Your Values Don't Fail on the Wall. They Fail in the Review.

About This Series

This is the third post in our seven-part Competency and Culture series. In our first post, we showed that most of a rating reflects the rater. In our second, we showed what happens when companies only score results and never score behaviour. This post is about the missing piece in both: how to describe your values so clearly that a manager can observe them, rate them, and be checked. Get this right, and the posts that follow, on culture fit, pay conversations, calibration, and competency design, all have a common language to work with.

Integrity appears in the published values of 65% of large companies. Researchers compared those values with more than a million employee reviews and found no link between what companies say they value and how employees rate them on it.

That should not surprise anyone who has worked in a growing company. Values are usually written in an offsite, printed on a wall, and added to the onboarding deck. Then the review cycle arrives, and the form asks about goals, delivery, and technical skill. The values are nowhere on it. Or they are there as a single line, "demonstrates our values," rated 1 to 5 with no description of what a 5 looks like.

That is where values fail. Not on the wall, where they look fine. In the review, where nobody knows how to score them.

Why do most company values change nothing?

The clearest evidence comes from the Culture 500 research by Donald Sull, Stefano Turconi, and Charles Sull, published in MIT Sloan Management Review. They compared the official values of more than 500 large organisations with over 1.2 million Glassdoor reviews, to see whether companies that emphasise a value are actually rated better on it by their own people.

They are not. The researchers found no correlation between the values a company emphasises in its published statements and how well employees say it lives up to them.

The problem is not the words. Integrity, collaboration, and respect are good values. The problem is that a word is not an instruction. "Collaboration" can mean sharing your work early, or agreeing with everyone in the room, or never pushing back on another team's plan. Ask five managers what it means and you will get five answers, and each of them will rate their team against their own version.

This is exactly the gap our first post described. When a value has no clear description, the rater fills the gap with their own standard. Rater bias does most of its damage in the space between a value's name and its meaning.

What happens when a stated value and daily reality come apart?

Safety has long been one of Boeing's stated core values. On 5 January 2024, a door plug blew out of an Alaska Airlines Boeing 737-9 MAX shortly after take-off. Nobody was killed, but the incident triggered one of the most detailed public reviews of a company's culture in recent years.

The following month, an expert panel convened by the US Federal Aviation Administration published its report on Boeing's safety culture, with 53 recommendations. It observed "a disconnect between Boeing's senior management and other members of the organization on safety culture." It also found that employees hesitated to report safety concerns for fear of retaliation, in part because the managers who oversaw those concerns also controlled performance evaluations, pay, promotions, and discipline.

Read that last point again, because it is the whole argument of this post. The value was on the wall. The review system was sending a different message. When the people who decide your rating and your pay are the same people you would have to report a problem about, the review system wins.

Boeing is a very different company from a 100-person scale-up, and we are not comparing the stakes. But the mechanism is the same at any size. A value only shapes behaviour if the review system rewards it and protects the people who act on it. Notably, Boeing's current values page now pairs each value with specific behaviours.

What does a value look like when you can actually observe it?

A value becomes measurable when you can describe it as something a person does, in a specific situation, that someone else could see. Not "shows ownership," but "raises a risk before the deadline, with a proposed fix." Not "collaborative," but "checks for an existing tool before building a new one."

The second thing that makes a value measurable is level. Ownership for a new engineer and ownership for a head of engineering are not the same behaviour. (We looked at how unclear ownership turns into a P&L problem in Role Ambiguity Isn't a Soft Problem, and at how to make ownership structural in Accountability Isn't a Personality Trait.) If you use one description for both, you will either set the bar too high for juniors or too low for leaders.

Here is what one value looks like when it is written for three levels, with three standards for each.

Three things make this work. Every cell describes something you could see or hear. Each level builds on the one below it, so the path to promotion is visible. And the "below the bar" row is as specific as the "role model" row, which is what makes a hard conversation fair.

This approach is not only for large companies. A growing number of scale-ups now publish their progression frameworks openly, describing the behaviour expected at every level. They do it because it makes promotion decisions more consistent, and because candidates can see what good looks like before they join.

How do you write behaviours from your company's own best and worst moments?

The worst way to write behaviour descriptions is in a workshop, starting from a blank page. You end up with polished sentences that sound right and describe nobody. The better way starts from what has actually happened in your company.

The method has a formal name, the critical incident technique, first described by John Flanagan in 1954, and it has been used to build rating scales for decades. In plain terms, it works in four steps.

1. 1Collect real moments. Ask a small group of experienced people (a few senior individual contributors, a couple of managers, and one or two of your strongest performers) to describe specific moments from the last year where a value was clearly lived, or clearly broken. Real situations, real outcomes. No names needed.
2. 2Pull out the behaviour. For each moment, write down exactly what the person did. Not what kind of person they are. "Shared the incomplete analysis two days early so the other team could plan" is a behaviour. "Great team player" is not.
3. 3Sort by level and standard. Group the behaviours by role level, and sort each group into below the bar, at the bar, and role model. You will find that most of your table writes itself.
4. 4Test it with a second group. Give the sorted behaviours, without labels, to a different group of managers and ask them to place each one. If they put a behaviour in a different place, the description is still too vague. Rewrite it until two groups agree.

The fourth step is the one most companies skip, and it is the one that makes the difference. It is a small calibration exercise before calibration ever happens. If two groups of your own managers cannot agree on what a behaviour means, they will not agree when it is attached to a real person's rating either.

How does this connect to the rest of the series?

Values written as behaviour are the shared language the rest of this series depends on.

They connect back to our first post. Rater bias does most of its work where descriptions are vague. A strict manager and a generous manager can both rate "ownership" fairly once they are looking at the same description of it.

They connect back to our second post. You cannot score how someone delivers until you have described what good looks like. The toxic producer is only visible when behaviour has its own column, and its own evidence.

They connect forward to hiring. In our next post, we look at why "not a culture fit" is the most expensive sentence in hiring, and how values written as behaviours give interviewers something better to test for than whether a candidate feels familiar.

The reframe underneath all of it is simple. A value is a promise about behaviour. A review is where you check whether the promise was kept. If the review cannot see the value, the promise was never really made.