Work From Home Productivity: What the Research Shows

What does work from home productivity actually measure?

Different studies measure different things, which is the source of most of the disagreement. Some count calls per hour, some count lines of code, some use manager ratings, and some simply ask people whether they feel productive. Those measures do not move together. A worker can answer more calls while handling each one worse; a team can ship the same number of features while accumulating rework.

The second problem is selection. People who choose remote work are not a random slice of the workforce. They differ in seniority, home setup, family situation, and — as one large study found — baseline productivity. A comparison that ignores who opted in will attribute the difference between types of worker to the location of the desk.

The third problem is the counterfactual. “Productive at home” compared to what? Very few studies can answer that. The ones that can are randomised, and they are worth reading closely.

What did the Ctrip experiment find?

The Ctrip study, published in the *Quarterly Journal of Economics* in 2015 by Nicholas Bloom, James Liang, John Roberts, and Zhichun Jenny Ying, is the most cited evidence that home working raises productivity. It is also the most misquoted.

The setting was a Chinese travel company’s call centre with about 16,000 employees. Of those who volunteered, 249 were randomly assigned either to work four days a week from home and one in the office, or to stay in the office full time. The experiment ran for nine months.

Measured performance rose by 13% in the home group. The authors decomposed it cleanly: about 9% came from working more minutes per shift, and about 4% came from handling more calls per minute. Attrition was roughly half that of the control group. Promotion rates for home workers were lower, though the estimate was imprecise.

That decomposition is the useful part. The gain was not brilliance; it was fewer sick days, fewer breaks, and a quieter room. The work was scripted, individually measured, and required almost no coordination — conditions that describe a call centre far better than most knowledge work.

> Common mistake: quoting the 13% as a general result for remote work. It is a result for volunteers in one routine job at one company, over nine months, where output per hour was already instrumented.

What happened when the same team ran it again?

In 2024, Bloom, Ruobing Han, and James Liang published a much larger experiment in *Nature*. It is closer to how modern teams work, and its headline finding is a null result.

The setting was Trip.com, the same corporate family as Ctrip, with 1,612 employees in engineering, marketing, finance, and accounting. Employees were randomly assigned to work two days a week from home or to stay in the office five days. The experiment ran six months, and performance was tracked for two years.

• Quit rates fell by 33%, from 7.2% to 4.8%.
• Performance grades showed no significant difference over two years. The study ran an equivalence test, the correct way to argue that an effect is small rather than merely undetected.
• Promotion rates did not differ, and lines of code did not change for the engineers.
• Managers’ beliefs about the productivity effect moved from −2.6% before the experiment to +1.0% after it.

The retention number is the real finding. At roughly $20,000 per departure, avoiding 2.4 percentage points of attrition across 1,612 employees is worth about $774,000 per cohort — a large benefit from a change that cost almost nothing in output.

The honest reading is not “hybrid makes people more productive.” It is “hybrid did not measurably change performance, and it reduced turnover.” Those are different claims, and the second is better supported.

Why did fully remote call-centre workers answer fewer calls?

A 2024 study by Jonathan Emanuel and Emma Harrington in the *American Economic Journal: Applied Economics* looked at call-centre workers at a large online retailer who chose to work remotely before the pandemic.

Remote workers handled 12% fewer calls per hour than office counterparts. The interesting question is how much was caused by remote work and how much was a difference in who chose it.

The pandemic provided a natural test. When offices closed and everyone worked from home, the gap narrowed to about 8%. The remaining difference cannot be caused by location, because everyone shared the same location. It reflects selection: the people who had opted into remote work were, on average, less productive to begin with.

So the rough split is 4 percentage points of causal effect and 8 points of selection. The study also found lower call quality and lower promotion rates among remote workers.

That finding should change how you read every remote-work statistic. If people self-select into remote work, that group will include a disproportionate share of people who were already struggling, and the averages will look worse than the policy deserves.

What happened to 10,000 IT professionals working from home?

The most detailed output data comes from a 2023 study by Michael Gibbs, Friederike Mengel, and Christoph Siemroth in the *Journal of Political Economy Microeconomics*. They analysed personnel and analytics data for more than 10,000 IT professionals at a large IT services firm.

Hours worked rose by about 30%, and 18% of that increase fell outside normal working hours. Output, measured in completed work items, fell slightly. Together, the authors estimated productivity fell by 8% to 19%, depending on adjustments for output quality.

The mechanism is visible: people worked longer, not smarter. That pattern fits a home environment that blurs work and rest, and coordination costs that eat into the working day. One caveat: this was an IT services firm with billable client work, where much of the job involves waiting on other people. The result describes that environment, not a solo analyst with no dependencies.

What does the wider evidence say?

Two more studies round out the picture.

A 2007 meta-analysis by Ravi Gajendran and David Harrison in the *Journal of Applied Psychology* pooled 46 studies covering 12,883 employees. It found small, mostly beneficial associations with self-reported performance — and one warning: telecommuting above about two and a half days a week was associated with worse coworker relationships.

A 2021 study by Prithwiraj Choudhury, Cirrus Foroughi, and Barbara Larson in the *Strategic Management Journal* followed US Patent and Trademark Office examiners who moved away from the office. They produced 4.4% more output with no increase in rework. Their work is individual, asynchronous, and measured by a clear output: a reviewed application.

