RTO Mandates vs. Remote Work: What the Evidence Says

An office mandate can improve attendance without improving the business.

Attendance, certainly. Performance, maybe.

That distinction gets lost when return-to-office policy becomes a public test of management authority. Amazon set a five-day expectation. JPMorgan moved employees on hybrid schedules to five office days. Neither decision settles the question for a different team.

An owner watching those decisions can reasonably ask, “Should I bring my team back too?” The useful answer is not pro-office or pro-remote. It is narrower:

Blanket RTO mandates have not shown a dependable general productivity gain. Hybrid work has the strongest randomized evidence for maintaining performance while improving retention. Fully remote work can perform well, but weak coordination, thin mentoring, and poor management can damage it. In-person work earns its cost when the task actually benefits from proximity.

I put the mandate papers beside the hybrid trials before deciding what they agreed on. Less than the headlines suggest. The studies do not identify one winning location. They show that task, career stage, coordination design, and employee choice change the result.

What the strongest RTO and remote-work studies say

The table below separates six questions that are often blended into one debate.

The most useful change came when I sorted the evidence by work arrangement instead of by conclusion. “Remote work” can mean an emergency move home, a permanent distributed company, two quiet days each week, or one office day each month. Those are different interventions.

The same is true of the office. A team brought together for mentoring and hard decisions is not experiencing the same policy as people commuting to sit alone on video calls.

RTO mandates do not have a general performance case

The University of Pittsburgh research is the cleanest answer to the broad claim that mandates improve a company. Yuye Ding and Mark Ma examined US companies in the S&P 500 and compared firms that imposed RTO mandates with firms that did not. Their difference-in-differences analysis found significant declines in employee satisfaction after mandates, but no significant change in financial performance or firm value.

That is not proof that every mandate failed. Firm-level measures are blunt. A policy could help one team, hurt another, and disappear inside the company average. Leaders may also value outcomes the paper did not measure well, such as apprenticeship, idea exchange, or a faster response during a crisis.

Still, the paper removes the easiest argument. An employer cannot point to the general evidence and say an RTO mandate is known to improve company performance. It is not.

I was close to writing a stronger anti-mandate verdict after that result. That would have repeated the same mistake in reverse. Absence of a broad company gain does not mean place never matters. It means a company should name the work problem, choose an arrangement that could fix it, and measure that result directly.

My view is firm here: if leadership cannot say what should improve besides office attendance, the mandate is not a productivity policy. It is an attendance policy wearing a productivity label.

Retention is not a side effect

The strongest randomized hybrid evidence comes from Trip.com. The company assigned 1,612 university-educated employees either to full-time office work or to a schedule with two home days each week. Over the trial and follow-up, hybrid work reduced quits by one-third. Performance grades did not fall. Promotion rates did not fall. Code output among engineers did not fall.

The manager result deserves equal attention. Before the experiment, managers expected hybrid work to reduce productivity. After experiencing the arrangement, their average view turned positive. The company then extended hybrid work more widely.

That sequence matters. Belief, experiment, measurement, decision. Not belief, mandate, enforcement.

The trial does have a boundary: it tested hybrid work at one Chinese technology company. It did not test a fully remote team. It also does not show that two home days are magic. The schedule was coordinated, the company already had formal performance reviews, and employees knew which days belonged where.

The retention concern becomes sharper in a separate working paper on Microsoft, SpaceX, and Apple. Researchers matched company policy changes with 260 million resumes. Their tenure-distribution analysis found that RTO policies shifted the workforce toward shorter tenure and roles below senior level. Departing workers tended to move to larger direct competitors.

Senior does not automatically mean better. Tenure does not guarantee current performance either. But losing experienced people removes company memory, mentoring capacity, and relationships that take time to rebuild. An RTO decision should price that risk before the resignation arrives, not after.

In delivery teams, location often becomes the suspect when owners cannot see handoffs, risk, and finished work clearly. The policy changes. The missing work record does not. That is the same management gap behind productivity paranoia, and an office badge does not close it. For teams working across countries, the employer's offshore management guide adds the engagement, payment, time-zone, access, and verification decisions that an RTO policy cannot settle.

The strongest case for the office is specific

Office advocates have a serious argument. It is just smaller than “people work better when watched.”

Physical proximity can improve feedback and learning. Natalia Emanuel, Emma Harrington, and Amanda Pallais studied software engineers at a Fortune 500 company from 2019 through 2024. Their revised working paper found that sitting near teammates increased coding feedback by 18.3 percent and improved code quality. The gains were concentrated among younger and less-tenured employees.

There was a cost on the other side. Experienced engineers wrote less code when they sat near colleagues. That is not necessarily a loss. They may have been spending time teaching, reviewing, and protecting future quality. It does show why an individual output score can punish the person creating team capability.

This is one reason mentoring juniors over video can be harder. The issue is not that a conversation cannot happen online. It is that small questions, overheard context, and unplanned feedback need more effort to start. Remote teams have to design those paths on purpose.

The result that changed my view most came in June 2026. A large multinational randomly assigned 248 fully remote customer-service employees either to remain remote or to attend one coordinated office day each month. In the monthly office-day trial, the office group eventually handled 7.8 percent more calls per hour, communicated more after meeting, received more manager feedback, and had one-third lower attrition.

The intervention was one coordinated day each month, with colleagues rather than alone at a desk.

That is a much better case for in-person work than five arbitrary badge swipes. It suggests that a small amount of coordinated contact can strengthen a remote system without removing the flexibility that helps people stay.

