Office, Hybrid, or Remote: Does Location Actually Change Productivity?

By Victor Hoang · Updated Oct 10, 2026

Every few months a new study "proves" that remote work is great, or that it's a disaster. Executives pick the one that matches their instinct and move on. The trouble is that the studies are usually both right. They just measured different jobs, different people and different outcomes.

Our view is simple. Location is a weak lever on its own. The variables that move output are the type of work, how much of it needs tight coordination, and whether the team runs reliable rituals for handoffs, feedback and decisions. Get those right and the building matters much less. Get them wrong and no location policy will rescue you.

This piece goes through the strongest field evidence, says plainly who each study actually observed, and ends with a table you can use to sort your own roles.

Key Facts

• In a randomized trial of 1,612 employees at Trip.com (engineering, marketing, finance), working from home two days a week cut quits from 7.20% to 4.80%, a one-third reduction, with no significant difference in performance reviews.
• In a nine-month randomized experiment at Ctrip, volunteer call-center employees who worked from home were 13% more productive than office peers.
• In a randomized trial in India's data-entry sector, workers assigned to work from home were 18% less productive than those in the office.
• At a Fortune 500 retailer, remote call-center workers answered 12% fewer calls per hour than on-site workers; the authors attribute about 40% of that gap to remote work itself and 60% to who chose remote jobs.
• At a Fortune 500 firm, sitting near teammates increased coding feedback by 18.3%, with the gains concentrated among younger, less-tenured engineers.

Why the Studies Seem to Contradict Each Other

Line up the headline numbers and the confusion is obvious. Ctrip found a 13% gain from working at home. The India data-entry trial found an 18% loss. A retailer's call centers found a 12% gap in the wrong direction. How can all three be true?

Because they describe three different situations. Ctrip's experiment involved call-center employees who volunteered to work from home, and workers who choose a setup tend to fit it. The India trial randomly assigned data-entry workers, including people who would have preferred the office, and the authors found that workers who preferred home were 27% less productive at home than in the office, while workers who preferred the office were 13% less productive at home. Preference and fit shaped the result as much as location did.

The retailer study adds a second lesson. Its authors used the pandemic office closures to split the gap in two and estimated that 40% reflected remote work itself and 60% reflected who had chosen remote jobs. So when someone quotes "remote workers are X% less productive," ask first: were they assigned or did they choose, and what were they doing?

Notice also what these studies have in common. Most observed call-center or data-entry work, where output is countable, tasks are individual and coordination needs are low. That's a narrow slice of the jobs you're probably deciding policy for.

What the Hybrid Trial Does and Doesn't Prove

The most cited piece of evidence is the Trip.com experiment published in Nature by Bloom, Han and Liang. It's a good study, and it's worth being precise about it.

The sample was 1,612 employees in the Airfare and IT divisions, including 395 managers and 1,217 non-managers, mostly university-educated engineering, marketing and finance staff. About half wrote code. The treatment group worked from home on Wednesdays and Fridays; the control group was in the office five days. The trial ran roughly six months.

What it found:

• Attrition was 7.20% in the control group and 4.80% in the treatment group, a 33% reduction.
• Performance grades showed no significant difference across four half-year review periods, and the paper found no evidence of a difference in promotions.
• The retention effect was concentrated among non-managers, where quits fell by 40%. Managers showed no significant change.

What it doesn't prove: that fully remote work is fine, that it works for early-career staff who need mentoring, or that it works for roles with heavy customer or physical presence. It tested two days a week, for degree-holding knowledge workers, at a large technology company in one city. As a result, it's strong evidence that a light hybrid schedule is a cheap retention tool for that profile. It isn't a license to generalize.

Where Location Does Cost You: Learning and Cross-Team Ties

If location mattered nowhere, we'd stop here. It does matter, in two specific places, and neither is "raw output."

The first is learning. In a study of software engineers at a Fortune 500 firm, Emanuel, Harrington and Pallais found that sitting near teammates increased coding feedback by 18.3% and improved code quality. The benefit was concentrated among younger, less-tenured engineers, while experienced engineers wrote less code when sitting near teammates. In plain terms, proximity moves effort from today's output to tomorrow's skill. If your team is mostly seniors with clear priorities, that trade may not be worth it. If you're building a bench of juniors, it probably is. (This is an NBER working paper, so treat it as strong but not yet final evidence.)

The retailer call-center study points the same way. Remote workers there were less likely to be promoted, and the authors found suggestive evidence that remote work reduced training and one-on-one time with managers.

The second is cross-team connection. A Microsoft Research analysis of employees' work patterns before and after the March 2020 work-from-home mandate found that cross-group collaboration time fell by about 25% and groups became more siloed, with more email and instant messaging in place of richer channels. That's a real cost for any company whose advantage depends on new ideas crossing team lines.

Neither cost is inevitable. Both are coordination problems, which is exactly where work design can help.

The Real Variables: Role, Coordination, Rituals

Pull the evidence together and three variables explain most of the variation.

Role type. Independent, measurable work with clear outputs travels well. Work that depends on apprenticeship, ambiguous judgment, or live customer contact travels badly. A support queue and a product-strategy team need different answers even inside the same company.

Coordination need. How many other people must you sync with to finish a task, and how often? The more handoffs, the more a team pays for distance, and the more it gains from explicit decision rights and written context. Our piece on coordination cost covers why this overhead grows faster than headcount, and why work about work outpaces real work covers what to prune.

Team rituals. This is the one leaders control most directly. Teams that onboard, give feedback, and decide on a predictable rhythm lose far less to distance than teams that rely on hallway osmosis. When the old office was doing that work invisibly, removing the office exposes the gap. When rituals are deliberate, the location barely registers. Our article on async-first versus remote-first covers the written-decision side of this; here the point is narrower and about where people sit.

A Decision Framework by Role

Use this to sort roles, not people. Start with the dominant work type, then adjust for the team's stage.

Two cautions on reading the table. First, "default" means a starting hypothesis to test for a quarter, not a policy to publish. Second, one team can span rows. A senior engineer and a new graduate on the same squad may need different schedules, which is uncomfortable but honest.

What to Do Next

Run this over about 60 days rather than announcing a company-wide rule.

1. Classify your roles. Put every role on the table above. Resist a single policy for all of them.
2. Pick one outcome per pattern. Retention for knowledge staff, quality for queue work, time-to-independence for new hires. If you can't name the outcome, you can't judge the policy.
3. Fix the rituals before the rules. Write down how decisions are made, when feedback happens, and how a new hire gets context. Our pieces on decision-making velocity and deep work as an organizational strategy are good starting points.
4. Protect the juniors. If you have a bench of early-career staff, schedule in-person overlap with their mentors. Don't leave it to chance.
5. Audit your meetings. Distance tends to expose bad meeting habits. If your team is drowning in them, read why meeting-heavy teams lose people before blaming the schedule.
6. Measure and revisit. Track quits, performance review distribution, promotion rates and time-to-productivity by role pattern. Adjust after a quarter. For the basics of managing output remotely, our library guide on remote work productivity is a useful companion.

The leaders who get this right usually don't have a better location policy. They have a clearer idea of what each role is for and what the team needs to function, and they treat the building as one input among several.