TL;DR. The four-day week has been tested in hundreds of workplaces since 2015, and the evidence splits cleanly in two. On health, it is consistent and it survives the strongest designs: a randomized Swedish study found better sleep and lower stress after 18 months, and a 2025 Nature Human Behaviour analysis of 2,896 employees in 141 organizations found improvements in burnout, job satisfaction and mental and physical health that did not appear in 12 control companies. On output, the evidence is weak. The famous revenue figures come from 38% to 61% of the companies in each pilot, one US and Irish resignation figure falls when weighted by company size and rises by 20% when it is not, and none of the private pilots that published average hours delivered the promised 20% cut. The newest data point, a German follow-up from February 2026, shows 70% of organizations still reducing hours two years on. The honest summary: a well-designed shorter week reliably makes people healthier and does not measurably damage the business in the organizations that chose to try it. That is a strong claim. It is not the claim most headlines make.
This piece belongs to our life design cluster, and it reads best alongside our global burnout data map and the evidence review on meeting-free days.
Why read the denominators
A CEO reading a pilot report asks one question before any other: how many of the participants produced this number? A student asks the second: what would the result have been without the intervention? Most coverage of the four-day week skips both. The reports themselves usually do not. The caveats are printed in the primary documents, often a few pages after the headline.
So we did the unglamorous work. We extracted the text of seven primary reports and one peer-reviewed trial, took each headline, and rebuilt it from the counts the authors published.
The pilots at a glance
Table 1. The major four-day week studies, by design strength (verified data from the primary reports)
| Study | Who | Size | Length | Comparison group | Hours actually cut |
|---|---|---|---|---|---|
| Sweden, Schiller and colleagues | Public sector workplaces | 33 workplaces, 580 people | 18 months | Yes, workplaces randomized | 25% (by design) |
| Six countries, Fan and colleagues | Organizations in six countries | 141 organizations, 2,896 people | 6 months | 12 control companies | Measured per person; varied |
| Portugal, Gomes and Fontinha | Private companies | 41 companies including 21 coordinated pilots; over 1,000 workers across all 41 | 6 months | 14 control companies, 160 workers, not randomized | 12%, from over 41 hours to 36.5 |
| Germany, Backmann and colleagues | Private organizations | 45 enrolled, 41 analysed | 6 months | Employees in organizations that did not switch everyone | 3.95 hours a week |
| Iceland, Reykjavik City and national government | Public sector | Over 2,500 staff, over 100 workplaces | 2015 to 2019 | Some comparison workplaces, not randomized | 40 to 35 or 36 hours |
| UK, Autonomy with Cambridge and Boston College | Companies and non-profits | 61 companies, about 2,900 workers | 6 months | None | 38 to 34 hours |
| US and Ireland, Schor and colleagues | Companies and non-profits | 33 companies, 903 employees | 6 months | None | 40.83 to 34.83 hours |
Two things stand out before any outcome is considered.
First, only one study randomized. The Swedish trial, published in the Scandinavian Journal of Work, Environment and Health in 2017, is the only design in the table that can rule out the possibility that the workplaces volunteering for shorter hours were already on a different path. The European Commission’s Joint Research Centre made this point directly in a 2023 methods review: private-sector pilots rely on before-and-after comparisons, and it concluded that their quantitative results should be treated as indicative rather than as robust evidence.
Second, the “four-day week” was rarely 32 hours. Computed from the published averages, realized cuts were 10.5% in the UK, 14.7% in the US and Ireland, and 12% in Portugal. Participants were keeping part of the fifth day. The German research team dropped the label altogether in its 2026 follow-up, writing about “work time reduction” because so many organizations had moved to flexible weekly or monthly hour targets instead of a fixed day off.
Health: the part that holds
If you only remember one section, make it this one.
The Swedish randomized trial reduced weekly hours by 25% with salary preserved for 18 months. Compared with the control workplaces, the intervention group slept 23 minutes longer on workdays, reported better sleep quality and showed lower sleepiness, perceived stress and worry at bedtime. The authors are careful to add that the effect sizes were small. The effects were still present at 18 months.
The largest dataset points the same way. Fan, Schor, Kelly and Gu analysed pre- and post-trial data from 2,896 employees in 141 organizations across Australia, Canada, Ireland, New Zealand, the UK and the USA. Burnout, job satisfaction, mental health and physical health improved, and the pattern was not observed in 12 control companies. The paper also identifies three mediators: better self-rated work ability, fewer sleep problems and less fatigue.