Specialized work can respond differently too. A 2025 working paper found that brokerage-firm mandates improved equity-analyst forecast accuracy and timeliness, while turnover rose. Analysts work under time pressure, interpret overlapping information, and learn through colleague interaction. The result is credible for that setting. It is not a blank cheque for every marketing, finance, support, or software team.

And some fully remote assignments perform worse. In a randomized study of data-entry workers in India, workers assigned home produced 18 percent fewer correct entries per minute. It was a narrow task, a short trial, and one labor-market setting. Still real. A balanced reading cannot pretend it away.

The uncomfortable conclusion is that opposite policies can both produce a gain, depending on the work. A remote employee can gain focus. A junior can lose feedback. A senior can produce less individually while making everyone around them better. A company can improve one measurable output and still raise turnover.

Why the studies appear to conflict

They conflict because “productivity” hides several outcomes and “remote” hides several systems.

The task changes the effect

Quiet analysis, writing, coding, and individual production can benefit from control over interruptions. Work built around rapid interpretation, apprenticeship, physical equipment, or dense coordination can benefit more from proximity.

The transition is not the steady state

An emergency move home tests disruption as much as location. A remote-first company with written decisions, clear owners, and deliberate gatherings is running a different system. Measuring both under one label creates a bad comparison.

Selection changes the workforce

People who choose remote work can differ from people suddenly sent home. People who comply with an RTO mandate can differ from people able to leave. After the policy, the company may be comparing two different workforces and calling the difference a location effect.

Measurement rewards what is easiest to count

Calls handled, data entries, code review comments, performance grades, forecast accuracy, revenue, and employee satisfaction answer different questions. Badge swipes answer only whether a badge was used. (A badge swipe is not a unit of output.)

This is also why firm-level research can look neutral while one role improves and another struggles. The average is not lying. It is averaging.

Coordination decides whether hybrid works

Hybrid can create the worst of both arrangements when schedules are uncoordinated. People commute without meeting the colleagues they need, while remote days remain packed with meetings. Fully remote work can create silos too. A natural experiment covering 61,182 Microsoft employees found that firm-wide remote work made collaboration networks more static and siloed, with fewer connections across groups.

That finding does not make a five-day office week the answer. It makes cross-team information flow a problem that management has to solve.

Make the office earn the commute

If a 60-person professional-services firm asked me whether to require several office days, I would ask which failure those days are meant to fix before choosing the days.

Use five tests.

1. Work: Which output, quality measure, decision, or customer result should improve?
2. Dependency: Which people need fast, high-context interaction with one another, and how often does that need occur?
3. Development: Who needs observation, feedback, coaching, or a denser learning environment?
4. Constraint: Which equipment, security rule, customer promise, or physical process requires a location?
5. Cost: What commuting, recruiting, retention, space, and focus costs will the policy create?

If leadership cannot answer the first question, stop. If the answer to the second is “everyone with everyone,” the organization chart is hiding the real dependency.

Then choose the location by work, not status.

I used to think the fairest hybrid policy was one company-wide number. It is simple, but simple can still be wrong. Equal treatment is not making a software engineer, lab technician, salesperson, and new support hire attend for the same reason. Fairness comes from a rule people can understand and a business need the rule can defend.

Test the policy before making it permanent

Treat a work-location change like an operating experiment.

Start with a baseline from one complete work cycle. Use measures the team already understands. Do not invent a new productivity score to justify the change.

Choose comparison groups where possible. A team testing coordinated office time can be compared with its own baseline or with a similar team whose arrangement did not change. Keep seasonality, staffing, and product deadlines in the notes. Not laboratory-perfect. Better than gut feeling.

State the review date before launch. Say which result would keep, change, or end the policy. That small promise stops a trial from quietly becoming permanent because leadership dislikes reversing itself.

Accountability does not require permanent observation

The fear underneath many mandates is simple: “If I cannot see people, how do I know the work is happening?”

Separate three records:

• Attendance evidence says where a person was or whether they were available under the agreed schedule.
• Time evidence says when work was recorded and which job, project, or customer it belonged to.
• Outcome evidence says what finished, how well it worked, and what happened next.

None can substitute for the others. Hours are needed for payroll, billing, capacity, and cost review. Attendance matters when a role or policy requires presence. Outcomes matter for performance. Turning one record into all three is how managers start rewarding visible activity instead of useful work.

The seven better measures in Activity Is Not Productivity give a stronger starting point for delivery and quality. The context-switching cost guide covers another issue an office can worsen when open-plan interruptions and scattered meetings consume focus.

Honestly, I think leaders are right to demand accountability. The part I reject is using location as a shortcut around the hard management work of defining commitments, reviewing quality, clearing blockers, coaching weak performance, and recognizing people who make the team better.

An office can help with those jobs. It cannot perform them for the manager.

A defensible RTO policy is narrower than a mandate

A strong policy names the purpose of in-person time, which roles it applies to, how exceptions work, and which results will be reviewed. It coordinates attendance so the needed people are actually together. It protects remote focus time instead of filling it with video calls. It gives managers discretion inside written boundaries, not secret exceptions for favored employees.

It also says when office attendance is the wrong tool.

Do not mandate a day to fix one underperformer. Manage the performance problem. Do not mandate a day because project status is unclear. Fix ownership and review. Do not mandate a day to create culture while rewarding only individual output. Repair the reward system.

Use office time when it can do something remote work cannot do as well. Protect remote time when it can do something the office cannot do as well.

The next useful RTO policy will not begin with a number of days. It will begin with the work.