Most companies already measure something. An annual engagement survey, a pulse tool, an eNPS score that moves two points a quarter and is discussed for ten minutes at the executive meeting. The complaint is almost never that there is no data. It is that the data stopped saying anything useful.
The backdrop does not help. Gallup found that global employee engagement fell to 20 percent in 2025, its lowest level since 2020. When the headline number is that low and that flat, a two point movement in your own score tells you nothing about whether anything you did worked.
The real problem is a category error. An engagement survey measures how people feel about working here. Culture is what the organization rewards, tolerates and punishes, which is a different thing entirely and shows up in behavior rather than sentiment. Measuring one and reporting it as the other is why so many culture programs run for two years with no evidence either way.
Separate the three things you are probably conflating
Engagement is an attitude: discretionary effort and attachment. Climate is the current weather, and it swings with a reorganization or a bad quarter. Culture is the underlying pattern of what actually gets rewarded, and it moves slowly.
They need different instruments and different cadences. Measuring culture with an engagement tool produces a number that goes up when people are happy, which is a useful thing to know and is not culture. A team can be delighted and still operating in a way that damages the company.
Decide which of the three you need before anyone selects a platform. Most executives asking to measure culture want the answer to a narrower question: is the behavior we said we wanted actually happening, and where is it not.
Write the behavior statements first
You cannot measure culture in the abstract. You measure the gap between specific behavior you have committed to and the behavior that occurs.
Take each stated value and convert it into two or three observable statements at a named level. “Accountability” becomes something like: when a commitment is going to be missed, the person who made it raises it before the deadline rather than at it. That is checkable. A reasonable observer could say whether it happened.
Six to ten statements is enough. If a statement cannot be confirmed or denied from the back of a room, rewrite it or drop it. This exercise alone tends to expose that the executive team never agreed what the values meant in practice.
Five sources that carry real signal
A survey is one input among several, and it is the weakest one on its own because it asks people to report on a system they are inside. Triangulate.
Decision archaeology
Pick ten significant decisions from the past year. Reconstruct each one: who raised it, how long it took, who had to approve, where it stalled, what evidence changed anyone's mind, and what happened to the person who disagreed. This is the single highest yield method available and almost nobody does it. Decisions are culture made visible.
Promotion and exit patterns
Look at who was promoted in the past twenty four months and write down what each of them is known for. That list is your real values statement. Then look at regretted exits, especially voluntary ones at high performing mid levels, and at who was quietly moved sideways rather than managed out.
Unprompted language
What people say when nobody supplied the categories is more informative than a rating scale. Open text responses, exit interview transcripts and public employer reviews all qualify. The MIT Sloan Management Review and Glassdoor Culture 500 project analyzed 1.2 million employee reviews across more than 500 large companies. It scored nine cultural values including agility, collaboration, execution and respect. That is a useful demonstration that unstructured text can be scored at scale, and you can run a smaller version on your own open text.
Calendar and forum design
Who sits in which meeting, and how long those meetings run. How much of each agenda is reporting rather than deciding, and how many meetings exist purely to prepare for another meeting. Structure encodes priorities more honestly than any values statement, and nobody games it because nobody thinks of it as a measurement.
What happens after a breach
Find two or three occasions in the past year when someone visibly acted against a stated value. Trace what followed. Nothing at all, a quiet word, a formal consequence, or a promotion three months later. Employees already know these stories in detail, and they weigh them far more heavily than any communication.
Cadence without survey fatigue
Culture does not move fast enough to justify monthly measurement, and asking monthly trains people to stop answering. A workable rhythm has three layers. One deep read a year, covering all five sources. A short behavior focused check at six months, on the two or three statements you are actively working on. And continuous passive signals such as promotion patterns and regretted attrition, which cost employees no time at all.
The discipline that matters more than frequency is the feedback loop. Publishing what you heard, what you will do and what you have decided not to do, within three weeks, does more for response quality next time than any incentive.
Turn the reading into something an executive can act on
A culture report that lands as a deck of percentages gets acknowledged and shelved. Report it as a small number of behavior gaps, each with the evidence behind it, each with a named owner and each with one system that would have to change.
Name the system explicitly, because that is the actionable part. Promotion criteria, the approval threshold, the escalation path, the performance conversation, who is invited to the Monday forum. Behavior follows the systems it is graded by, and pointing at a system gives the executive something they can actually change on Thursday.
Be alert to whether the gaps cluster at the top. When the same patterns keep appearing across every function, the origin is usually the executive team itself, and the behavior of that group is what needs measuring first.
The challenges nobody puts in the vendor demo
- Measuring the thing changes it. Once people know which behaviors are scored, reported behavior improves faster than actual behavior. Keep some observational sources that nobody is preparing for.
- Anonymity is thinner than you promise. In a function of nine people, cutting by tenure and level identifies individuals. People know this, so honesty drops precisely in the small units where the problems concentrate.
- Response rates are themselves a reading, and most reports ignore it. A department that drops from 80 percent to 45 percent participation has told you something more important than its scores.
- Benchmarks import someone else's problem. Comparing against an industry average can reassure you that you are normal while your own strategy requires you to be quite different from normal.
- Nobody wants the answer when it implicates the sponsor. The most common cause of a measurement program going quiet is not bad methodology. It is a finding that points at the person who commissioned it.
What it looks like when the measurement is actually working
You can tell within two cycles. Executives argue about the findings rather than about the methodology, which means they believe the data enough to disagree with the conclusions. Someone can name the system that changed as a result of the last read, by name, without checking their notes.
The language moves from adjectives to situations. Instead of “trust is low in operations,” the conversation becomes “three of the last five escalations from operations were raised after the deadline rather than before it.” Middle managers start requesting their own unit data rather than dreading it.
Above all, the measurement stops being an event. It becomes the thing that settles arguments about whether the culture work is doing anything, which is the only reason to run it. That distinction between measuring behavior and measuring mood is also the line between culture change and change management, and it decides which kind of program you should be funding.
If your current instrument has stopped producing arguments worth having, that is usually a design problem rather than a data problem. Our organizational development work starts by reading the systems that govern behavior, which is where the measurable signal lives.
Frequently Asked Questions (FAQs)
What is the difference between a culture survey and an engagement survey?
An engagement survey measures how people feel about working for you, which predicts effort and retention. A culture survey measures which behaviors are rewarded, tolerated and punished, which predicts how decisions get made. The scores can move in opposite directions, and a company can have high engagement alongside a culture that is quietly damaging the business.
How often should company culture be measured?
Once a year for a full read across multiple sources, with a shorter behavior specific check at six months on whatever you are actively working to change. Passive signals such as promotion patterns and regretted attrition can be tracked continuously because they cost employees no time. More frequent surveying lowers response quality without improving the picture.
Can you measure culture without a survey?
Yes, and the non survey sources are often stronger. Promotion and exit patterns, the reconstruction of real decisions, meeting and forum design, and what followed visible breaches of a stated value all describe behavior rather than sentiment. A survey is best used to test a hypothesis those sources have already generated.
What are the most useful culture metrics to track?
Regretted voluntary attrition by manager, internal promotion patterns against your stated behavior standards, and decision cycle time on significant decisions. Add participation rate changes by unit, and the gap between each behavior statement and observed practice. Avoid single composite culture scores; they move too little to guide action and hide the variation that matters.
How do you measure the gap between stated and lived culture?
Convert each stated value into two or three observable behavior statements, then gather evidence on each from at least two independent sources, one of which is not a survey. The gap is the difference between the commitment and the observed frequency. Reporting it that way keeps the conversation on specific situations rather than on whether people like working there.