I started this project after meeting a family that had mortgaged their house to pay a recruitment agency that then failed to deliver the apprenticeship it had promised their daughter. That single case is not evidence of anything at scale — but it's why I built what follows.
Epistemic status: Rough, first-pass estimate based on my own platform's data and public World Bank / German government figures, not an academic study. I built and run this platform myself; I am not a disinterested researcher, and I'd genuinely welcome pushback on the assumptions below. Numbers marked (?) are the ones I'm least confident in.
International educational migration to Germany has an established track record in this community — see Malengo's shallow investigation on university migration, since funded by GiveWell and Open Philanthropy. Vocational (apprenticeship) migration is the same underlying mechanism — moving a person from a low-income labor market to a high-income one — but with a shorter pathway (3 years vs. a full degree), a lower bar to entry (no university admission required), and, in the version I've built, a near-zero marginal cost per candidate reached (~$15-25 vs. Malengo's ~€12,000/student), because the intervention is information and direct matching rather than a cash stipend. The tradeoff: I don't yet have rigorous data on what share of registered candidates actually complete an apprenticeship, which is the biggest open uncertainty in this post.
Germany has a structural shortage of apprentices and offers apprentice wages that are unremarkable by German standards but very high relative to average incomes in much of Africa and parts of Asia. A large, real information gap exists between German companies with open apprenticeship positions and international candidates who could fill them.
That gap is currently filled almost entirely by commercial recruitment agencies, which typically charge candidates fees in the range of €2,000+ for "placement help," frequently financed through family loans at informal interest rates far above what a German apprentice wage can service. This is a fairly clean case of information asymmetry: candidates cannot verify in advance whether an agency's promises (guaranteed placement, guaranteed visa) are deliverable, and agencies have no binding mechanism forcing them to deliver. The predictable result is a market that extracts fees from the side with the least bargaining power, with no correction mechanism, since a burned candidate rarely has recourse or public visibility to warn the next one.
This market failure is two-sided, and the second side is easy to miss. Agencies typically also charge the hiring company a placement fee. That prices out exactly the employers who most need international apprentices: small trades and craft businesses (Handwerksbetriebe) without a dedicated HR department or recruiting budget. The result is a system that both overcharges vulnerable candidates and structurally excludes the small companies with the most open positions — a lose-lose that a zero-commission, direct-matching model removes on both sides at once.
There is also a personal-responsibility dimension worth naming honestly, on both sides. Agencies sell certainty that doesn't exist — "guaranteed placement," "guaranteed visa" — to candidates who then don't have to weigh their own odds realistically, and to HR departments who outsource the judgment call of "is this a plausible hire" to a paid third party instead of making it themselves. Removing the agency doesn't just remove a fee; it puts the decision, and the responsibility for it, back where it actually belongs on both sides of the match.
Countries with large populations of German-language learners and youth un/under-employment, where I have direct partner contact: Vietnam, Kenya, Uganda, Zimbabwe, Rwanda, Morocco, India, Pakistan, Malaysia.
Some real, sourced anchors on the German side: according to Germany's Federal Office for Migration and Refugees (BAMF, "Monitoring zur Bildungs- und Erwerbsmigration"), around 8,000 people arrived in Germany for school-based or company-based vocational training (Berufsausbildung) in 2022, up 48% year-on-year. Notably, the country pattern for this specific migration category differs from the much better-known university-migration pattern: India and China, which dominate academic migration to Germany, rank only 10th–13th for non-academic education migration, while Vietnam alone accounts for close to a fifth of it, with over 95% of Vietnamese non-academic education migrants receiving titles specifically for company-based vocational training. This is one useful, sourced data point illustrating that vocational migration follows a different country pattern than the university-migration literature (including Malengo's own analysis) would suggest — it doesn't mean Vietnam is the whole opportunity; my own partner base spans Africa (Kenya, Uganda, Zimbabwe, Rwanda), South Asia (India, Pakistan), and North Africa (Morocco) as well, and I don't yet have equally granular sourced numbers for all of them.
A conservative sum across ~20 comparable countries still suggests at least 20,000 people/year who are both eligible (language level, age) and actively seeking exactly this outcome. I want to be explicit that this 20,000 figure is a weak, back-of-envelope estimate, not a derived number: it comes from extrapolating the one sourced data point above (Vietnam's ~1/5 share of an 8,000/year national vocational-migration inflow, implying a Vietnam-sized "interested pool" well above the realized inflow) across a comparable number of focus countries, rather than from country-by-country demand data I actually have. It should be read as "plausibly tens of thousands, not hundreds or millions," not as a number to build a precise cost-effectiveness model on.
