Microsoft Japan gave roughly 2,300 employees five consecutive Fridays off in August 2019 without cutting their pay — and sales per employee came in 39.9% higher than the same month a year earlier. The experiment also forced a rethink of wasted work: 30-minute meetings became 46% more common, while employees increasingly replaced meetings with online collaboration.

In August 2019, Microsoft Japan closed its offices every Friday and gave roughly 2,300 employees five days of special paid leave. When the month ended, sales per employee were 39.9 per cent higher than in August 2018, despite a working calendar that was 25.4 per cent shorter.

The number quickly became shorthand for a four-day week producing 40 per cent more. Microsoft itself made a more cautious statement. Eight days after announcing the results, it amended its page to say the 39.9 per cent figure was accurate but had not been achieved by the challenge alone. Various factors had contributed.

That correction is not a footnote to hide. It is the boundary between an arresting comparison and a causal claim. The experiment did not prove that removing Fridays produced the sales increase. What it documented more directly was how a shorter calendar forced a company to reconsider meetings, communication and the routine costs of keeping an office open.

Five Fridays became paid days away from work

Microsoft Japan called the project Work-Life Choice Challenge Summer 2019. Under the programme announced in July, all five Fridays in August became special paid leave. Employees kept their salaries and did not have to spend annual holiday entitlement.

The company closed every office on those days. There was an operational exception: employees who could not take a Friday because of customer services or business requirements could arrange the special leave on another date by the end of September. This was therefore a broad company experiment, not an absolute shutdown of every function.

The project arrived during Japan’s wider effort to confront long hours and their human cost. National work-style reforms had begun taking effect in April 2019, and the country’s health ministry continues to maintain a dedicated programme for preventing death and illness from overwork. Microsoft Japan’s stated theme was to work for a shorter time, rest well and learn.

What the 39.9 per cent number actually measured

Microsoft defined labour productivity as sales divided by employee headcount. Its official results page reported that this measure was 39.9 per cent higher in August 2019 than in August 2018. It was not a count of tasks completed, code written or customer requests handled per working hour.

It was also a comparison between two different months a year apart, not a randomised controlled trial. Sales can move because of contracts, product cycles, pricing, demand and many other changes. There was no simultaneous Microsoft Japan control group working five days under otherwise identical conditions.

Microsoft’s clarification, dated 8 November 2019, is unusually direct. The figure was factual, it said, but was realised through various factors and not through this challenge alone. The company moved the number away from the list of direct project outcomes to prevent misunderstanding. Reporting the 39.9 per cent result is justified; calling it the isolated effect of Fridays off is not.

The missing day forced meetings to become smaller

Teams were asked to make meetings 30 minutes by default, keep attendance to five people where possible and use Microsoft Teams intensively. Staff could replace a meeting with chat or hold it online instead. A contemporary Japan Times report described the trial as an effort to improve efficiency in a country known for long working hours.

The share of meetings lasting 30 minutes rose 46 per cent compared with August 2018. This does not mean the number of meetings increased by 46 per cent. It means half-hour meetings became more prevalent within the meetings being held. The share of remote meetings rose 21 per cent compared with the April-to-June 2019 period.

That distinction reveals the experiment’s most concrete management lesson. A shorter week worked as a constraint. With less shared time available, organisers had to decide whether a meeting needed an hour, whether all invitees were necessary and whether the conversation required a meeting at all.

Less office time reduced inputs as well

The number of working days in August fell 25.4 per cent from a year earlier. Electricity consumption fell 23.1 per cent and printed pages fell 58.7 per cent, although both resource figures used August 2016 rather than August 2018 as the baseline. Different comparison years make the numbers less tidy than many summaries imply.

Employees broadly approved. Microsoft Japan’s survey found that 92.1 per cent rated the four-day-work, three-day-rest arrangement positively, while 94 per cent gave a positive rating to the summer project as a whole. Survey approval tells us how respondents felt about the trial; it does not establish lasting effects on retention, health or output.

The international attention was immediate. The Guardian’s coverage helped spread the image of an entire 2,300-person workforce receiving five consecutive Fridays off with no salary reduction. That description captures the scale, even if the causal language attached to the sales figure often became stronger in retelling.

Why one August cannot settle the argument

A one-month test has obvious limits. August can be seasonal, employees knew the arrangement was temporary, and a technology company already equipped for digital collaboration is not representative of hospitals, factories, shops or public services requiring continuous coverage. The experiment also bundled time off with meeting rules, communication changes and support programmes, making their separate effects impossible to identify.

It is also worth separating this design from a compressed four-day week. The fifth day became paid leave; the published plan did not describe employees packing Friday’s hours into four longer days. The label “four-day week” is used for both genuine reductions in working time and schedules that compress the same weekly hours. Evidence from one model cannot simply be transferred to the other.

Broader evidence has since strengthened the case that reduced working time can improve wellbeing. A 2025 Nature Human Behaviour study examined six-month, organisation-wide four-day-week interventions and found improvements across several wellbeing outcomes. Those organisations prepared by reorganising work before reducing hours. The evidence concerns a model of planned redesign, not the mechanical deletion of one day.

The same pattern appeared in a large British trial, where burnout fell and most employers continued the arrangement. Silicon Canals has previously examined what that trial found about shorter weeks. The strongest studies still do not promise identical results for every occupation or operating model.

The most useful result was the redesign

Microsoft Japan’s sales comparison deserves attention because it contradicted the fear that a missing day must produce a proportionate business loss. It should not be treated as proof of a 39.9 per cent causal gain. Microsoft explicitly warned against that interpretation.

The operational results are harder to sensationalise and easier to use. Half-hour meetings became more common, remote collaboration increased, offices consumed less electricity and most employees preferred the arrangement. The company had placed a firm limit on available time, and routines that once appeared fixed suddenly became negotiable.

That is why the experiment still matters. It did not show that every organisation can remove Fridays and expect sales to jump. It showed that roughly 2,300 employees could receive five paid Fridays away while management rebuilt ordinary work around the constraint. The calendar change attracted the headlines; the forced examination of wasted work was the part other companies could actually study.