Two numbers get quoted about the UK four-day week pilot. One is roughly 35% revenue growth. The other is 1.4%.
The second is the one that describes the trial. The first compares performance to the equivalent period in previous years, which folds in everything else that happened to those businesses over that time. It is not a measurement of the four-day week, and it is the number that ended up in most of the coverage.
Starting there because the evidence on this is genuinely interesting, and it is poorly served by both the advocacy and the dismissal.
What the pilot actually found
The UK pilot ran from June to December 2022 across 61 companies and about 2,900 workers, coordinated with researchers and documented by the Autonomy Institute. It remains the largest trial of its kind.
The strongest result is the employers' own revealed preference. Of the 61 organisations, 56 continued with the four-day week after the trial ended, and 18 made it permanent. Companies that have just run a six-month experiment are well placed to judge whether it damaged them, and the overwhelming majority chose to keep going. That is a more persuasive signal than any single metric in the report.
The wellbeing results are the best-documented of the quantitative findings: 71% of employees reported reduced burnout, 39% reported less stress. Self-assessed staff turnover fell by 57%, from company reports.
Revenue rose about 1.4% on average over the trial. Modest, but the relevant comparison is not "did revenue grow impressively" — it is "did a 20% reduction in scheduled hours damage the business", and the answer appears to be no.
A later study published in Nature Human Behaviour, covering 141 organisations over six months, found no decline in self-assessed performance. Same direction, same caveat about self-assessment.
Where the evidence is weak, honestly
Three limitations matter, and advocates tend to skip them.
Selection. Organisations volunteered. They were disproportionately small, services-based, and led by people who already suspected it would work. You cannot generalise from that to a hospital, a factory floor or a 40,000-person bank.
Self-report. Burnout and stress are subjective states, so self-report is the right instrument for them. Productivity and performance are not, and most of the productivity evidence here is self-assessed. "People did not feel less productive" is a weaker claim than "output was maintained", and the two get conflated.
Duration and the Hawthorne problem. Six months of a well-publicised trial that everyone wants to succeed is not the same as year three, when it is just the schedule and nobody is watching. The 2026 follow-up position is that no comprehensive public register tracks which pilot employers still run it, with counts of 54 or 56 of 61 depending on the source, which tells you how thin the long-run data is.
None of this makes the four-day week a bad idea. It makes it an idea with good evidence for retention and wellbeing, suggestive evidence for output neutrality, and almost no evidence about durability at scale.
The mechanism, which is the part worth copying
What the successful trials mostly did was not work the same way in less time. They cut meetings.
That is the consistent operational finding across reports: shorter and fewer meetings, less synchronous coordination, more asynchronous updates, and a sharp reduction in work that existed to demonstrate work. The schedule change forced a prioritisation exercise that nobody had been willing to run on its own.
Which raises an uncomfortable question for anyone considering this. If the mechanism is removing low-value coordination, you can do that on five days. The four-day week works partly because it is a hard constraint that makes the prioritisation unavoidable — but the gain is in the prioritisation, not the day off.
Our analysis of the cost of meetings is the version of this exercise you can run without changing anyone's contract, and productivity theatre covers the category of work that tends to get cut first.
How to measure your own trial
If you are going to try this, the measurement design matters more than the policy design.
Baseline for at least six to eight weeks before anything changes. Without that you will be comparing against recollection, and recollection will tell you whatever the trial's champion wants to hear.
Choose output measures specific to your work rather than generic productivity. Tickets closed, features shipped, cases resolved, revenue per head, delivery dates met. Generic measures invite the self-assessment problem.
Then instrument the three displacement patterns directly. Total hours actually worked, not scheduled — if hours did not fall, nothing has been tested. After-hours and weekend activity, which catches spillover. Workload distribution across the team, which catches transfer. And meeting load, which tells you whether the prioritisation actually happened or whether everyone just has the same calendar in four days.
Working-pattern data answers all four of those, and it is the measurement most trials skip in favour of a survey. Our note on spotting burnout early covers the after-hours signal specifically, and capacity planning covers the distribution question.
Run it long enough to get past the novelty. Six months is the conventional minimum and it is a minimum.
Who this probably does not suit
Shift and coverage-based work, where reducing individual hours means hiring rather than prioritising, and the economics are completely different. Client-facing roles with external availability expectations, unless the clients are in on it. Teams already running at genuine capacity with no low-value coordination left to cut, which is rarer than people claim but does exist.
And organisations whose real problem is something else. A four-day week will not fix unclear priorities, a broken toolchain or a manager nobody wants to work for. It will make all three more visible, which is useful but not what was advertised.
Where ProdView fits
ProdView measures the things a four-day week trial needs to track and usually does not: actual hours worked versus scheduled, after-hours and weekend creep, meeting load, focus time, and how workload distributes across a team. Aggregate-first, activity metadata only, never screen content, with employees seeing the same dashboard managers do — which matters here, because a trial measured in secret produces compliant answers rather than true ones.
When not to pick us: your output measure is the thing that needs tracking and it lives in your delivery, CRM or support system. Measure it there. We tell you about hours, load and fragmentation, not about whether the work was good.
Try before you commit
Three seats are free forever, which is enough to establish a baseline before a trial starts — and the baseline is the part that cannot be reconstructed later. The meeting cost calculator prices the coordination you might cut.
Pilot figures as reported by the Autonomy Institute and subsequent peer-reviewed work; several widely circulated figures do not survive checking, so verify against the primary reports before using any of them in a board paper.