Home → Leadership
How to Give Negative Feedback
You walk into the room. Two minutes of small talk, then "you are doing really well, but...". The moment they hear that "but", they close. Sound familiar?
- The sandwich works once. By the second time the other person has learned the pattern and hears praise as a threat.
- The pattern is simple: event → impact → expectation. What happened, what it caused, what happens next.
- Then stop talking. The most important information usually arrives after that silence.
- Late feedback is not a correction, it is a surprise. Do not wait for the performance cycle.
About that sandwich
The books all gave the same recipe: a compliment, the criticism, another compliment. The logic even looks sound — deliver the message without triggering defensiveness.
Here is what actually happens: people work out the pattern after about two uses. By the third they are not even listening to the opening line; they are waiting for the bad part. That costs you two things, both quietly:
- Genuine praise loses its value. One day you thank someone for real good work and they sit there waiting for the catch. You spent your own currency.
- The real message softens into nothing. The person leaves the room not knowing what to change. You think you spoke, they think they listened, nobody understood anything.
And we have all heard the worst version: "Overall it is going great, there are some small things but nothing serious, keep going." That is not feedback, it is an escape.
What to use instead: event → impact → expectation
Three steps. It looks short and dull, but this is what works in the field.
Say what happened with no interpretation. Whatever a camera would have recorded. "You are careless" is an interpretation; "there were no tests in the PR" is an event. Start with interpretation and the conversation moves from behaviour to character, and there is no coming back from there.
Explain what it caused. Not abstractly: who was affected, how much time went, what was delayed? Most people genuinely do not know the consequences of their behaviour — nobody ever told them.
Say what you want from now on in one sentence. "Be more careful" is a wish, not an expectation. "From now on do not open a PR without tests" is an expectation; you can measure it tomorrow.
The same situation, two versions
"You did some really good work this sprint, your energy in the team is great. One small thing, the tests seem a bit thin, but that happens. Overall I am very happy, keep it up."
Result: the person leaves thinking you are happy. Nothing changes.
"The last two PRs had no unit tests. One of them caused a production issue on Friday evening; Cem and I spent three hours on it on Saturday. From now on I want you not to open a PR without tests. If something is blocking you there, let us talk about it separately — what happened?"
Result: what happened, what it cost, and what is expected are all clear.
And then be quiet
This is the hard part. After you say your piece there is a silence, and that silence is uncomfortable; you want to fill it, soften it, add "it is not a big deal really". Do not.
Because the real information comes after that silence. In the example above the actual answer was: "I was writing tests, but the test environment has been broken for two weeks and I cannot run them. So I dropped it."
So there is no diligence problem here; there is an infrastructure problem and nobody said anything. With the sandwich you would never have learned that, because the other side would not have realised a serious topic was on the table.
Timing: why immediately?
The most common mistake is stockpiling feedback for the performance cycle. The logic even seems reasonable: "let me not knock their morale now, we will discuss it at the review anyway."
But in the three months in between, that person repeated the same behaviour maybe forty times and learned it was fine, because nobody objected. What they hear at the review is not a correction but a betrayal: "if it bothered you for three months, why did you not say so?" And they are right.
Nothing in a performance review should be heard for the first time. Everything discussed there should have been said at least once already, while the event was fresh. That one rule carries about half the trust in a manager relationship.
Three hard cases and what I actually do
"But it happened in front of the team"
Criticise in private, praise in public. The one exception: if wrong information is spreading to the team, correct it there — but correct the information, not the person. "One correction: that service has no retry, so let us not build it that way" and move on. The part about the person comes later, privately.
"I said it, they did not accept it"
Usually one of two reasons: the event part was not concrete enough (it landed as an opinion), or there really is a reason you do not know. One question separates them: "How do you see it?" Listen to the answer. If they are right, saying "you are right, I did not have the full picture" does not reduce your authority. It does the opposite.
"This is the third time I have said this"
If you are giving the same feedback a third time, you are past the feedback stage. At that point the conversation changes, and you have to say so explicitly:
"This is the third time we are discussing this, so this conversation will be different. Over the next two sprints we will track this: [concrete measure]. If we are still here after two sprints, we will make a bigger decision together. If something is blocking you, tell me now and let us fix it."
That is blunt. But leaving it vague is blunter: the person carries on for months not knowing how serious it is, and then one day meets a decision they never saw coming. Clarity hurts less than the bad news itself.
A small habit: measure yourself
For a year I roughly logged the feedback I gave. Two things showed up, and both annoyed me:
- Most negative feedback went to the same two people. Not because the others worked badly, but because I talked to those two more.
- Most positive feedback went to visible work. Whoever ran the demo got praised; whoever kept the infrastructure alive went unnoticed.
Neither is about intent; both are about attention. When you do not measure it you think you are being fair. When you measure it you find out you are not.
Checklist
- Is what I am about to describe an event, or my interpretation?
- Can I state the impact concretely (who, how much, what was delayed)?
- Is my expectation measurable, or is it "be more careful"?
- Is this the first time I am saying it, or the third?
- How long ago did it happen? (More than a week means I am late.)
- Could there be a reason I do not know about? Will I ask?
- Will I manage to stay quiet after saying it?
Conclusion
Negative feedback is hard not because it upsets the other person, but because we cannot sit with our own discomfort. So we soften it, delay it, wrap it in a sandwich — all for our comfort.
In reality, most people are not upset by being spoken to clearly. What upsets them is everyone knowing something is wrong and nobody telling them.