People do not leave companies, they leave managers. It is the most quoted sentence in human resources, it has been in circulation for twenty-five years, and it is almost always accompanied by a figure: 70 percent.
The figure is real and it is published. It does not measure what the slogan says it measures.
What the 70 percent actually is
The number comes from Gallup, and the company states it carefully. On its own page explaining why great managers are so rare, the wording is that managers account for at least 70 percent of variance in employee engagement scores across business units.
Read that again with the qualifications intact.
Variance, not level. It says how much of the difference between business units is associated with the manager, not how much of any individual's engagement the manager causes. A company where every unit scores identically would have no variance for a manager to account for, and the statistic would say nothing at all.
Across business units, not across individuals. The unit of analysis is a team, not a person.
Engagement, not turnover. The outcome is a score on an engagement instrument. Resignation is a different outcome with a different literature.
At least, and an estimate. Gallup's own phrasing includes both hedges.
Gallup's evidence base for this is substantial in size. The same page reports engagement measured across 27 million employees and more than 2.5 million work units over two decades. The scale is not the issue. The issue is that a statistic about the variance in engagement scores between teams has been restated, step by step, as a statistic about why individuals resign.
The chain of restatement
The path from the research to the slide has four steps, and each one is small enough to seem harmless.
- Managers account for at least 70 percent of the variance in engagement across business units.
- Managers account for 70 percent of engagement.
- Engagement drives retention, so managers drive retention.
- People leave managers, not companies.
Step two drops the variance framing and the unit-level framing. Step three substitutes an outcome. Step four converts a partial association into an exclusive one, which is the only part that is obviously wrong on its face: nobody claims pay, workload, commute, career path and job security have no effect.
The slogan itself predates the statistic. It is generally traced to First, Break All the Rules, published in 1999 by Marcus Buckingham and Curt Coffman, both then at Gallup, and built on a large programme of interviews with managers. Interviews with managers are a legitimate source for a claim about what good managers do. They are not a design that can establish why employees left, because the employees who left were not the people being interviewed.
The part that cannot be checked
One feature of this claim is unusual for something so widely repeated: the underlying data has never been released for independent examination.
Gallup is a commercial research firm and its engagement database is its principal asset, so this is commercially rational and it is not an accusation of anything. But it does mean that the most quoted quantitative claim in people management cannot be re-analysed, replicated or checked by anyone outside the firm that produced it. When a vendor's 70 percent appears on a slide next to a peer-reviewed effect size, the two are not the same kind of object, and the audience is rarely told which is which.
What the turnover research says instead
Voluntary turnover has its own literature, and it is independent of the engagement industry.
The most comprehensive recent synthesis is Rubenstein, Eberly, Lee and Mitchell's 2018 meta-analysis in Personnel Psychology, which pooled the research on the antecedents of voluntary employee turnover and updated the effect sizes for both established and newer predictors. Among the correlations it reports are organisational commitment at −0.29, job security at −0.23 and engagement at −0.20, all in the expected direction: more of each, less quitting. It also finds that people who quit tend to do so earlier in their tenure, with a correlation of −0.27 between tenure and voluntary turnover across 117 studies.
Those are modest correlations, which is the first honest thing to say about turnover research: no single variable explains much of it. The second is that the strongest predictors in the list are attachments to the organisation and to the job, not measures of the immediate supervisor.
That does not mean managers are irrelevant. A manager shapes workload, recognition, development, whether a promise about a promotion is kept, and much of what a job feels like day to day, and all of those feed into commitment and engagement. The manager's contribution is real and it is mostly indirect, which is a less quotable finding than the slogan and a more useful one.
Why the correction matters practically
Believing that people leave managers rather than companies leads to a specific and expensive response: manager training as the retention strategy.
Manager development is worth doing, and the evidence behind it is reasonably strong, particularly at the junior end. But if a team is losing people because the pay bands have fallen behind the market, or because there is no route from the role to any other role, then training the manager is an answer to a question nobody asked, and it will be measured against a retention target it cannot move.
The diagnostic question is whether exit patterns cluster by manager or by something else. That is answerable inside most organisations with data they already hold, and it takes an afternoon. If leavers cluster under particular managers and not others, the slogan is doing useful work in that organisation. If they cluster by grade, by tenure band, by location or by function, it is not.
What is independently well documented is that the manager population itself is under strain: manager engagement fell five points in a year, considerably faster than the workforce they manage. A retention strategy that consists of asking that population to try harder has a problem regardless of what the 70 percent means.
Common questions
Is it true that people leave managers, not companies? Not as stated. The figure usually cited measures how much of the variance in engagement scores between business units is associated with the manager. It is not a measure of why individuals resign, and no claim that one factor exclusively explains turnover survives contact with the research.
What does Gallup's 70 percent figure mean? That managers account for at least 70 percent of the variance in employee engagement scores across business units. It concerns differences between teams, it concerns engagement rather than turnover, and Gallup states it as an estimate.
Where does the phrase come from? It is generally traced to First, Break All the Rules by Marcus Buckingham and Curt Coffman, published in 1999 and based on Gallup interview research with managers.
What actually predicts voluntary turnover? A 2018 meta-analysis in Personnel Psychology reports organisational commitment at −0.29, job security at −0.23 and engagement at −0.20 among the predictors, along with a tendency for leavers to go earlier in their tenure. All the correlations are modest; turnover is not explained by any single factor.
Do managers matter for retention at all? Yes, largely indirectly. Managers shape workload, recognition, development and whether commitments are kept, all of which feed into the commitment and engagement measures that do predict turnover.
Can Gallup's engagement data be independently verified? No. The underlying database is not released for external review. That is commercially normal and it does mean the claim cannot be replicated outside the firm.
How should we work out whether managers are driving our turnover? Look at whether leavers cluster by manager or by something else, such as grade, tenure band, location or function. Most organisations can answer that from data they already hold.
Sources
- Gallup, Why Great Managers Are So Rare
- Gallup, State of the Global Workplace 2026 Report
- Rubenstein, Eberly, Lee & Mitchell (2018), Surveying the forest: A meta-analysis, moderator investigation, and future-oriented discussion of the antecedents of voluntary employee turnover, Personnel Psychology 71(1), 23-65
