Gallup's 70% manager claim is real. Almost every retelling of it is not.

Provenance: Method not published. The finding is the author's own and consistently stated, but no method, sample or technique is published, so nobody outside the organisation can check it. This describes how the figure is sourced, not whether it is true, and it is not a claim that anyone acted in bad faith.

Gallup's sentence, from pages 10 and 18 of State of the American Manager, is this:

Managers account for at least 70% of the variance in employee engagement scores across business units.

Three parts of it carry the limits, and all three are routinely dropped.

• At least. Gallup's own hedge. The figure is an estimate and they say so. Retellings state it as exact.
• The variance. A statement about differences between things, not about the level of anything.
• Across business units. The unit of analysis is the team, not the person. This word does the most work and goes missing most often.

What between-unit variance means

Picture every team in a company scored for engagement. Some score high, some low, and the spread between them is the between-unit variance. The claim is that at least 70% of that spread is associated with the manager.

That is worth knowing, and it is narrower than it sounds. It says nothing directly about how much of one person's engagement their manager drives, because people on the same team also differ from each other, and that variation sits inside a team rather than between teams. The claim does not reach it.

It also says nothing directly about turnover. Engagement and turnover are related, and Gallup publishes separately on that link, but a finding about engagement variance is not a finding about who quits.

What is not published

Neither the report nor Gallup's own article about it gives a named statistical method, a sample size for this particular analysis, a technical appendix, or a peer-reviewed paper to follow.

The report does give the scale of Gallup's wider research: engagement measured for 27 million employees and more than 2.5 million work units over two decades. That is the size of the programme, not the sample behind this estimate. The two get run together constantly, which is how a figure with no published method acquires the authority of 27 million data points.

None of this is an accusation. Gallup holds workplace data almost nobody else has, and the claim has been consistent across a decade of their publications. But a number carrying this much weight should be checkable, and this one is not.

How it gets restated

Four versions in wide circulation, and what is wrong with each. That 70% of engagement is caused by managers swaps variance for level, and association for cause. That managers control 70% of how engaged you feel moves the unit from the team to the individual, where the claim does not operate. That managers drive 70% of employee performance changes the outcome entirely. That 70% of turnover is down to the manager changes it to something this figure does not measure.

We found this on our own pages

Four pages on this site cite the claim. All four quoted Gallup correctly. Three then explained it in a way the claim does not support: one said most of what an engagement survey measures traces back to the manager, which is total variance and includes the differences between people on the same team; one substituted retention for engagement; and one slid from an engaged team to a team heading for the exit.

All three were corrected on 2 September 2026, and the quote itself was never touched on any of them. We are reporting it rather than fixing it quietly because a site whose whole argument is that a finding should travel with its boundary does not get to make this particular mistake in private. The trace that found it is the same one you are reading.

What the figure will and will not carry

It is sound for arguing that manager quality is where engagement differences between teams concentrate, for justifying investment in manager selection and training over company-wide perks, and for explaining why two teams under identical policies get different results.

It will not carry any claim about one person's engagement, any claim about turnover, performance or productivity, or any presentation of it as a measured and reproducible result rather than an estimate. It will not survive dropping at least, variance, or across business units, which is where the limits live.

How to cite it

Quote the sentence whole. Every qualifier in it is doing work, and the wording is the only thing carrying the boundary. A claim that cannot be independently checked should at the very least be repeated accurately.