One finding from that paper matters more for individuals than any headline. Larger reductions in an individual’s own hours were associated with larger well-being gains, but larger reductions at company level were not. The policy on paper did less than the hours people actually stopped working.
The German pilot, which gave participants smartwatches, measured 38 more minutes of sleep per week than in its control group. Portugal found the share of pilot workers rating their mental health very good or excellent doubled from 15% to 30%, and the share struggling to balance work and family fell from 46% to 17%.
Different countries, different designs, different research teams, one direction. That is what a robust finding looks like.
Output: the part that does not
Here the headlines and the denominators part company.
Table 2. Headline figures rebuilt from their denominators (CEOtudent editorial framework, derived from counts published in the primary reports)
| Headline as usually reported | What the report shows | Share of pilot behind the number |
|---|---|---|
| UK: revenue rose 1.4% | Size-weighted, start to end of trial, 23 companies | 23 of 61, which is 37.7% |
| UK: revenue up 35% on the prior year | 34.5% size-weighted, comparison period to trial period, 24 companies | 24 of 61, which is 39.3% |
| UK: resignations fell 57% | 2 to 0.8 per 100 employees; authors state they cannot say the trend is statistically significant | Not significant by the authors’ own account |
| US and Ireland: revenue rose 8% | 8.14% weighted; the same 16 companies show 30.15% unweighted | 16 of 33, which is 48.5% |
| US and Ireland: revenue up 38% on 2021 | 37.55% weighted; the same 20 companies show 31.22% unweighted | 20 of 33, which is 60.6% |
| US and Ireland: resignations fell | 1.76 to 1.70 weighted, a 3.4% fall; 1.99 to 2.39 unweighted, a 20.1% rise | 22 of 33; direction depends on weighting |
| UK: 92% of companies continuing | 56 of 61 immediately after the trial; 18 said permanent | Permanent: 29.5% |
| US and Ireland: none planning to stop | 18 definitely continuing among 27 respondents | Definitely continuing: 18 of 33, which is 54.5% |
| Germany: 73% continuing | 30 of 41 in October 2024; in February 2026, 70% of 40 respondents still reduce hours | At least 28 of 45 enrolled, which is 62.2% |
| Iceland: 86% of workers now on shorter hours | Or gained the right to shorten; baseline contract cuts the report lists include 35 minutes a week for private-sector shop workers and 65 minutes a week in the public sector | 1.5% and 2.7% of a 40-hour week |
None of this means the pilots failed. The UK and US reports disclose these details themselves, and weighting by company size is a defensible choice. What the table shows is that the revenue figures describe a minority of each cohort, that small samples make the direction of some effects unstable, and that “continuing” and “permanent” are different claims.
The European Commission’s methods review singled out the resignation example: the weighted rate suggested a small improvement that was highly likely to be statistically insignificant, because the unweighted data pointed the other way on a small sample.
The German pilot was more conservative from the start. Its executive summary reports that revenue and profit showed no significant difference compared with the previous year, and it frames the productivity story accordingly: stable output with fewer hours implies some productivity gain. That is a defensible inference. It is also the most anyone can claim.
The public sector adds one piece of administrative evidence that does not depend on surveys. South Cambridgeshire District Council, which in July 2025 voted to adopt the model permanently, reports that an independent analysis by the universities of Salford, Bradford and Cambridge found 21 of 24 monitored services improved or stayed the same. The council defines its model as 100% of the work in around 85% of contracted hours for 100% of pay.
Focus: what the pilots actually changed
The honest answer on focus is that the pilots measured it indirectly, through what people reported changing.
In Germany, employees compensated for fewer hours by reducing distractions (65%), optimizing processes (63%) and changing meeting structures (52%). Fewer used dedicated focus time (32%) or new digital tools (25%).
In Portugal, 75% of companies made at least one organizational change, and the most common was shorter meetings. Then comes the most useful single number in the entire literature for anyone planning a shorter week: among pilot companies that made two or more process changes, 8% went back to five days. Among those that made one change or none, 38% did.
The UK report adds a caution. By the end of the trial, 62% of employees said their pace of work had increased, even though 78% reported no increase in workload. Average work intensity rose only slightly, and individual answers were split: 36% felt more intensity, 31% less.
