KPIs That Actually Change Behavior — Lessons From 7 Years of HR With 650 Employees

A KPI spreadsheet takes one afternoon. A changed behavior doesn’t.

I spent seven years as HR director in an FMCG company with over 650 employees: AquaTrade and SINCER, the group’s HoReCa brand. There I built the performance management system from scratch, digitized the HR processes and trained the company’s managers. That is also where I understood why most of the indicators we put on paper move absolutely nothing.

Here is what I learned, without the jargon.

Why most KPI systems don’t work

They measure things the person doesn’t control. You give a salesperson an indicator that depends on production, on logistics, on the exchange rate and on the market. They look at it the way they look at the weather: they comment on it, they don’t influence it. From the day an indicator becomes “bad luck,” it stops changing behavior. It only generates excuses.

There are too many of them. A role with ten rows of indicators doesn’t have ten priorities. It has none. When everything counts equally, people do what they already knew how to do and tick the rest at the end of the month.

They’re reported too rarely. An indicator you see once a quarter isn’t a management tool. It’s an autopsy. By the time you find out, the quarter is already lost, and the discussion becomes about blame instead of correction.

They’re not connected to anything except the bonus. If the only moment the indicator comes up is when pay is calculated, you’ve built a negotiation system, not a performance system. Smart people will optimize the number, not the work. And they’re right: that is exactly what you asked them for.

Result or behavior — the distinction that changes everything

People talk a lot about “leading” and “lagging” indicators. You don’t need those terms.

You need one distinction: the result is what happened, the behavior is what you do to make it happen.

Last month’s sales figure is a result. It’s true, it matters, and it’s completely dead — there is nothing left to do with it. The number of new clients contacted this week is a behavior. You can act on that one today, at four o’clock.

The same logic everywhere:

• Staff turnover is a result. The number of one-on-one conversations a manager has with their team is a behavior.
• Customer complaints are a result. Following the communication standard at reception is a behavior.
• Margin is a result. The number of quotes sent without an automatic discount is a behavior.

The rule I give managers: results get reported, behaviors get managed.

A director who looks only at the month’s number is driving with their eyes on the rear-view mirror. They can see exactly where they have been. About where they are going — nothing.

What I actually did at AquaTrade

650 people, FMCG, field pace. We didn’t have the luxury of a system that looked beautiful on paper and couldn’t be used.

I built the performance management system starting from one question: “what does this person have to do differently on Monday morning?” If an indicator didn’t answer that question, I took it out.

I digitized the HR processes so the data would stop arriving a month late. An indicator that arrives a month late isn’t an indicator, it’s a story.

I trained the managers. That was the hardest part and the most important one. A KPI system without managers who know how to hold the conversation is nothing more than a spreadsheet. The manager is the one who turns a number into a decision.

And I developed from scratch a tool for measuring satisfaction, motivation and engagement at work. Here is the nuance a lot of companies miss: satisfaction and engagement are early signals. Turnover is the result that shows up much later. If you look only at how many people left, you always find out after there is nothing left to do.

The result, company-wide: staff turnover dropped by 10%.

Not because of one magic reason. Because we started measuring things people could actually influence, and talking about them before they turned into resignations.

What makes a KPI work

Out of everything I tried, four things mattered.

Few. Two, three at most for one role. If you can’t name someone’s indicators from memory, neither can they.

Influenceable by the person carrying them. The test is simple: ask the person “what do you do tomorrow to move this number?” If they don’t have a concrete answer, the indicator is the wrong one. Not the person.

Reviewed often. Weekly, not quarterly. An indicator you see in time is a steering wheel. One you see too late is an incident report.

Tied to a real conversation, not just to a bonus. The number is the excuse for the discussion, not its purpose. The question isn’t “why didn’t you hit the target?” but “where are you stuck and what do you need to get past it?” When the indicator becomes the person’s own tool instead of the boss’s whip, behavior changes. Not before.

How to pick your two or three indicators, practically

Don’t start from the list of indicators. Start from the result you want and work backwards.

Ask yourself three things, in this order.

What result do I want, in a single sentence? Not three. One. If it doesn’t fit into one, it isn’t clear to you yet either.

What do people do, concretely, on the days when that result shows up? This is where all the work is. Look at the good months and the good people and describe what they do differently. Not what they say they do. What they do.

Which of those things can be counted weekly, without bureaucracy? If measuring an indicator requires an extra hour-long report, the indicator will die on its own within two months. Pick what is already visible in the systems you have.

What’s left after those three questions is short. Exactly as it should be.

Where I learned this first

In 2008 I was working at Coca-Cola Bottling Chișinău, on recruitment, development and internal communication. Those were the years of a major production reorganization: from 360 employees down to 220, in a year and a half.

In a context like that you learn fast that a number is never just a number. People read it as a message about themselves. If the number arrives with no explanation and no conversation, they write the explanation themselves — and it’s almost always worse than reality.

I’ve had a rule ever since: no indicator gets launched until I know exactly who holds the conversation about it, and how often.

I took that rule with me — as a business trainer at StarNet and Moldcell, through the seven years of HR, and since 2023 into the Rare People projects. It hasn’t changed.

The link to sales: the 90-day test

I use the same logic in training projects.

I measure before and after, but I read the sales figure at 90 days. In between, I watch behaviors: are the scripts being used, are the checklists being followed, is the follow-up happening, is the objection handled or is a discount thrown in to make it go away faster.

Behaviors move in two or three weeks. The number comes after. If the behaviors haven’t moved, there’s no point waiting for the number — I already know the answer.

Training changes people. The system keeps the change. And a well-built set of indicators is exactly the mechanism that keeps it.

What you can do this week

Open your KPI file. For every row, ask one question: can this person influence this number through what they do tomorrow?

Everything that gets a “no,” move into a results report, where it belongs. What’s left with a “yes” — and there won’t be much — those are your real indicators. Discuss them weekly, with the person, face to face.

If you want us to look together at what you have now, book 30 minutes with me. Free, no strings.

I’ll tell you straight what holds up and what is just a nice-looking spreadsheet.