Give people a free Friday and some of them will spend Thursday night paying for it. That was roughly what happened at Safeguard Global, whose 2023 four-day workweek pilot produced three distinct failure modes: employees who crammed five days into four and burned out, employees who worked Fridays anyway without saying so, and employees who followed the rules and felt they were falling behind. The company dropped the fixed schedule and let people choose their own hours and days.
The easy reading is that the four-day week doesn't work. We think the better one is narrower and more useful: the four-day week fails when it is imposed as a single schedule, because roles and people do not organise their time the same way, and a mandate turns that difference into hidden overwork. The replacement Safeguard chose, measuring outcomes instead of hours, is the right move. It is also harder than it looks, because outcome metrics are far better at spotting people who do too little than people who do too much.
Writing in HR Brew, Mikaela Cohen reports that CEO and founder Bjorn Reynolds launched the pilot to help recruitment and retention at the Austin-based workforce management company, which has roughly 1,000 remote employees in 78 countries. Each employee is now evaluated on three role-dependent outcomes; a customer service rep, for instance, is measured on account growth, customer satisfaction and points of failure. Reynolds' instruction to staff is simple: "Choose the hours that fit best for the work product you need, and if you can do that in four days, great, if you do it in five [days], but you're taking two half days, whatever works for you."
The schedule changed and nothing beneath it did
Inc. reports that the original trial covered about 20 engineers working across 26 countries plus one senior product leader, and that the leader usually worked at least part of the day off while managers had to remind others to log off. That is a small, engineering-heavy sample, and nobody should treat it as a verdict on the four-day week in general. What it does show clearly is what happens when a schedule changes and nothing beneath it does. Inc. puts it well: "A new policy cannot produce new results if leaders leave the system underneath it unchanged."
The larger trials make the same point from the other direction. A Nature Human Behaviour study of 2,896 employees across 141 organisations in six countries drew a sharp line between a genuine four-day week, with fewer hours at the same pay, and a compressed schedule in which people work longer days to earn the day off. The compressed version is the one that burned out Safeguard's staff. The participating companies also spent a preparation period redesigning work and cutting low-value tasks such as unnecessary meetings. Hours fell by about five a week on average, but not everyone reduced their hours by the same amount.
So even in trials built carefully to succeed, workers did not converge on a single schedule, which is exactly what a mandate assumes they will do. Buffer, after three years on a four-day week, found that 21 percent of its staff still worked more than four days, and responded by formally designating the fifth day as an overflow day for whoever needs it. That sits somewhere between a rigid rule and Safeguard's open choice, and it amounts to the same admission: one calendar does not fit every job.
A shared day off is fairness on paper
This is a serious argument, and in a synchronous team sharing one time zone it carries real weight. A collective day off is easier to defend than an individual one. But the UK results, summarised by UK Research and Innovation, came from organisations that committed to a 20 percent cut in hours with no cut in pay. Safeguard's pilot, by its own account, left some employees working longer hours Monday through Thursday to earn the Friday off. Holding up the UK data as proof that structure beats flexibility compares an explicitly shorter week with one that, for at least some people, was a squeezed one.
The fairness claim also rests on a premise that doesn't survive contact with most organisations. A uniform schedule treats a customer service rep fielding a stream of tickets and an engineer on a long build as though their work had the same shape. A schedule everyone shares is fair only if everyone's work has the same shape, and it never does: one person finds Friday genuinely free, another finds it a debt that comes due on Monday. The adherents at Safeguard who felt they were falling behind were paying for that equal treatment. And for a pilot spread across 26 countries, let alone a company operating in 78, a common day off was always common more in name than in practice, since one region's Friday overlaps with another's Thursday or Saturday.
Where the opposing view is right is on the risk. Flexibility can let overwork back in through the side door, and a policy of choice can become a policy of nobody ever being off. The answer to that risk is to measure the right things.
A schedule everyone shares is fair only if everyone's work has the same shape, and it never does.
Outcome metrics catch slackers, not martyrs
Reynolds' own example shows how the system is meant to work. If a rep's customers raise more tickets than average and wait longer than average for answers, that is an anomaly worth a conversation, which may turn out to be about working patterns, the client, or something qualitative. It is a sensible design, and it is built to detect shortfall.
An employee who hits all three metrics by working evenings and weekends looks like a star on an outcome dashboard. The secret Friday worker from the pilot doesn't disappear under a choose-your-hours model; they simply stop being visible, because there is no longer a Friday to be caught working on. Outcome measurement tells you whether the work got done. It says nothing about what the work cost. Neither HR Brew's reporting nor the research we reviewed offers evidence that outcome-based measurement improves retention, so for now Safeguard's model is a reasonable bet that has yet to be proven.
Before handing employees control of their calendars, leaders need a few things in place:
- A short set of outcome measures per role, specific enough that a manager can tell good work from busy work.
- A deliberate pass at removing low-value tasks, so freedom of schedule isn't spent sitting in the same meetings at different times.
- Some signal of sustained overwork, whether from hours data, workload reviews or plain conversation, sitting alongside the performance metrics.
- Managers prepared to open the anomaly conversation in both directions, including with the people who are overdelivering.
Safeguard was right to drop the mandate, and right that outcomes are a better basis for flexibility than seat time. The four-day week was supposed to give people time back, and a fixed schedule could not do that for jobs with different shapes. But a system that measures only results will reward the people giving up their weekends without a word, and call it performance. Freedom to choose your hours is a benefit only if someone is still paying attention to how many of them you work.
