How to Give Feedback to Employees That Changes Behavior

Most feedback is received well. The person nods, says that is fair, thanks you for saying it. Then nothing changes, and six weeks later you are having the same conversation with more edge in your voice.

Here is the version almost everyone gives. "Hey, one small thing. It would be good to see you being a bit more proactive on the client stuff. Anyway, just something to think about."

Polite, honest, impossible to act on. No moment, no consequence, no request. Nothing in it says what to do differently on Monday, so it gets filed under things your manager would prefer.

Most advice on how to give feedback to employees stops at be specific and be kind. Both are true and neither is enough. Feedback that changes behavior is mechanical: a window, three parts, and a second half that happens days later. Managers do the first half and skip the rest.

Feedback has a shelf life of about a week

After a week, two things have gone wrong.

The first is memory. You remember the moment clearly because it bothered you. They have sat through thirty meetings since. When you describe it they are not weighing your point, they are reconstructing the event, and the conversation becomes an argument about whether it happened that way. You cannot fix a behavior you are still negotiating the existence of.

The second is what the delay says. If it was worth raising, it was worth raising then. Waiting tells them either that it was not really a problem or that you save things up. And they have now done it four more times with your silence behind it, so you are not correcting an incident. You are correcting a habit you funded.

When you have already missed the window

Do not pretend it is fresh. If the behavior is a pattern, use the old incident as an example rather than the charge, and name the delay out loud.

If it was a one-off that never came back, let it go and catch the next one within a day. A stale incident costs more trust than it buys. If the pattern has been running for months, this is not feedback any more, it is an underperformance conversation.

The three parts, and what breaks when one is missing

Constructive feedback is not a tone. It is a structure with three pieces: something observable, its effect, and a request. Drop one and the failure is predictable.

The observable thing

What you saw or heard, with a time attached. Not a trait, not a summary. "You are not very collaborative" is a judgment about who they are, and the only responses available are to agree or to argue. Most people argue. "In yesterday's planning call you cut Priya off twice before she got to the estimate" is a fact, and facts can be discussed, corrected or accepted. That is the whole difference. Pick one instance, the clearest one.

The effect

Without this, feedback is a preference, and preferences get ranked against every other preference the person is already managing. The effect is what happened next because of the thing. "It came across as dismissive" is thin. "She had flagged a risk to me that morning, never raised it in the room, and we found it in QA on Thursday" is a cost. The ask is now about the project, not your taste in meeting etiquette.

If you cannot name an effect, there may not be one, and what you have is a style difference you were about to deliver as a correction.

The ask

This is the part that gets dropped, and it is why polite feedback evaporates. "Just something to think about" is not an ask. Neither is "be more proactive", which names a quality instead of an action.

An ask is something a person could do next Tuesday and know afterwards whether they did it. Would the two of you agree, a week later, on whether it happened? If not, it is still a mood.

In private, or in the room

The default is private. Not because the feedback is shameful, but because an audience changes what the other person is doing. In front of six colleagues they are not weighing your point, they are managing how they look. You get agreement. You do not get change.

There is a real exception: when the meeting is about to proceed on a wrong basis. A stale number, a commitment nobody can keep, a decision forming on facts you know are out of date. Waiting until the afternoon means the room walks out with the wrong plan.

That version is one sentence, about the work and not the person, and then you hand the floor back.

The behavior underneath it, if there is one, gets handled later and privately. Two conversations, not one: the room gets its correction, the person gets the feedback without seven witnesses.

Checking that it landed, which is not asking whether it makes sense

"Does that make sense?" gets a yes every time. Saying no means admitting you were not following, to the person who writes your review. It is built to be answered one way, so the answer carries no information. Same with "are we good?" and "any questions?"

What works is asking them to say the change back in their own words, then listening for whether the answer is specific.

The first question separates the people who got it from the people who are being agreeable. "I will be more mindful in meetings" is a mood. "I will not respond to a number until the person has finished" is an agreement. Someone who accepts everything in four seconds is ending the conversation, not entering it.

The second question is where the useful information lives, and most managers skip it. Half the time it reveals something you did not know: they interrupt because the meeting always runs long and they have a hard stop, or the draft is late because they are waiting on someone you have never spoken to. Then the ask changes, because it was aimed at the wrong thing.

Written feedback reads harsher than you meant it

Put the same words in a message and they land about two notches heavier. Predictably, every time.

Speech carries what softens it: pace, a pause, your face, the fact that they can answer in the same second. Writing strips that out and adds three things. It is permanent. It can be re-read at midnight, when it sounds worse than it did at four in the afternoon. And it can be forwarded.

It also changes what the thing is. A spoken note is feedback. A written note about behavior reads as a record, and they start wondering who else has seen it.

The line runs between the work and the person. Written feedback on an artifact is fine and often better: comments on a doc, notes on a deck, a review on a pull request. Feedback about how someone behaves belongs in a conversation.

When you have no choice, keep it to one incident, put the ask in the first three lines, and offer a conversation at the end. Then read it back as the person receiving it on a bad day, because that is who will open it.

The follow-up, which is the part nobody does

Feedback given once is an event, and people wait out events. The change happens at the second mention, and that mention almost never comes. You delivered the hard part, felt the relief of it, and moved on. Three weeks later the behavior is back and it is somehow a new topic. Effective feedback is not one conversation, it is two.

The first is the same day, in writing, three lines. Not a warning, not a file for HR. It exists because two people leave the same room with different memories of what was agreed, and the written one survives. "Following up on this morning: on the next planning call you will let people finish before responding. We will check in on the 18th."

The second is at the next instance, and this is the one that makes behavior stick.

Most managers only follow up when it fails, which teaches the team that feedback means trouble. Naming the change out loud costs thirty seconds and is the reason it survives past week two. Your regular one-on-one is where this belongs, with the review date written down so it happens on a schedule rather than when you next get annoyed.

If you reach a third mention and nothing has moved, you no longer have a feedback problem. You have a harder conversation to run.

This does not work if the person cannot see the effect you are describing. Feedback about how someone comes across in meetings is a report on your experience, not a fact they can check, and it lands as an opinion unless you can name the moment and what happened next. When you cannot, say that you are describing an impression, and ask what they saw.

Knowing the three parts is not the same as saying them

None of this is hard on paper. Observable thing, effect, ask, then follow up twice. It fits on an index card.

The difficulty arrives when the person is sitting across from you, they look stung at the word twice, and the sentence you rehearsed comes out with three qualifiers stapled to the front. Feedback fails in the delivery, not in the planning. Saying the ask flat and then staying quiet while they react is a physical skill. It gets easier around the fifth time, and reading does not count as one of the five.