The Benelux business case: what temp turnover and incidents cost

Most training business cases are built on a return percentage: completion up this much, retention up that much, payback in so many months. Those figures rarely say what was measured, against what, in which company. A finance director discounts them, and rightly.

There is a better case available in Belgium and the Netherlands, built entirely from numbers that anyone can check: the staffing federation's own figures, the federal accident agency's statistics, the Dutch labour inspectorate's reports and the published sanction rules. It does not promise a return. It shows where the cost sits and which part of it instruction controls.

Figures in this post are as of September 2026: Federgon 2025, Fedris 2024, the Arbeidsinspectie monitor 2024 and the Fedris contribution amounts for 2026.

Layer one: every new face is a first shift

Belgian temp work in 2025 came to 218 million hours, spread over 740,216 different people, 302,832 of them job students. That is about 295 hours per person over the year. It is an average across students, day contracts and long placements, not a turnover rate for your site. But it gives the order of magnitude: 100,000 hours of temp work means roughly 340 different people.

Each of them needs instruction before they start. In Belgium the user company gives the workstation instruction; in the Netherlands the agency must inform the worker about the risks before the work starts and the hirer carries the employer's Arbowet duties. Who owes what, and how to deliver it before the first shift, is set out in the temp-worker onboarding guide.

That instruction is not wasted when the person moves on. According to a HIVA-KU Leuven study for Federgon, 4 in 10 temps are later hired by the company where they worked. The first instruction is often the start of a permanent employee's induction.

The cost of this layer is simple arithmetic on your own numbers: the number of first shifts per year, times the minutes of instruction each one takes, times the hourly cost of the person giving it and of the worker receiving it. Most operators have never multiplied it out, because the time is spread across shift leads at every site.

Layer two: temps are hurt more often

Belgian employers in the private sector reported 97,591 workplace accidents in 2024, according to Fedris. The average frequency rate, accidents with at least one day of incapacity per million hours worked, was 12.66. In the staffing sector (NACE 78) it was 21.83. That is the lowest figure in the series (45.97 in 2012) and still well above the average. Per 100,000 hours: about 2.2 lost-time accidents in staffing against 1.3 across the private sector. Age adds to it: workers aged 20 to 29 had 48 accidents per 1,000 workers, against an average of 38.

The Dutch picture points the same way. In the accidents the Nederlandse Arbeidsinspectie investigated in 2024, the victim was a temp worker in 18 percent of cases, while temps are an estimated 4 percent of the working population. In industry it was 24 percent of victims against about 8 percent of the workforce. For temps the most common accident type was contact with work equipment or an object: the kind of risk a workstation instruction is meant to cover.

In Belgium, part of this now lands on the user company directly. Since 16 March 2024, accidents of temps count in the user company's risk index, the index Fedris uses to select companies with an aggravated accident risk. Before, only permanent staff counted. A company whose index is at least twice its sector's and five times the private-sector average can be selected, and then pays a flat contribution: for 2026, EUR 4,370.30 below 50 full-time equivalents, EUR 2,913.54 for each further 50, up to EUR 21,851.52. It also has to draw up and run a three-year action plan with its insurer's prevention service, followed up every quarter.

The contribution is the visible part. Absence, replacement, the investigation and the lost production are your own numbers, and usually larger.

Layer three: what an inspection adds

When an accident is investigated, the sanction rules multiply.

The Netherlands. Arbo fines are set in seven base categories from EUR 340 to EUR 13,500 for employers with 500 or more staff, scaled down to between 10 and 80 percent for smaller employers. When the violation led to an accident with permanent injury or hospitalisation, the fine is multiplied by four. Not reporting such an accident straight away, so that the inspector can no longer investigate it, carries a separate base fine of EUR 50,000.

Belgium. Social criminal and administrative fines are calculated by multiplying the base amount, and that multiplier rose from eight to ten on 1 February 2026. A level 3 criminal fine now runs from EUR 2,000 to EUR 20,000; level 4 from EUR 6,000 to EUR 70,000. The Social Criminal Code often multiplies the fine by the number of workers concerned, up to 100.

For staffing firms, the Dutch authorisation regime from 2027 adds another layer; operational drift covers it.

Building the case with your own numbers

Put the public anchor next to your figure for each line. Where you do not have the figure, that is the first finding.

Two rows carry most of the argument. The instruction load is certain: it happens every time someone new starts, whether you plan for it or not. The accident exposure is probable: the rates say temps are hurt more often, and the risk-index change means their accidents are now counted against the site where they work.

What this case does not claim

It does not claim that better instruction removes accidents, or by how much. No public source gives that number for your sites, and inventing one would put you back where the return-percentage cases started. What the data supports is narrower and harder to dismiss: temps turn over fast, they are hurt more often, and the costs of an accident now reach the user company through the insurer and the inspector. Instruction before the first shift, recorded per person and per site, is the part of that chain an operator controls.

Whether the instruction then worked is a separate question with its own evidence: what people did on the floor afterwards, not whether they finished a module. After an incident covers the other end of the chain.