I run Azubis-Now, a platform that connects candidates directly with German training companies — no agency, no commission, no loan. I built it myself (13+ custom WordPress/WooCommerce plugins) after 30 years in German HR/staffing.
Since December 2025 (~7 months): ~90 registered candidates, partner schools in Zimbabwe, Kenya, Rwanda, an institutional partner in Germany (EBG gGmbH), built with zero marketing spend, purely through SEO (average Google position 9.3, growing ~20%/week in impressions as of mid-2026).
Revenue model: candidates pay a one-time €39 membership fee (vs. €2,000+ for an agency) — not zero, but two orders of magnitude below the status quo, and I've said I'd drop this to near-zero with adequate outside funding.
Worth stating plainly: unlike many migration interventions, the primary bottleneck here isn't capital. It's information and coordination — getting the right candidate and the right small employer to find each other reliably, without a paid intermediary sitting between them and extracting rent from the fact that they can't easily find each other on their own. The WordPress build, the SEO work, the €39 fee — none of that is the actual innovation. Reducing the bottleneck to "can two sides find each other" is.
Planned scaling mechanism: a performance-based "country ambassador" model — a local, trusted person per country/region, paid a share of the membership fee and only for verifiable outcomes (via unique referral codes, the same mechanism I already use for partner schools). Ambassadors also run informational workshops. This matters beyond growth: many candidates arrive scared and distrustful, having already heard about or been burned by agency scams, and a workshop-based, in-person explanation from a trusted local face does more to counter that fear than anything the website alone can do.
Where this breaks without funding: the model currently runs on my own unpaid, full-time labor. Past a certain candidate volume, that stops being viable without additional (likely virtual/remote) staff support — this is one of the more concrete uses I'd have for outside funding, distinct from the platform's own running costs.
Cost side: Annual platform running costs (hosting, tools) are roughly $500–2,000/year(?), excluding my own unpaid labor. Over ~90 candidates in 7 months, that's a marginal cash cost of very roughly $15–25 per registered candidate. This number understates the true cost: it excludes the opportunity cost of my own full-time, unpaid labor (roughly 18 months so far), which an EA-style cost-effectiveness analysis should properly count. A fairer accounting would value that time at some reasonable wage and add it to the denominator — which I haven't done here, so treat the $15–25 figure as a lower bound on cash cost, not a complete cost-effectiveness number.
Benefit side, per successfully placed candidate:
The key open uncertainty: I deliberately do not track whether individual registered candidates go on to secure and complete an apprenticeship. Unlike Malengo's model, which has a defined admissions/support process with natural touchpoints for outcome tracking, mine is non-custodial by design — candidates apply directly to companies themselves, and I don't mediate or follow the process — and I've chosen not to build tracking infrastructure for it, on data-protection grounds (this concerns minors in several source countries, and I don't want to hold that data). This is a real, structural limitation on how confidently anyone can state a placement rate, not a temporary data gap. A rough placement-rate assumption (say, 15–30%) would still make the intervention highly cost-effective given how low the cost side is, but that number is currently an assumption, not a measurement. One thing I could realistically do without building custodial tracking infrastructure: an anonymous, voluntary follow-up survey (opt-in, no identifying data retained) to at least get a rough, self-reported placement-rate range — I haven't done this yet, but it's a clear next step if I want this analysis to hold up to scrutiny.
Malengo and Azubis-Now serve different populations toward different outcomes — a full Bachelor's degree and eventual white-collar career versus a 3-year trade qualification — and I don't think it's meaningful to rank one as "more effective" than the other from the numbers here. What I'd say more carefully: even under a pessimistic placement-rate assumption, and even after accounting for the labor-cost caveat above, the cash cost per candidate reached is low enough that this pathway probably deserves a place in the same conversation as educational migration, rather than being absent from it, as it currently is.
My main uncertainty is no longer whether this pathway exists — it clearly does, and at unusually low cash cost per candidate reached. The open question is how large its real-world impact turns out to be once candidates use it at real scale, particularly the placement rate this post can't yet answer. That seems like a tractable empirical question, not a reason to wait — I'd welcome pushback on any of the assumptions above, especially from anyone who's evaluated a similar early-stage, pre-scale intervention before.
