Meta's three-day office requirement took effect ten days ago. Amazon's has been running since May. Google has tied attendance to performance conversations. In August a video conferencing company asked staff living within commuting distance to come in twice a week, which the internet enjoyed more than the company did. Meanwhile the measured share of work done from home has barely moved in over a year. Survey data across large economies keeps landing in the same place, office occupancy indices in major cities have plateaued at roughly half of pre-2020 levels, and the gap between what is being announced and what is being observed has stopped narrowing.
Every mandate is an assertion about productivity. Almost none of them are accompanied by the measurement that would test it
That is the structural problem with this debate. Executives announce a return on the basis of collaboration, innovation and culture, three things nobody is measuring. Employees respond with commute time and focus time, two things they measure daily and personally. Both sides then reach for research, and the research is genuinely mixed, which means everyone can cite something. What almost no organisation does is measure its own outcome, before and after, against its own stated reason.
Three claims, and what the evidence actually supports
Individual productivity. The weakest claim. Studies of task-based output generally find remote work equal or slightly better for defined, measurable work, with some notable exceptions for complex collaborative tasks. Anyone asserting a general productivity gain from attendance is picking a fight they will lose, because the other side has the commute arithmetic. Collaboration and weak ties. Substantially stronger. Analyses of communication patterns consistently show networks becoming more siloed under full remote: people talk more to their immediate team and much less to everyone else. The connections that generate unplanned work are the ones that decay, and they decay quietly. Onboarding and early career. The strongest of the three, and the least contested by anyone who has managed a junior team remotely. Learning by overhearing, absorbing judgement by proximity, and receiving correction in small unscheduled doses are all hard to reproduce on a calendar. The honest case for the office is about new joiners, cross-team connection and ambiguous work. It is a good case. It is weakened every time it is presented as a productivity claim.
Attendance is the wrong metric, and it produces exactly what it measures
Mandates enforced on badge data generate badge behaviour. People arrive, take their scheduled video calls with colleagues in other buildings and other countries, and leave. The office is full, the car park is full, and the interaction the policy was designed to create has not occurred. If the purpose is connection, the meaningful measure is not how many days someone attended. It is whether the people who need to work together were in the same place on the same day. Most organisations enforcing three days a week have never calculated their team overlap rate, which is the only number that would tell them whether the policy is working.
Coordinated beats frequent
Two days chosen by the team is worth more than three days chosen by each individual. This is arithmetic rather than opinion: with individuals selecting their own days, the probability that a given pair of colleagues coincides is low, and it falls as the team grows. So set the anchor at team level, let teams pick their own days, and make the company-level rule a floor rather than a schedule. This is also the version employees object to least, because the constraint has a visible purpose.
The costs that do not appear in the announcement
Attrition does not fall evenly. It concentrates among people with caring responsibilities, people with long commutes, and specialists who can find another role in a fortnight — which is to say, disproportionately among the people you would have chosen to keep. The hiring radius contracts at the same moment, and any role you have struggled to fill becomes harder. And there is a credibility cost that compounds. A mandate justified by a productivity claim, with no measurement attached, teaches the organisation that leadership makes assertions it does not test. That lesson gets applied to the next announcement.
What to measure instead
Keep the list short, and baseline it before the change: team overlap days per fortnight, time to productivity for new joiners, the rate at which people form working relationships outside their immediate team, regretted attrition segmented by commute distance, and internal mobility. If those numbers move in the right direction, you have an argument. If they do not, you have an expensive attendance policy and a workforce that knows it.
| Stated reason | Measure to consider |
|---|---|
| Cross-team connection | Team overlap and working relationships outside the immediate team. |
| New-joiner development | Time to productivity with role-appropriate assessment. |
| Retention and opportunity | Regretted attrition and internal mobility, with proportionate segmentation. |
Qualitative summary of this article's source text, not a measured outcome or performance estimate.
Practical Guidance for Workplace Policy Analysis
- State the actual reason, and avoid productivity claims you cannot evidence.
- Set anchor days by team, not by individual choice or company decree.
- Measure overlap, not attendance.
- Baseline the metrics before the policy changes, not after.
- Treat onboarding as a separate case with a stronger requirement.
- Segment attrition by commute distance and watch it monthly.
- Document exceptions openly rather than granting them quietly.
- Design around the local calendar, not a head office template.
