One question Gallup uses to gauge if people are engaged at work sounds too personal to belong – but it turns out to be one of the strongest predictors of retention and well-being, a quiet reminder that even jobs ultimately run on human friendship

The first time Gallup tested it, just 30% of employees said yes — and those who did turned out to be seven times as likely to be engaged at work. The question itself is Item 10 of Gallup’s Q12 employee engagement survey, and it reads: “I have a best friend at work.” Not a colleague you respect. Not a manager who supports you. A best friend.

I thought it was a typo the first time I read it. Sitting in a list of perfectly sensible workplace statements, the kind about clear expectations and having the right tools to do your job, was a line that read more like something a child might ask. It feels out of place. Too personal to belong in a survey a company hands round at the end of the year. And yet that is exactly the point, and the reason it has stuck around for decades.

A quick note before I go further. I am not a psychologist or an organizational researcher. The findings here come mostly from one organization’s research, and they describe patterns across large groups of employees, not rules about you or your particular job. Take them as something to think about, not as a verdict on your own working life.

Workplace and human development leader Annamarie Mann has called the item “among the most controversial Gallup has asked in 30 years of employee engagement research.” Executives in particular tend to have a strong reaction to it, she notes. Friendship sounds like a soft thing, a nice-to-have, certainly not something you would measure alongside productivity.

What trips people up, by Gallup’s own account, is one small word. “They get stuck on the word ‘best,’” Gallup notes, because it feels exclusive, and most of us would struggle to name a single best friend among our coworkers. A good friend, sure. A best friend feels like a lot to ask of a place you go to earn a living.

So why not soften it? Why not change “best” to “good” or “close,” or drop the friendship language altogether and spare everyone the awkwardness? Gallup tried. According to its own analysis, “the item lost its power to differentiate highly productive workgroups” from mediocre ones. The blunt, slightly uncomfortable wording was doing real work. Sand it down and it stopped telling you anything.

Gallup says it uses this kind of deliberately extreme wording in several of its items precisely because it separates high-performing teams from low-performing ones on things like productivity, profitability, safety and retention. Tom Rath and Jim Harter put it plainly: early research showed that “having a ‘best friend’ at work was a more powerful predictor of workplace outcomes” than simply having a friend or a good friend. That is their account of their own research, not a settled law of human relationships but the discomfort the word causes seems to be the same thing that makes it useful.

Employees with a best friend at work are also more likely to produce higher quality work and have higher well-being. Gallup also says the item has, if anything, grown more important since the pandemic, even as more of us work remotely. Whatever else a person needs to be happy at work, it helps a great deal to have someone there they would want to grab a coffee with.

This fits my experience, too. I ran an adult language school in Vietnam for a few years, my first real management job. I went in assuming engagement was something you could build with the right perks and the right pep talks. It is not. You cannot manufacture it. It came down, almost every time, to whether the work and the person fit, and whether they had people around them they liked. The teams that hummed were the ones where people stayed for each other.

What I keep getting stuck on is this: friendship outpredicts most of the things companies actually spend money on. Compensation reviews, training budgets, engagement initiatives, the whole apparatus. And the strongest single signal anyone has found is whether two people on a team happen to like each other enough to call it a friendship. That is not really something a company can buy, schedule or roll out. It either happens between people or it doesn’t. Which leaves the uncomfortable question of what, exactly, an organization is supposed to do with a finding like that.