The Four-Day Week: What the Trials Measured, and What They Could Not

Short answer: the evidence is better than it was two years ago and weaker than the coverage suggests. The largest study, published in Nature Human Behaviour in 2025, followed 141 organisations and 2,896 workers across six countries and found real improvements in burnout, job satisfaction and health. It also found that the average week shrank by about five hours, not eight, that no trial has been randomised, and that the outcomes were self-reported.

The German pilot that did use objective measurement found effects in the same direction and considerably smaller in size. Both point at something the headlines miss: it may not be the day off doing the work.

The largest trial

Improvements appeared in burnout, job satisfaction, mental health and physical health relative to controls, and the researchers identified the mechanism: the gains were mediated by reduced sleep problems and fatigue. Less tired people reported feeling better, which is neither surprising nor trivial.

The results indicate that income-preserving 4-day workweeks are an effective organizational intervention for enhancing workers' well-being.

Fan, Schor, Kelly and Gu, Nature Human Behaviour, 2025

The five-hour figure reframes the whole debate. "Four-day week" sounds like 40 hours becoming 32. In practice the average reduction was around five hours, with wide variation: some cut eight or more, others far less. Many organisations ran compressed or staggered versions rather than closing on Fridays. If you are arguing about a 32-hour week, most of this evidence is not about that.

The trial that measured rather than asked

The German pilot, run by Julia Backmann's team at the University of Münster across 45 companies for six months, is the methodologically interesting one, because it did not rely on surveys alone. Participants wore activity trackers and gave hair cortisol samples, which integrate stress hormone exposure over weeks rather than capturing a moment.

Read the objective column against the self-reported one. Employees and managers perceived productivity gains; the financial records showed changes that were not statistically distinguishable from the previous year. Sleep improved by 38 minutes across a whole week, which is roughly five minutes a night — a real effect, and not the transformation implied by "better sleep" in a headline.

This is the most useful thing in the entire literature: when you measure instead of asking, the direction holds and the magnitude shrinks.

Does it last? The question just got answered

Every earlier pilot ended with the same unanswered question: six months is a honeymoon, what happens after?

The Münster team published a two-year follow-up in May 2026. Of the participating organisations, 70% had kept some form of reduced working time, and a further 22% had adapted their model rather than abandoning it. That is 92% still running some version two years on, with financial performance stable across the cohort.

This is the strongest durability evidence available on the question, and it is genuinely new. It does not prove the arrangement causes good outcomes — companies that liked it kept it, which is what you would expect either way — but "it quietly reverted after a year" is now a claim contradicted by data rather than merely unsupported.

What the trials cannot tell you

Four limitations, all acknowledged by the researchers.

No randomisation. Companies volunteered. Organisations willing to trial a four-day week are, almost by definition, already more attentive to staff wellbeing than average. Nothing here separates the intervention from the kind of employer that would attempt it.

Self-selected workers within self-selected firms. People who dislike the arrangement can leave the study or the company; those who thrive stay and answer the follow-up survey.

Narrow sector coverage. The Nature sample skews heavily to professional services and office work in English-speaking countries. Manufacturing, retail, healthcare and shift work are barely represented, and those are the settings where coverage costs make the arrangement hardest.

Outcomes measured by asking. The Nature study's health findings are self-reported. The German pilot is the corrective, and it is one pilot of 45 companies.

The finding that should change how you read all of it

Here is the part that gets buried. In the German pilot, the participating companies did not simply remove a day. Before and during the trial:

• over 60% reduced distractions and streamlined processes
• more than 50% cut the number or length of meetings
• 25% introduced new digital tools

The Nature study describes the same pattern: organisations underwent pre-trial work reorganisation to improve efficiency and collaboration.

So the intervention being tested is not "one fewer day". It is "cut the meetings, remove the interruptions, and also take a day off" — and no trial has separated those. It is entirely possible that a substantial share of the benefit comes from the reorganisation, which any team could do on Monday without changing anyone's contract.

That possibility is consistent with what the interruption research already shows. Mark, Gudith and Klocke found that interrupted work is completed faster but at a significant cost in stress, frustration and effort. Remove half the meetings and you have removed a large part of that cost, with or without a four-day week. We covered the mechanism in attention residue and the practical version in how to focus better.

The sectors the evidence does not cover

Worth stating plainly, because the gap is not random. It falls on exactly the jobs where the arrangement is hardest.

The Nature sample was heavily weighted toward professional services and office work in six English-speaking countries. That leaves out most of the workforce, and the omission is structural rather than accidental: the four-day week as trialled works by removing inefficiency from knowledge work. Cut the meetings, protect the focus blocks, and the same output emerges from fewer hours.

That mechanism does not exist in coverage-based work. A ward needs a nurse present. A shop needs someone behind the till. A production line runs or it does not. There is no meeting to cut, because the hours are the output. Reducing them means either hiring more people or covering less time, both of which are cost questions rather than efficiency questions.

So the honest scope of the evidence is: income-preserving four-day weeks improve wellbeing in organisations where output is not a direct function of hours present. Extending that to healthcare, retail, hospitality, logistics or manufacturing is an extrapolation the trials do not support in either direction — it has largely not been tested.

If you cannot get a four-day week

Most readers of this cannot implement one, so here is what the same evidence suggests you can copy.

The reorganisation is the transferable part. Over 60% of German pilot companies reduced distractions and streamlined processes; more than half cut meetings. None of that requires a contract change or anyone's permission beyond your own team.

Audit recurring meetings once. The pilot companies did this before the trial started. Cancel or halve anything that exists because it always has, and reclaim the time as protected work rather than as more meetings.

Protect one block, not a whole day. The mechanism identified in the Nature study ran through sleep and fatigue, not through the specific day off. Reducing chronic overload is the active ingredient.

Ask for compressed hours rather than reduced hours. A 40-hour week in four days is a different and much easier ask than a 32-hour week, it requires no pay discussion, and a meaningful share of the pilot organisations ran exactly that variant.

Take the actual evidence to the conversation. If you are making the case to an employer, the two useful facts are that financial performance was statistically unchanged in the German pilot, and that 92% of those organisations were still running some version two years later. Those answer the two objections an employer will actually raise.

What a reasonable conclusion looks like

The direction is consistent. Across dozens of trials, thousands of workers, six countries and both self-reported and objective measures, wellbeing outcomes move the same way. That consistency is worth something even without randomisation.

The magnitudes are modest. Five minutes more sleep a night. Revenue changes that did not reach significance. Sick-day reductions that did not either.

The durability is now evidenced. 92% of German participants still running some version after two years is the newest and arguably most decision-relevant fact available.

The mechanism is unresolved, and that is the useful gap. If you cannot get a four-day week, the trials suggest the components you can copy — fewer meetings, fewer interruptions, protected focus time — may be carrying more of the effect than the day itself.