If you ask an employee how likely they are to leave, they’ll answer. And their answer will be reasonably related to how well they fit your company’s culture.
If instead you look at who actually left, the relationship shrinks. A lot.
That gap, between what people say they’ll do and what they end up doing, is the most uncomfortable finding in the entire culture-fit literature, and almost nobody selling assessments will tell you about it. We will, because it’s exactly the point where this starts being useful to you.
The promise that gets made badly
The standard industry pitch goes roughly like this: “hire for culture and cut your turnover by 40%.”
That sentence was built backwards. Somebody saw companies with good culture and low turnover, inferred causality, rounded up, and put it on a landing page. The problem isn’t that it’s false; the problem is that it can’t be audited. And when a number can’t be audited, it isn’t science: it’s marketing in a lab coat.
The honest question is a different one. How much of a person’s tenure is explained by their alignment with the organization, and how much by things that have nothing to do with you: a competitor’s salary offer, a move across the country, a new boss, a new baby?
The answer exists, it’s been measured, and it’s more modest than what you were sold. It’s also more useful.
The evidence, with its real numbers
There are two meta-analyses that are required reading here, and they’re worth reading carefully.
The first is Kristof-Brown, Zimmerman & Johnson (2005), published in Personnel Psychology. It’s the largest synthesis that exists: 172 studies and 836 effect sizes across different types of fit (with the job, with the organization, with the group, with the supervisor). What interests us is person-organization fit, or P-O fit: how well someone’s values and characteristics match the company’s.
These are the corrected values they report for P-O fit:
| What was measured | Correlation (ρ) | Studies / N |
|---|---|---|
| Intent to quit | −0.35 | 43 studies, 34,276 people |
| Intent to quit, measured on values | −0.46 | 32 studies, 18,222 people |
| Organizational commitment | +0.51 | 44 studies, 36,093 people |
| Job satisfaction | +0.44 | 65 studies, 42,922 people |
| Actual turnover (who left) | −0.14 | 10 studies, 2,157 people |
| Tenure | +0.03 | 28 studies, 6,036 people |
The negative sign in the first rows is the right one: more fit, less intent to leave. And the second meta-analysis, Verquer, Beehr & Wagner (2003), with a smaller sample (21 studies), points the same direction with smaller magnitudes: −0.21 with intent to quit, +0.28 with satisfaction, +0.31 with commitment. It also finds something that matters: when fit is measured as value congruence, the effects are larger.
Now the part that has to be said out loud.
A correlation of −0.35 is not a silver bullet. It explains around 12% of the variance in intent to quit. It’s a real signal, robust, replicated across more than thirty thousand people, and it still leaves most of the story unexplained. Anyone promising you that a culture assessment “eliminates” turnover is lying to you, and probably never even read the paper they’re citing.
And the drop from −0.35 to −0.14 is not a detail. It’s the difference between what people declare and what people do. A third meta-analysis, Arthur, Bell, Villado & Doverspike (2006), estimated the validity of P-O fit for predicting actual turnover at 0.24, somewhat better, but warned that the credibility interval included zero. Translation: there are studies where the effect practically disappears.
This doesn’t invalidate culture fit. It sizes it. And it tells you where to look.
Why values predict better than satisfaction
Look at the table again. The strongest correlation with the intent to stay doesn’t show up when you measure “fit” in general, but when you measure value congruence (−0.46 versus −0.35). Both meta-analyses land on this independently.
It makes sense when you think about what each thing is.
Satisfaction is a state. It goes up when you’re promoted, drops when your manager changes, sinks in March and recovers in April. It’s a thermometer, and like any thermometer it tells you today’s temperature, not whether the house is well built.
Values are a structure. How much autonomy matters to you versus predictability. Whether open conflict energizes you or drains you. Whether you need to see the impact of your work or it’s enough to do it well. That doesn’t change with a good quarter, and that’s why it predicts better at twelve and twenty-four months.
