The four-day week we tried for six months

We announced it at a Monday morning meeting in February and the reaction was roughly what you would expect: delight from six people, immediate arithmetic from one, and a very quiet question from the person who runs our customer inbox about who would be answering it on Fridays. That question turned out to be the whole trial in miniature.

Six months later we kept it for most of the team, abandoned it for one function, and changed how we would run the trial if we did it again. What follows is the honest version, including the bits we got wrong, because the case studies you read about four-day weeks tend to be written by people who are selling something.

First, decide which four-day week you mean

This is where most conversations go wrong before they start, and we very nearly made the same mistake.

Reduced hours at full pay means the same salary for four days instead of five. Total hours fall — thirty-two instead of forty, say — and the bet is that output holds because the wasted fifth of the week was never productive anyway. This is what people usually mean by "the four-day week", and it is a genuine cost to the business if the productivity bet does not come off.

Compressed hours means the same total hours squeezed into four longer days. Forty hours becomes four tens. There is no productivity bet at all — you are rearranging, not reducing — but there is a real fatigue question, and ten-hour days do not suit everyone, particularly people with caring responsibilities and school runs.

A nine-day fortnight splits the difference and is far easier to staff, because half the team is off on alternate Fridays rather than the whole business closing.

We chose reduced hours at full pay, thirty-two hours over four days, with the business closed on Fridays. Choosing the closed-Friday version was the decision that caused nearly all our subsequent problems, and I would not make it the same way again.

The holiday sum nobody had done

Three weeks in, our bookkeeper asked a question that stopped the whole thing for a fortnight.

Statutory holiday entitlement in the UK is 5.6 weeks. For someone working five days a week, that is 5 × 5.6 = 28 days, which is also the statutory maximum — the entitlement is capped at 28 days however many days a week you work. For someone working four days a week, it is 4 × 5.6 = 22.4 days.

So what happens to the 28 days our staff already had in their contracts? There are only two honest answers, and both cost something.

If you keep the entitlement at 28 days while they work four-day weeks, you have quietly given everyone an extra week off on top of the day a week you just gave them — because 28 four-day-week days is nearly seven weeks of leave rather than 5.6. If you reduce the entitlement to 22.4 days, which is entirely lawful and still gives them the full 5.6 weeks, it *feels* to the team like you have taken six days of holiday away in the same breath as giving them a day a week. We had done none of this thinking, and the first person to work it out understandably assumed the worst.

We spent four months planning the productivity case and about ten minutes on holiday entitlement. Guess which one nearly ended the trial in week three.

Bank holidays make it worse. Ours fall disproportionately on Mondays, and our closed day was Friday, so a bank holiday Monday cost a four-day-week employee a quarter of their working week while costing a five-day employee a fifth. Over a year that is not trivial, and the fairest handling — converting the whole entitlement to hours rather than days, and deducting actual hours taken — is the one that requires the most explaining. It is the same arithmetic that trips people up in how much holiday staff are actually entitled to.

Do this sum before you announce anything. In hours, on paper, for every contract type you have.

Run it as a trial, properly

A four-day week for the same pay is a change to terms and conditions, and you cannot simply impose it — even a change everyone wants needs to be agreed, and it needs to be clear what it is. We wrote a short trial letter to each employee: the change, the start date, the end date, an explicit statement that terms revert automatically on that date unless separately agreed, and a note that we might end the trial early with four weeks' notice if it was not working.

That last sentence is the important one. Without it, you have a very difficult conversation ahead of you if the trial fails, because after six months of a four-day week the informal position is that it is now the norm, and unpicking it looks like a pay cut. The mechanics of doing this correctly are set out in how to change an employee's contract terms lawfully, and they are worth following even when everybody is enthusiastic.

One thing we did not anticipate: three people asked whether the four-day week meant they could also make a flexible working request for something else — earlier starts, a permanent home day. It is a fair question and the answer is yes, those routes run separately, with their own statutory timetable and consultation duties. Worth knowing before you are asked, and covered in flexible working requests: what small employers must do.

What worked and what broke

For the work that is project-shaped — where a person owns an outcome and controls their own time — it worked well and quickly. Two of that group said within a month that they were getting more done, and the reason they gave was not "I work harder on Thursdays". It was that the four-day frame killed the meetings. We went from eleven recurring internal meetings a week to four, because with 20% less time nobody would tolerate the standing forty-minute one that existed to update people who could have read an email.