Read together, these point to a mechanism rather than a miracle. The shorter week works where it forces a redesign of how time is spent, especially meetings, and it strains where it simply compresses the same habits into fewer hours. If you want the underlying data on meeting load, our analysis of the true cost of meetings sets it out.
Durability: the newest data
Most pilots report at six months, which is roughly when novelty fades. The German team went back.
In October 2024, 73% of the 41 organizations said they would continue, 20% had decided against continuing and 7% were undecided. By February 2026, 40 of the 41 answered a follow-up survey. Of those, 70% still operated some form of reduced hours, 30% did not, and 22% had adapted their original model, most often because the fixed format felt too rigid for daily operations. Organizations that stopped cited workload peaks and volatile business conditions. One suspended the model during an order peak and reinstated it later.
The follow-up also surveyed 167 employees. Those still on reduced hours reported significantly lower burnout than the control group, and so did those whose organizations had reverted, which the authors describe as a tentative sign of a residual effect. Among the organizations that had gone back, one third said they had concrete plans to reintroduce a working time adjustment, and those plans favoured compressing the same hours into fewer days rather than reducing total hours.
The researchers are explicit that these are self-reports from a subset and do not permit causal conclusions. Taken as they are offered, they suggest that the arrangement is durable for most adopters and fragile for organizations with volatile demand.
What the evidence supports, graded
Table 3. Evidence strength by outcome (CEOtudent editorial framework, based on the studies above)
| Outcome | Best available evidence | Grade | What you can responsibly say |
|---|---|---|---|
| Sleep and stress | Randomized Swedish trial, plus control-group comparisons in six countries, Portugal and Germany | Strong on direction, small in size | Reduced hours improve sleep and lower stress |
| Burnout and job satisfaction | 141-organization study with control companies; Portugal control group | Moderate to strong | Improvements are consistent and not explained by general trends in control firms |
| Retention and absenteeism | Small-sample company data; authors report non-significance; German absenteeism not significant | Weak | Direction is favourable; the size is unknown |
| Revenue and output | Data from 38% to 61% of firms; weighting changes the magnitude; German financials not significantly different | Weak for gains | No measurable harm in organizations that chose to try it |
| Focus and work design | Self-reported changes; Portugal reversal rates by number of process changes | Descriptive | Redesign, not hours alone, predicts whether it sticks |
| Durability | German two-year follow-up | Emerging | Most adopters keep it; volatile-demand organizations struggle |
The phrase that carries the most weight in the table is “organizations that chose to try it.” Every private pilot was self-selected. The Portuguese researchers wrote that their method, based on companies selecting themselves, prevents the results from being generalized, and that the results inform the debate but do not justify implementation by legislation. That is the student’s discipline applied to one’s own data, and it is rarer than it should be.
How to run your own test
If you lead a team, run a business or control your own schedule, the evidence gives a usable protocol. It is our construction, derived from the findings above.
Table 4. The CEOtudent Shorter-Week Test Protocol (editorial framework)
| Step | What to do | Why, from the evidence |
|---|---|---|
| 1. Name the outcome | Choose one primary outcome in advance: health, retention or output | The JRC review flags the lack of pre-registered outcomes as a core weakness |
| 2. Redesign before reducing | Make at least two process changes first, starting with meetings | Portugal: 8% reversal with two or more changes, 38% with one or none |
| 3. Track realized hours | Log hours actually worked, not the schedule | Pilots cut 10.5% to 14.7%, not 20%; individual reductions predict gains |
| 4. Keep a comparison | Use a team that has not switched, or the same months last year | Before-and-after comparisons cannot separate the policy from seasonality |
| 5. Report both views | Look at weighted and unweighted results, and at who did not respond | One resignation figure changes sign with weighting; survey panels lost 22% to 48% of respondents |
| 6. Decide the reversal rule in advance | Write down what result by what date would end the test | German reverters cited demand peaks; decide what you would do before one arrives |
| 7. Check again at a year | Six months captures novelty; two years captures fit | 30% of German organizations had stopped by February 2026 |
For individuals, step 3 is the one to keep. The 2025 study found that the hours you personally stop working matter more than the policy your organization announces. A freelancer who protects a genuine day off is closer to the evidence than an employee whose company adopted a four-day week and still answers messages on Friday. Our recovery protocol covers how to make rest time actually restorative.
