., de Moraes, G., Petrelli, A., Soytas, M.A. and Rivera, N. (2026) Unlocking Opportunities: Exploring the Potential of Public-Private Partnerships in Employment Services Delivery. SPJ Policy & Technical Note. Washington DC: World Bank Group. 

Photo credit: EFE Tunisia, SAGE Training, September 2025