The Regional Angle
Three things make this debate land differently here, and the first is that it is largely imported. Offices in the Gulf refilled far earlier and far more completely than those in North America or northern Europe, and for most regional employers the question of whether people come in was settled two years ago. The live issue is therefore not whether to mandate attendance but what happens inside multinationals whose global policy is more flexible than local management practice. Regional staff read the group's hybrid working policy, observe that it is not applied in their office, and draw conclusions about whether regional roles are equivalent. Meanwhile regional leadership, operating in a business culture where presence signals commitment, regards the global policy as something written for another market. That gap is worth resolving explicitly rather than leaving as an unspoken local variation, because it surfaces in exit interviews and in the recruitment market long before it surfaces in a policy review. The second is contractual and immigration-related, and it is widely ignored. Employment contracts here specify a workplace, work permits are tied to an employer and often to a specific emirate or city, and the flexibility that staff actually exercise is not a day at home but several weeks working from their home country during the summer or around family obligations. That pattern is extremely common, almost always informal, and carries consequences nobody has assessed: personal tax residency, social security exposure, and the risk that an employee performing substantive work from another jurisdiction creates a taxable presence for the entity. The arrival of corporate tax in the United Arab Emirates this year makes the permanent establishment question considerably less theoretical than it was. The fix is administrative rather than restrictive — a short approval process for working outside the country, with a day limit, a record, and a check against the countries where you already have entities. The third is that the useful axis of flexibility here is seasonal rather than weekly. The genuine pressures on the regional calendar are Ramadan, with its reduced statutory hours and shifted energy through the day; the summer exodus when schools close and families travel for six to eight weeks; and the midday outdoor working restriction that already establishes, in law, the principle that schedules bend to conditions. A policy built on a Tuesday-to-Thursday template designed in California addresses none of these and will be ignored during the periods when flexibility matters most. Build the regional version around the actual calendar: formalise Ramadan arrangements rather than improvising them annually, make the summer pattern an approved arrangement with clear expectations, and accept that the weekly split is the least important variable in the equation.
The objection worth taking seriously
The strongest objection is that this treats an executive judgement as though it were a research question. Senior leaders can see things that do not appear in a dashboard: that a product team has lost its edge, that the graduate cohort is not developing, that decisions are taking three weeks because nobody will make them in a message thread. Demanding a controlled study before acting on that is a recipe for two more years of drift, and by the time the data is unambiguous the damage is done. There is also the unavoidable fact that the lease is signed and the building is largely empty, which is a real cost that no amount of measurement makes disappear. The judgement point is fair. Organisations are not laboratories, and waiting for clean evidence is itself a decision with consequences. But it argues for a different announcement rather than no measurement. Say that the leadership judgement is that the organisation is losing cross-team connection and failing its junior staff, that this is a judgement rather than a finding, and that here are the four numbers that will be tracked to see whether the change helps. That version is honest, it is hard to argue with, and it survives contact with a sceptical workforce. The version that asserts a productivity gain, produces no evidence, and enforces on badge data is the one that costs you the people you wanted to keep. The sunk cost of the building, incidentally, is not an argument for filling it — it is an argument for subletting part of it.
Common Questions
How many days is right?
There is no general answer, and the number matters less than whether the days are coordinated. Two aligned days will deliver more interaction than three unaligned ones.
Should new joiners have a different requirement?
Yes, and it is the easiest part of the policy to defend. The case for proximity is strongest in the first six months and weakest for experienced staff on well-defined work.
Does tying attendance to performance reviews work?
It produces attendance. Whether it produces the outcome the attendance was meant to deliver is a separate question, and it carries employment law risk in several jurisdictions.
What should we expect over the next twelve months?
Expect more large employers to link attendance to review outcomes and, in some cases, to compensation, and expect the first serious disputes about whether that is lawful where flexible working rights exist. Expect the aggregate remote share to stay roughly where it is, because the market has already found its equilibrium and announcements have stopped moving it. Expect quiet exemptions for hard-to-hire specialists to become the real policy inside most organisations, regardless of what the written one says. And expect property strategy to catch up, with desk ratios falling and the released space converted to the collaboration settings the mandates were supposedly about.
Workplace Policy Analysis — we baseline the four numbers that tell you whether your attendance policy is doing what you said it would, and build the version that fits the regional calendar rather than a head office template.