That was the real finding, and it is slightly deflating: a big chunk of our productivity gain was not a new way of working. It was the removal of things we should have removed years earlier and had never had a reason to confront.

The function that broke was customer-facing. Our inbox and phones needed covering five days a week because our customers work five days a week, and there is no amount of internal efficiency that changes that. We tried rotating cover, which meant one person's four-day week landed on a different day each week, which is worse than a five-day week — it has all the coordination cost and none of the reliable long weekend. After ten weeks we moved that team to a nine-day fortnight with fixed alternate Fridays, and it has been fine ever since.

The other thing that broke was quieter. One person simply worked Fridays anyway. Not because we asked — because the work was there and he is conscientious. We only found out from the timestamps on his emails. If you run this trial, watch for that person specifically; a four-day week that becomes a five-day week with one unpaid day is worse than no change at all.

The measure we should have set on day one

We started with no baseline, which meant that for the first two months every discussion about whether it was working was just people asserting their prior view more loudly. That is the single biggest thing I would change.

Before you start, write down three or four numbers you already collect and could defend: jobs completed per week, revenue per head, response time to a customer enquiry, absence days. Not a survey about how people feel — those come back positive on day one of any trial and tell you nothing. Then set a review date and agree in advance what result would end the trial. Ours, eventually, was that revenue per head should not fall over two quarters and average first-response time to a customer should not exceed one working day. Both are imperfect. Both were infinitely better than arguing from anecdote.

Six months on, would I recommend it? For a business where most work is project-shaped and internal, yes — and the honest reason is not the day off. It is that a four-day week is the most effective forcing function I have ever seen for deleting the low-value work that accumulates in every company, which is the same discipline underneath building a business that doesn't need you every day. If your work is coverage-shaped, a nine-day fortnight gets you most of the benefit without the customer-facing problem.

And whichever version you pick, do the holiday calculation first.

Common questions

Is a four-day week legal in the UK?

Yes. There is no separate law for four-day working: it is simply a change to an employee's contractual hours, so it is governed by ordinary contract and working time rules. Because it changes terms and conditions, it needs to be agreed rather than imposed, even where everyone welcomes it. Employers commonly run it as a trial with a written letter setting out the change, a fixed end date, an automatic reversion to previous terms unless separately agreed, and a right to end the trial early on notice. Getting that reversion clause in writing at the outset matters, because after several months the four-day week feels like the norm and withdrawing it without a documented basis looks like a unilateral pay cut.

How does holiday entitlement work on a four-day week?

Statutory entitlement is 5.6 weeks, so it converts with the working pattern: five days a week gives 5 × 5.6 = 28 days, which is also the statutory cap, while four days a week gives 4 × 5.6 = 22.4 days. Both are the same 5.6 weeks. The practical problem is that staff on 28 days who move to four-day weeks will see a reduction to 22.4 days as leave being taken away, even though it is legally correct and proportionate. Keeping them on 28 days instead effectively grants nearly seven weeks of leave. Bank holidays complicate it further where they cluster on your non-working day. Calculating entitlement in hours rather than days is the cleanest way to keep it fair.

What is the difference between a four-day week and compressed hours?

A four-day week in the usual sense means reduced hours at full pay — for example thirty-two hours instead of forty, with the wager that output holds because the least productive fifth of the week disappears. Compressed hours means the same total hours rearranged into four longer days, so forty hours becomes four ten-hour days. Compressed hours involve no productivity bet and no reduction in paid time, which makes them far easier to justify commercially, but ten-hour days do not suit everyone, particularly people with caring responsibilities or school runs. A nine-day fortnight is a third option that is much easier to staff, because only part of the team is off on any given day.

How do you cover customer-facing work on a four-day week?

Usually not by closing the whole business on the same day, which is the mistake most trials make. Customers still operate five days a week, so any role whose value is availability rather than output needs continuous cover. Rotating which day each person takes off tends to be the worst of both worlds: it creates the coordination overhead without giving anyone a reliable long weekend. A nine-day fortnight with fixed alternate days off usually works better, because coverage is predictable and staff still get a regular extra day. Watch also for the conscientious employee who quietly works their day off anyway — a four-day week that becomes an unpaid fifth day is worse than making no change at all.