There’s a methodological nuance that reinforces the point, and it’s one the industry prefers not to mention. In that same 2005 meta-analysis, when fit is measured by asking the person how well they feel they fit, the correlation with commitment climbs to 0.77. When it’s measured by comparing the person’s profile against the organization’s actual profile, it falls to 0.32.
That difference is not evidence that “feeling like you fit” is more powerful. It’s common-method bias: you’re asking the same person, in the same questionnaire, two versions of the same question. That’s why we distrust any culture score that comes out of a single self-perception, and why fit has to be calculated against an organizational profile measured separately, not asked about. The full logic is in the culture-fit guide, and the instrument we use to do it is in the OCEAN model.
Two different turnovers you attack differently
Here’s the practical implication that gets lost in the debate.
When someone leaves, your company records “an exit.” But there are at least two completely different phenomena hiding in there, and confusing them makes you spend the budget in the wrong place.
Early turnover: it’s a selection problem
Someone who quits in the first six to twelve months almost never leaves because you failed them as an employer. They leave because they should never have come in. The process didn’t detect a misalignment that was already there the day they signed: values that clash with how decisions get made, an expectation of autonomy the role doesn’t give, a tolerance for conflict incompatible with the team.
That gets attacked before the hire. It’s the only turnover an entry assessment can move, and it’s exactly the terrain where P-O fit has something to say.
Late turnover: it’s a management problem
Someone who leaves in year three, after two good reviews, rarely leaves because of an original cultural misfit. They leave because of a manager, a ceiling, a promotion that never came, a project that bores them.
No selection assessment prevents that, and promising otherwise would be precisely the kind of overselling we criticize. That gets attacked with development, internal mobility, career paths. With the same measurement engine, yes, but asking the data a different question: what we cover in the talent lifecycle.
If your turnover is concentrated in the first year, you have a front-door problem. If it concentrates after the second, you have an inside problem. Different budgets, different teams, different tools. Look at it before you buy anything.
Why the engagement survey arrives late
The annual engagement survey is the most beloved and worst-used instrument in people management.
Not because it’s badly built. Because it measures, almost always, attitudes: satisfaction, engagement, declared intent. And attitudes are the symptom, not the cause. By the time someone marks a 4 out of 10 on “I’d recommend this company,” the misalignment that produced that 4 has been running for months. You already missed the window.
There’s another, quieter problem: the survey asks the people who stayed. The people whose misalignment was strongest are already gone, which is why the climate of a high-turnover team sometimes looks fine. It’s what attraction-selection-attrition theory has been pointing out since the eighties: organizations homogenize themselves because whoever doesn’t fit leaves. Your survey measures the survivors.
Use the survey. It’s useful for what it is: detecting problems with management, with bosses, with workload. But don’t ask it to predict who’s going to leave, because you’re measuring the smoke while the fire has already taken half the house.
What can be predicted at the front door, and what can’t
Let’s be explicit, because this list is where all the honesty of the matter lives.
You can reasonably estimate:
- How aligned a person’s values are with your organization’s, measured with separate, comparable instruments.
- Which working styles they’ll clash with and which they’ll flow with, based on stable personality traits.
- The relative probability of early misalignment compared to another candidate with the same technical level.
You cannot predict, and be suspicious of anyone who says otherwise:
- Whether this specific person will still be here in three years. Nobody can. The correlation with actual turnover is 0.14 to 0.24 depending on the meta-analysis, and that’s a population-level tendency, not an individual forecast.
- How they’ll react to a manager who hasn’t hired them yet, to a restructuring that hasn’t happened yet, or to an offer they haven’t received yet.
- Whether they’ll “be happy.” Psychometrics doesn’t measure that. Life does.
The value of measuring fit isn’t in certainty. It’s in the shift in probabilities. When you make fifty hiring decisions a year, moving a correlation from 0 to 0.35 in your selection criteria changes the aggregate visibly. On a single hire, it may change nothing. That’s how prediction works in behavioral science, and anyone offering you more is selling something else.