The pattern is not about remote versus office. It is about task interdependence. Studies that find gains involve work one person can complete alone and that can be counted. Studies that find losses involve other people, tacit knowledge, or quality judgements that output counts miss.

Where do the gains and losses come from?

The gains are easy to reproduce. Commute time disappears, interruptions drop, and deep work becomes possible in blocks the office rarely allows.

The losses are subtler. Feedback gets delayed. A junior employee who would have overheard a senior colleague handling a difficult client misses that lesson. Weak-tie relationships decay, and those produce unexpected help across teams. Onboarding stretches from weeks into months.

There is also a measurement effect. Managers who cannot see work assume it is not happening — the “productivity paranoia” Microsoft documented, where 87% of employees said they were productive and 85% of leaders said hybrid made it hard to be confident. Employees who feel distrusted respond with visibility theatre: status messages, quick replies, calendar padding.

What do the numbers look like in practice?

Two worked examples show how the same evidence produces different answers for different teams.

Example 1: the retention value of two home days. A 500-person company with 15% annual attrition loses 75 people a year. At $18,000 per replacement, a policy that cuts attrition by a third, as in the *Nature* study, avoids 25 departures and saves $450,000. If the same policy cuts measured productivity by 2% against a $40 million payroll, it costs $800,000. If the effect is zero, as the study found, it returns $450,000. The decision hinges on an effect size most companies never measure.

Example 2: the hours trap. If a team’s output index stays at 100 while weekly hours rise from 40 to 52, output per hour falls from 2.50 to 1.92 — a 23% decline. The Gibbs study found a 30% rise in hours with slightly lower output: the same arithmetic from the other direction. Longer hours are not a productivity signal; they are often the symptom of coordination that no longer happens for free.

Does hybrid beat fully remote?

For most teams with interdependent work, the evidence favours hybrid, for a narrower reason than advocates give. The *Nature* trial tested two days at home and found no performance cost plus a large retention benefit. Emanuel and Harrington find a causal penalty of about 4% for fully remote work in a routine job, before selection. Gibbs finds 8–19% in coordination-heavy IT work.

None of those studies tested four or five home days against two. The honest conclusion is that two home days is the best-tested configuration, not the optimal one. And hybrid is not free: compress all collaboration into three office days and those days fill with meetings.

What actually predicts success?

The research points to a short list of conditions, and none of them is “trust” in the abstract.

Work that can be defined as an output. If you can say what done looks like, location matters less. If you cannot, remote work exposes the gap rather than creating it.

Written communication as a default. Teams that write decisions down lose less when people are apart. It is a skill that can be hired for and trained.

Managers who measure results. The Microsoft data suggests the confidence gap is a management gap. A team with clear weekly commitments and a visible board does not infer productivity from green status dots.

Deliberate office time. If people come in to sit on video calls, the office is a commute with worse coffee. Reserve office days for onboarding, planning, and difficult conversations.

Fair evaluation. Remote workers can be penalised twice: by visibility bias in promotions, and by the selection effect that made them remote. Calibrate reviews against written output.

What are the trade-offs and common mistakes?

Treating surveys as evidence. Self-reports are useful for morale and useless for productivity. They correlate with commute length and home-office quality more than with output.

Rolling out full remote for apprenticeship roles. Junior employees learn by proximity. A fully remote graduate programme can work, but it requires structured mentoring most companies never build.

Ignoring selection. If remote work is optional, the people who take it are not a random sample. Compare like with like, or run a proper pilot.

Confusing hours with output. The IT study is the clearest warning: 30% more hours and slightly less output. A dashboard that counts activity will reward the wrong behaviour.

Skipping the pilot. The evidence justifies a trial, not a mandate. Run a six-month pilot with a control group, a pre-registered productivity measure, and a retention measure. It is not expensive at company scale.

How do you measure work from home productivity honestly?

Pick two or three outcome measures before the policy starts and keep them stable for at least two quarters: cycle time from assignment to done, rework rate, client satisfaction, voluntary attrition. Add a quality measure, because volume alone rewards speed at the expense of everything else.

Then make the work visible without surveilling people. That means a shared board where commitments, owners, and due dates live, and a weekly review that looks at outcomes rather than activity.

Doitify is built for that layer: projects, tasks and sub-tasks, checklists, dependencies, workload views, and reports in one workspace, plus remote-team management features for distributed groups. Doitify is our product, so treat this as a description rather than a review. A team that only needs a shared task list will do fine with something simpler; the value appears when output has to be tracked across projects and people without falling back to status meetings.

Conclusion

The evidence does not support a verdict on remote work, only conditional statements. Hybrid at two days a week is close to performance-neutral and good for retention. Fully remote work is fine for individual, well-defined tasks and costly for work that depends on other people. Self-reported productivity is not a measurement.

The practical implication is unglamorous. Define outputs, write things down, keep shared time, and measure the effect before scaling the policy. Companies that do that get a defensible answer in two quarters. Companies that argue from anecdote keep having the same argument, with better slides each year.

If this post on work from home productivity was helpful, you might also enjoy How to Write a Project Status Report (5 Steps) and Team Productivity Metrics Every Manager Should Track (2026).