Most management fashions never face an experiment. The four-day week has faced several, which makes it possible to do something unusual in this genre: separate what the trials actually found from what advocates and skeptics say they found. Both camps have been selective.
The centerpiece is the study published in Nature Human Behaviour in 2025 by Wen Fan, Juliet Schor and colleagues: a six-month trial of a four-day week at full pay covering 2,896 employees in 141 organizations across six countries (the US, UK, Canada, Australia, Ireland and New Zealand). The measured results were clean and consistent. Burnout fell, job satisfaction rose, and self-reported mental and physical health improved, a pattern absent in the twelve companies used as comparisons. The study found a dose-response relationship at the individual level: the more a given worker's hours actually fell, the larger their well-being gain, which is the signature one expects if the hours themselves are doing work. The researchers also traced the mechanisms: improved sense of work ability, fewer sleep problems and reduced fatigue carried most of the effect. Alongside it sits the British trial run by the 4 Day Week Foundation from November 2024 to April 2025: roughly 1,000 employees across 17 organizations, at the end of which all 17 chose to keep the shorter week, echoing earlier UK, US and Canadian pilots in which nearly every participating company retained the policy.
What the trials cannot show
Now the other ledger. None of this is a randomized experiment. The organizations in the Nature Human Behaviour study volunteered, and its authors are explicit that the design is pre/post with comparison firms, not an RCT; the well-being outcomes are self-reported by workers who knew what was being tested and, in many cases, wanted it to succeed. The business-outcome evidence is thinner still. In the UK trial, only 4 of the 17 organizations supplied revenue data at all (three saw increases; the one decline used a different comparison period), and the report's own authors acknowledge the findings cannot support causal inference. Michael Sanders of King's College London put the sharpest version of the critique on record: firms that volunteer for four-day-week trials are precisely the firms already positioned to succeed at them, and the foundation's director conceded that future trials need a randomized control element. The headline that "every company kept it" is real, but it is a statement about satisfied volunteers, not about the average firm.
One more finding deserves its asterisk restored. Workers in the large trial reported feeling just as productive in four days as in five. That is a fact about perceived productivity, honestly labeled as such in the research and promptly relabeled as measured output in the retelling. Objective productivity data at scale simply does not exist yet for this policy.
The overlooked variable: what happened before day one
The most practically important detail of the Nature Human Behaviour trial is the one both boosters and critics skip. Participating organizations did not simply delete Friday. They spent the run-up weeks redesigning work, cutting meeting load, tightening handoffs, automating low-value routines, and the authors state plainly that the measured effects belong to the package of redesign plus reduced hours, not to the reduction alone. This reframes the whole debate. The four-day week may function less as a schedule than as a forcing device: a hard constraint that finally makes an organization do the process hygiene it could theoretically have done at any time, but never would without a deadline. Skeptics read this as deflating (the day off didn't do it). It is equally readable as the strongest practical argument: if it takes an attractive, morale-raising constraint to trigger the redesign, the constraint is earning its keep twice, and the dose-response finding suggests the recovered time then adds its own effect on top.
The honest scoreboard
Where does the evidence actually stand? Established with reasonable confidence: large, consistent well-being improvements across thousands of workers and multiple countries, with plausible measured mechanisms; near-universal voluntary retention of the policy by participating firms. Not established: causal effects on revenue, output or profitability; generalization beyond self-selected, mostly small, mostly anglophone organizations; durability of effects once novelty fades. Anyone quoting the trials as proof the four-day week "works" for business is ahead of the data. Anyone dismissing them as hype must explain away a well-measured, mechanism-backed, dose-responsive improvement in human functioning, replicated in trial after trial, at apparently tolerable cost to the volunteers.
Firms keep a policy they could freely drop; workers sleep more and burn out less; the promised collapse in output remains unobserved among participants. The four-day week's evidence is incomplete. Its null hypothesis, that five days is the right number, was never tested at all.
Sources
- Nature Human Behaviour / Fan, Schor, Kelly & Gu (2025). Work time reduction via a four-day workweek. Six-month trial, 2,896 employees, 141 organizations, 6 countries: improvements in burnout, job satisfaction, mental and physical health versus 12 comparison companies, with a dose-response pattern at the individual level; quasi-experimental, not randomized. https://www.nature.com/articles/s41562-025-02259-6
- Nature News (2025). Coverage of the largest four-day-week trial. Workers reported feeling just as productive on four days; the study is the largest intervention trial of the policy to date. https://www.nature.com/articles/d41586-025-02295-2
- CNN Business (2025). UK four-day week trial (4 Day Week Foundation, Nov 2024-Apr 2025). ~1,000 employees across 17 organizations; all 17 kept the shorter week, but only 4 provided revenue data, and researchers acknowledge the trials cannot support causal inference; self-selection critique by Michael Sanders (KCL). https://edition.cnn.com/2025/07/03/business/four-day-week-trial-uk-intl