If you want the number for your own company, not for a meta-analysis, the cost of those failed decisions is broken down in the cost of a bad hire.
What to do this week
Three things, without buying anything.
One. Split your turnover in two. Take the voluntary exits from the last twenty-four months and separate them by tenure: under twelve months and over twelve months. If most fall into the first group, your problem is selection and no engagement initiative is going to solve it. It’s the cheapest analysis there is and almost nobody runs it.
Two. Write your values as observable behaviors, not as nouns. “Transparency” isn’t a value, it’s a word on a wall. “Around here, bad news gets said in the meeting, not in the hallway” is one, and you can actually evaluate someone against it. If you can’t describe your culture with sentences a reasonable candidate could reject, you don’t have a defined culture: you have a poster. How that translates into a measurable profile is covered in why generic assessments fail at culture fit.
Three. Look at the ones who stayed. The profiles of your high performers with more than three years are your best calibration data, and you already have them. Which values they share, which traits repeat. That’s the profile you should be comparing candidates against, not an ideal written by a committee. It’s literally what assessing your current team is for.
Culture doesn’t retain on its own, and anyone telling you an assessment cuts your turnover by 40% is selling you a correlation nobody measured.
What the data does show, with thirty years of research behind it, is that value alignment moves the needle in a consistent, moderate, verifiable way. On one decision you won’t notice it. On a hundred, you will. And that’s the only promise we can make without inventing anything.
Meet Talento Index and measure values fit with the same engine you use to hire. You bring the numbers; we make them comparable.
Sources
- Kristof-Brown, A. L., Zimmerman, R. D., & Johnson, E. C. (2005). Consequences of individuals’ fit at work: A meta-analysis of person–job, person–organization, person–group, and person–supervisor fit. Personnel Psychology, 58(2), 281–342. Corrected P-O fit correlations: intent to quit ρ = −0.35 (k = 43, N = 34,276); intent to quit with value measures ρ = −0.46 (k = 32, N = 18,222); organizational commitment ρ = 0.51; job satisfaction ρ = 0.44 (k = 65, N = 42,922); actual turnover ρ = −0.14 (k = 10, N = 2,157); tenure ρ = 0.03 (k = 28, N = 6,036).
- Verquer, M. L., Beehr, T. A., & Wagner, S. H. (2003). A meta-analysis of relations between person–organization fit and work attitudes. Journal of Vocational Behavior, 63(3), 473–489. 21 studies: intent to quit ρ = −0.21; job satisfaction ρ = 0.28; organizational commitment ρ = 0.31. Value congruence increases effect sizes.
- Arthur, W., Bell, S. T., Villado, A. J., & Doverspike, D. (2006). The use of person–organization fit in employment decision making: An assessment of its criterion-related validity. Journal of Applied Psychology, 91(4), 786–801. Estimated criterion-related validity of 0.24 for turnover (k = 8, N = 2,476) and 0.15 for performance; the lower bound of the credibility interval included zero in both cases.
- Schneider, B. (1987). The people make the place. Personnel Psychology, 40(3), 437–453. Attraction-selection-attrition framework.
Related OCEAN+ profiles
Discover which personality dimensions to look for in each role.
Related Articles
Self-Awareness: The Predictor of Whether Coaching Works
The gap between how someone sees themselves and how their team sees them predicts whether feedback lands or bounces. What the science says, and how to use it.
MBTI for Hiring: Why It Doesn't Work
50% of people get a different MBTI type when they retake the test. Here's what psychology says about its validity — and why it matters for your hiring process.
Talen.to's 10 behavioral archetypes (and why we calibrated them with our own data)
These aren't archetypes borrowed from a management book. They're centroids computed in a 7-dimensional space from real field data. How the model works, what it predicts, and why in-house calibration changes everything.