When the evidence does not apply to you
When demand is volatile and delivery is contractual. The German organizations that stopped did so because they could not guarantee reliable delivery during peaks. If your revenue depends on responding within hours, the evidence from agencies and software firms may not transfer.
When the week is compressed, not reduced. Doing 40 hours in four days is a different intervention. The health findings above come from reducing total hours. They say nothing about ten-hour days.
When you cannot change how the work is done. Every piece of evidence on durability points to redesign. A shorter week layered on an unchanged meeting calendar is the arrangement most likely to revert.
This is where the CEO and the student meet. The CEO decides to run the experiment and owns the reversal rule. The student reads the denominators and refuses to call a six-month self-selected pilot a proof. For the longer arc of that trade-off, see our piece on the 40-year career.
Frequently asked questions
Does a four-day week increase productivity?
The best available answer is that it maintains output in the organizations that have tried it, which implies higher output per hour. The evidence for revenue gains is weak because the figures come from a minority of firms in each pilot and are sensitive to weighting. The German study, the most cautious, found no significant change in revenue or profit.
Is the health benefit real or just enthusiasm?
It is the best-supported finding in the literature. It appears in a randomized trial with 18 months of follow-up, and in large studies with control groups. Effect sizes in the randomized trial were small, so expect a meaningful improvement rather than a transformation.
How many companies keep the four-day week after a trial?
It depends on what you count. In the UK, 56 of 61 continued immediately and 18 called it permanent. In Germany, 70% of responding organizations still reduced hours two years after the start, which is at least 28 of the 45 that enrolled.
Did Iceland move to a four-day week?
No. The trials cut the week from 40 hours to 35 or 36. The report states that roughly 86% of the workforce has moved to shorter hours or gained the right to shorten them, but the baseline contract reductions it lists for large groups are 35 minutes a week in the private sector and 65 minutes in the public sector.
What is the single best predictor that a shorter week will stick?
Process change. In Portugal, companies that made two or more organizational changes reverted 8% of the time, against 38% for those that made one change or none.
Does this apply to freelancers and solo workers?
Partly. You cannot use a control group, but you can use the same months from last year as a comparison, and the finding that individual hours matter more than company policy applies directly to you.
Sources
Fan, W., Schor, J. B., Kelly, O., and Gu, G. Work time reduction via a 4-day workweek finds improvements in workers’ well-being. Nature Human Behaviour, 2025, volume 9, pages 2153 to 2168.
Schiller, H., Lekander, M., Rajaleid, K., Hellgren, C., Åkerstedt, T., Barck-Holst, P., and Kecklund, G. The impact of reduced worktime on sleep and perceived stress: a group randomized intervention study using diary data. Scandinavian Journal of Work, Environment and Health, 2017, volume 43, issue 2, pages 109 to 116.
Lewis, K., Stronge, W., Kellam, J., Kikuchi, L., Schor, J., Fan, W., Kelly, O., and colleagues. The results are in: the UK’s four-day week pilot. Autonomy, February 2023.
Schor, J. B., Fan, W., Kelly, O., Gu, G., Bezdenezhnykh, T., and Bridson-Hubbard, N. The Four Day Week: assessing global trials of reduced work time with no reduction in pay. Four Day Week Global, 2022.
Gomes, P., and Fontinha, R. Four-Day Week: Results from Portuguese Trial, Final Report. July 2024.
Backmann, J., Hoch, F., Hüby, J., Platz, M., and Sinnemann, M. F. The 4-Day-Week in Germany: First Results of Germany’s Trial on Work Time Reduction, 2024; and The 4-Day-Week in Germany: A Follow-Up Two Years After the Start of the Trial, 2026. University of Münster with Intraprenör, Berlin.
Haraldsson, G. D., and Kellam, J. Going Public: Iceland’s Journey to a Shorter Working Week. Alda and Autonomy, June 2021.
Cuello, H. Assessing the Validity of Four-day Week Pilots. JRC Working Papers Series on Social Classes in the Digital Age 2023/08, European Commission Joint Research Centre, Seville, 2023.
South Cambridgeshire District Council. Four-day working week: analysis of performance data, including the independent report by the universities of Salford, Bradford and Cambridge published July 2025.
All shares, percentage changes and hour reductions in Tables 1 and 2 were computed by us from the counts and averages published in these reports and checked by an independent script.
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
This post is also available in:
