---
title: "Hybrid Working and Retention: What a Randomised Trial Found About Quit Rates"
description: "A randomised trial of 1,612 employees found hybrid working improved job satisfaction and cut quit rates by a third, with no measurable effect on performance grades or promotions. Here is what that means for hiring."
author: "Marcus Telle"
category: "Reading the Market"
date: 2026-10-16T17:45:03.024Z
canonical: "https://theartofpeople.io/blog/hybrid-working-staff-retention-trial-hiring"
---

# Hybrid Working and Retention: What a Randomised Trial Found About Quit Rates

![cook, at home, to bake](https://cdn.pixabay.com/photo/2021/02/24/15/21/cook-6046729_1280.jpg?w=1200&q=75)

> A randomised trial of 1,612 employees found hybrid working improved job satisfaction and cut quit rates by a third, with no measurable effect on performance grades or promotions. Here is what that means for hiring.

Almost every argument about working from home runs on conviction and anecdote. One of them was run as a proper experiment instead, and the result is worth knowing if your job involves persuading people to join a company.

The study reached Nature in 2024. Over six months, 1,612 staff at a technology firm in China were [split at random into two arms of a controlled trial](https://pmc.ncbi.nlm.nih.gov/articles/PMC11208135/): one arm in the building every weekday, the other in for three days and at home for the remaining two. Nicholas Bloom wrote it up with Ruobing Han and James Liang.

## What the Trial Actually Measured

The abstract does not equivocate. The authors record a hybrid pattern that [lifted job satisfaction and cut resignations by a third](https://pmc.ncbi.nlm.nih.gov/articles/PMC11208135/). That fall reached statistical significance in three groups: staff below management level, women, and anybody with a long journey to work. Equivalence testing found no effect on performance grades across the following two years of reviews.

Two further findings matter for anyone hiring. Across the same two-year window, promotion rates were indistinguishable between the arms. And among the engineers in the sample there was no effect on the volume of code written. The standard anxieties, that flexibility quietly erodes output or stalls a career, did not appear in [this trial](https://nature.com/articles/s41586-024-07500-2).

Randomisation is what sets this apart. Almost everything written about remote work contrasts populations who picked their own arrangements, and such populations differ in a dozen respects that have nothing whatever to do with buildings. Chance did the picking here, which makes the difference between the two arms fairly attributable to the pattern itself.

## Plausible Explanations, None of Them Proven

A trial of this kind establishes an effect without explaining it. Several readings are reasonable, and each is a hypothesis you could test inside your own company rather than a conclusion you can borrow.

Commuting is the obvious candidate. The fall in resignations was significant for people with long journeys, and two mornings reclaimed from a long one is a tangible return in hours and in energy. Nobody needs a spreadsheet to feel it.

Autonomy is another. Anybody allowed to shape a portion of their own week tends to describe the feeling as being trusted. Trust resists measurement, and it shows up anyway in how somebody talks about their employer at a dinner party.

Life stage is a third. The same result appearing among non-managers and among women points towards flexibility mattering most to the people whose weeks contain the least slack. That matches what recruiters hear constantly from candidates and rarely write down. On the related question of why good people sometimes leave with no visible build-up, our piece on [the resignation that arrived from nowhere](/why-your-best-employee-just-quit-without-warning-mny1) covers the signals that were there all along.

## From One Trial to the British Picture

A single company in a single country is a thin basis for a policy, so it helps to set the trial beside national figures. The Office for National Statistics, drawing on 5,494 adults in work within its Opinions and Lifestyle Survey, reports that [28% of workers in Great Britain were hybrid workers between 8 January and 30 March 2025](https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employmentandemployeetypes/articles/whohasaccesstohybridworkingreatbritain/latest). That average conceals a steep gradient. In the [same release](https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employmentandemployeetypes/articles/whohasaccesstohybridworkingreatbritain/latest), 45% of workers earning £50,000 or more hybrid worked, against 8% of those earning under £20,000. Flexibility in Britain is, for now, largely a benefit enjoyed by the better paid, which is worth knowing before you describe yours as generous.

The retention claim also has a British echo. In its 2025 report on flexible and hybrid working practices, the CIPD found that [3% of employees said they had left a job in the previous year because of a lack of flexible working](https://www.cipd.org/uk/knowledge/reports/flexible-hybrid-working/), which it puts at roughly 1.1 million workers. The same report has 91% of employers offering some flexible arrangement and 65% requiring a minimum number of days on site, so the argument in most British workplaces is no longer whether, only how many and which. The report states no sample size on its landing page, so read it as a well-sourced indication rather than a measurement.

One more thing belongs on the table, and it is not a statistic. In Great Britain, employees can ask to change when, where or how they work from their first day, and [gov.uk states that they may make two such applications in any twelve-month period](https://www.gov.uk/flexible-working/applying-for-flexible-working). The rules differ in Northern Ireland. A UK government factsheet dated February 2026 says the Employment Rights Act 2025 will add a duty to consult before refusing a request and a test of whether the refusal is reasonable, and that [these changes will take effect in 2027](https://assets.publishing.service.gov.uk/media/69a6bf1ca2495f2d259f1397/factsheet-flexible-working-feb-2026.pdf). We write as editors, not as lawyers, and anyone altering contracts should take advice in their own jurisdiction. The point for a hiring plan is simply that your pattern will be tested one request at a time, whatever the policy document says. A comparable trial of a different lever is covered in our piece on [what the four-day week trials found](/four-day-week-hiring-uk-trial-results).

## Name the Days, Out Loud, Early

Candidates raise location early, and vague answers cost you more than unpopular ones. It helps to have one sentence prepared and memorised: Tuesday and Wednesday are shared days in the building, and the remaining three are yours to arrange. Specific beats flexible every time, because specific can be planned around.

Supply the reasoning alongside the rule. A line about doing the week's planning face to face reads as deliberate design. A line about attendance being expected reads as an argument somebody lost.

Then write it down twice, once in the advert and once in the offer letter. Nothing then surprises anybody in month two, and surprise is one of the more reliable precursors to a new starter concluding they were sold a different job. Those people begin looking quietly, and you find out last.

## Point the Flexibility Where It Lands Hardest

The trial suggests the effect concentrates among staff with long journeys, among non-managers and among women. Where your hardest vacancy draws mainly on those populations, lead with the working pattern instead of burying it beneath the benefits list. Somebody travelling ninety minutes in each direction reads that line before they read anything else on the page.

Ask the question too, rather than only answering it. Asking what their ideal week would contain costs ten seconds, tells you what is worth putting on the table, and gives you room to be straight about the parts of your pattern that will not suit them. Honesty at that stage saves a withdrawn acceptance later. The reverse case, where a company tightens its policy instead, is worth reading about in our piece on [what a return-to-office mandate does to hiring](/return-to-office-mandate-hiring-impact).

## Put a Price on Every Resignation

Turn the debate into arithmetic. Total up the agency fee or advertising spend, the weeks of work nobody is doing, the interviewing hours taken out of your managers' diaries, and the ramp-up period before a replacement is fully productive. For most skilled roles that total represents a serious fraction of an annual salary. Against a number like that, a meaningful reduction in resignations pays for a great deal of flexibility, even in a small team. Our guide to [calculating your turnover rate](/employee-turnover-rate-how-to-calculate) gives you the denominator for that sum.

The same arithmetic disciplines the argument in the other direction. If a rigid pattern costs you two departures a year from a team of ten, then the pattern has a price, and the person defending it should know what that price is. Put the figure on the whiteboard. A dispute about preferences becomes a decision about money, and those conversations finish faster.

## Run Your Own Small Trial First

The reason this research is useful is that somebody treated it as an experiment. Instead of announcing a company-wide policy and then defending it for a year, give one team a pattern for a single quarter. Log the resignations, the output and the way people describe the arrangement, then make the call. Nothing about that will be as clean as a randomised study, and it still beats settling the matter by whoever is loudest.

Be equally wary in the opposite direction. Flexibility is not free, and it is not universally applicable. Some work genuinely needs people in one room, and some teams need considerably more shared time than others. The test of a reasonable rule is whether you can describe it to a candidate in one sentence and not feel the need to apologise.

How much of this needs writing down changes with size. Our piece on [how hiring shifts between the fifth employee and the twenty-fifth](/nuances-of-hiring-at-each-stage-5th-to-25th-employee) covers that progression. Five people can settle a pattern over coffee. Twenty-five cannot, and the attempt produces exactly the kind of quiet resentment that shows up later as a resignation.

## The Call We Would Make on Mandated Days

Here is our position, and it is about practice rather than principle. Most companies are fighting the wrong battle. They argue about the number of mandated days, write the number into a policy, and then spend a year defending it. On the evidence above, the number is the least interesting variable in the room. What predicts whether a pattern survives is whether somebody owns the days and can explain them without apologising.

So we would set the pattern narrow, write it in the advert and the offer letter, and then handle individual requests generously and on their merits, rather than setting the pattern generously and policing it tightly. The second approach looks kinder and ages badly: every exception becomes a precedent somebody else heard about, and the manager spends the year adjudicating instead of managing. Teams that go the other way tend to find the conversation stops being about buildings within about two quarters.

The cost nobody budgets for sits here as well. A hybrid pattern asks more of a manager, not less. Somebody has to notice who is quietly drifting, run the planning conversation that used to happen by accident beside a desk, and make sure the two people who never come in are not slowly written out of the interesting work. That is real hours, and it comes out of the same week as everything else.

Work that is genuinely co-located sets a hard limit: a workshop, a laboratory, a kitchen, a ward. There, flexibility has to be built from rotas and predictable shifts rather than from location, and pretending otherwise insults everybody. Outside that, the test we would apply is the one at the end of this piece. Can you say your pattern in a single sentence to a candidate, and does the sentence need an apology?

## What the Trial Cannot Tell You

Be careful about how far this travels. The sample is one technology company in one country, observed across half a year, with a specific pattern of three office days and two at home. It says nothing about fully remote arrangements, nothing about four-day weeks, and nothing about what happens to a team that has never shared a room at all. Nor does it settle the question of what the arrangement does to people who join afterwards, since every participant in the study already knew their colleagues.

Treat it, then, as strong evidence against one specific fear rather than as a mandate for a particular policy. The fear it addresses is that letting people work from home part of the week quietly costs you performance and promotion. On the evidence of this trial, over two years of review data, it did not.

## Where to Get Advice

If you are altering contracts, ask an employment adviser in your own jurisdiction, or a recruiter who works your market. Contractual terms, expenses for home working and the right to request flexible arrangements all differ by country, and this article is editorial commentary rather than legal advice.

Outside help earns its fee when you cannot work out whether a pattern is doing any good. Someone who talks to the people leaving you and the people considering you is watching both halves of the market, and will know what your competitors put on the table this quarter.

On its own the trial settles nothing. One employer, one country, six months. What it hands you instead is a sharper question than whether anybody works hard at home. Ask what each departure costs this company, and what you would genuinely change to stop the next one.

## FAQ

### Does hybrid working really reduce resignations?

In the randomised trial published in Nature in 2024, covering 1,612 employees at a Chinese technology company, hybrid working improved job satisfaction and reduced quit rates by one-third. The effect was significant for non-managers, female employees and people with long commutes. It is one company over six months, so treat it as strong evidence rather than a universal result.

### Did performance suffer in the trial?

No measurable effect appeared. Equivalence testing found no difference in performance grades over the next two years of reviews, no detectable difference in promotions across the same period, and no effect on the volume of code written by the engineers in the sample.

### What should I put in a job advert about our working pattern?

One specific sentence naming the days, plus the reason behind them. A named pattern a candidate can plan around is far more persuasive than the word flexible, and repeating it in the offer letter prevents the month-two surprise that makes new starters feel misled.

### How do I know whether our pattern is costing us people?

Price a resignation. Add the agency fee or advertising cost, the weeks the work goes undone, your managers' interviewing hours and the ramp-up time for a replacement. Then compare that against your departures. A recruiter who talks to both your leavers and your candidates can tell you what competitors are offering.


---
Source: https://theartofpeople.io/blog/hybrid-working-staff-retention-trial-hiring