Onboarding new employees: from months to weeks on the shop floor

A new operator has been on the job for three weeks. The machine does something he does not recognise: a sound, a deviation, an alert that is not in the instructions. He stops his work, walks across the floor, and looks for the colleague who does know. That colleague stops his own work, looks for ten seconds, adjusts a setting, and walks back. Two minutes. That same week it happens seven more times.

Both men are doing exactly what you would expect of them. And yet this is the most expensive form of knowledge transfer a production environment knows. Not because something goes wrong, but because every question pulls two people out of production at once, and because afterwards the answer is again recorded nowhere. The next new employee starts from zero.

Onboarding in manufacturing is rarely slow because of a lack of motivation or training. It is slow because the knowledge someone needs to become independent is, for the most part, captured nowhere.

Why onboarding in manufacturing takes months, not weeks

There are two very different moments in an onboarding path, and companies often confuse them.

The first is when a new employee masters the basic operations. They can operate the machine, know the sequence, know where the buttons are. In most production roles that takes one to three months. It is visible, it is measurable, and it creates a sense of progress.

The second moment is when they are independent. When they know when to deviate from the standard, which machine reacts differently under which conditions, which customer has an exception that was never written down, and when a deviation is harmless versus when to stop the line. In knowledge-intensive manufacturing that takes six to twelve months, and sometimes longer. Broader figures on time to productivity point in the same direction: eight months to a year to full productivity is a common rule of thumb, though that is an industry norm rather than hard research.

The problem sits exactly in that gap between “can do it” and “can do it alone”. You learn the first phase from a manual. You learn the second from a colleague, or not at all.

That gap is not a local phenomenon either. The European Centre for the Development of Vocational Training, Cedefop, found that in 2024 twenty-two occupations count as shortage occupations across the EU, and that the most widespread shortages are precisely in metalworking and manufacturing. Whoever finds people also has to genuinely bring them up to level. The labour market no longer delivers ready-made craftspeople.

What the onboarding period really costs

The cost of onboarding is rarely booked, because it appears on no invoice. It is spread across three places.

• The new employee produces below level for months. They are on the payroll, but for the first months they run at a fraction of their eventual output. With an onboarding path of six to twelve months, that is not a warm-up week but a substantial part of a year.
• The experienced colleague produces less. Every question interrupts their own work. In a survey of Dutch technical professionals (TechBarometer, ROVC), 45 percent report that mentoring comes at the expense of their own work.
• Mistakes cost material and quality. Scrap and rework in manufacturing typically sit somewhere between half a percent and just over two percent of revenue, with a clear difference between companies that have it under control and those that do not. Inexperienced hands sit on the wrong side of that spread, and a mistake that reaches the customer costs more than the scrap itself.

There is a fourth cost that only becomes visible later. The same research shows that 37 percent see new employees being deployed independently too soon. That is what happens when the pressure rises: onboarding does not get shorter, it gets cut off. The bill arrives afterwards, in the form of mistakes, uncertainty, and people who leave again within a year.

Why training, manuals, and buddy systems fall short

Companies try to close this gap with the means they know. All three work, up to the point where it really counts.

Formal training covers only a small part of what someone learns. The well-known 70-20-10 model from the Center for Creative Leadership estimates that about seventy percent of learning at work comes from practice, twenty percent from colleagues, and ten percent from formal training. The exact ratio has since been criticised and should not be read as law, but the direction is undisputed: most craftsmanship arises on the floor, not in a classroom.

Manuals and work instructions describe the standard. That is exactly why they do not help at the moment something goes wrong, because that moment is by definition a deviation from the standard. The knowledge that makes a new operator independent is about exceptions, and exceptions are hard to write down in advance. Research into tacit knowledge elicitation for shop-floor workers shows that implicit knowledge is hard to put into words and hard to capture in procedures and documents (Kernan Freire et al., TU Delft, CHI 2023).

The buddy system does work, but it does not scale. One new employee alongside one experienced colleague is a good solution. Four new people alongside the same colleague is a production problem. And the quality of the onboarding depends on who happens to be on shift that day.

On top of that, the handover itself is bumpier than managers think. In the same ROVC research, 35 percent of organisations say they have insufficient insight into who the most important knowledge holders are, and 40 percent of younger technicians have insufficient insight into the crucial practical knowledge of older colleagues. The willingness to share is there. The structure is missing.

What does work: knowledge available at the moment of the question

If seventy percent of learning happens on the floor, then the question is not how to organise more training. The question is how to make the knowledge that is already on that floor available at the moment someone runs into it.

That fundamentally changes the role of the experienced operator. Today they answer the same question eight times to eight different people. In the other model they answer it once, at the moment they solve the problem themselves, and the answer is findable for everyone afterwards.

Four things make that possible.

• Capture it in the moment, during the work. Knowledge you try to reconstruct at the end of the shift has already lost half its detail. And knowledge that creates extra work does not get captured.
• Let people talk instead of type. A craftsperson explains their knowledge effortlessly out loud and laboriously in writing. Sticking to text picks the form that fits worst with the person who holds the knowledge.
• Make it searchable, not archivable. A new operator has no need for a folder of documents. They need an answer to one concrete question, within a minute, while standing at the machine.
• Build in a human control step. Before something counts as valid knowledge, someone with craft knowledge confirms or corrects it. Otherwise nobody trusts the system, and a knowledge source that is not trusted is not used.

The gain is not in the first month. Basic operations were learned quickly anyway. The gain is in months three to nine, the phase in which the new employee would otherwise remain dependent on whoever happens to be standing next to them.

Why this is a growth question, not an HR question

For companies that want to scale up, this is not a side issue but the brake itself. The order book can be full and the machines can be in place. If you only have two people who can onboard new colleagues, your growth speed equals their availability.

That brake will be tightened further in the coming years. In the Netherlands, 58 percent of employers still found the labour market tight in 2025, and in manufacturing 53 percent of vacancies were hard to fill (UWV Employer Survey 2025). Germany, where many of our suppliers and customers are based, is ageing fastest of the entire European Union: in 2024, 24.0 percent of workers were aged 55 to 64, against an EU average of 20.1 percent (Destatis, 2026). For German mechanical engineering alone, the Institut der deutschen Wirtschaft expects that by 2034 around 296,000 people will leave against roughly 118,000 entrants, a net gap of 178,000 skilled workers.

The sum is uncomfortable but simple. More experienced people are leaving than entering, the entrants need six to twelve months, and the people who have to mentor them are the same people you pull out of production. Whoever tries to solve that by recruiting harder is solving the wrong half of the problem.

Shortening onboarding time is therefore not a personnel theme. It is the cheapest capacity expansion there is.