Performance Conversations That Don't Damage the Relationship

Most managers have a version of the same worry. If I tell them what is not working, they will take it badly, the atmosphere will change, and I will have made things worse. So the concern gets parked. It reappears months later in a formal review, wrapped in praise on both sides, and the person walks out unsure what they were meant to hear but quite sure they were being judged.

The usual advice is to deliver it more gently. That treats honesty as the risk. The research suggests the risk is somewhere else: in what the conversation is about, and in when it happens. Get those right and an honest performance conversation is one of the things that builds a working relationship, not one of the things that wears it down.

Why do managers think honest feedback will hurt the relationship?

Partly because they misjudge what the other person wants. In a set of five experiments with 1,984 participants, published in 2022 in the Journal of Personality and Social Psychology, Nicole Abi-Esber and colleagues found that people consistently underestimate how much others want to receive constructive feedback. The more someone underestimated it, the less likely they were to give it. The most effective correction was simple perspective-taking: asking givers to imagine they were the person who needed to hear it.

You can see the result in UK workplaces. In the CIPD's Good Work Index 2026, a survey of 5,074 UK employees carried out between December 2025 and February 2026, fewer than six in ten employees (57%) agreed that their manager gives them useful feedback. Fifteen per cent never meet their manager to discuss performance at all, and nearly half of that group (49%) say this is adequate. When silence is that normal, the first honest conversation feels bigger than it should, to both people.

Does feedback actually damage people?

Sometimes, and the way it does so is the useful part. The best-known study on this is Kluger and DeNisi's 1996 meta-analysis in Psychological Bulletin, covering 607 effect sizes and 23,663 observations. Feedback improved performance on average. But over a third of feedback interventions made performance worse, and that was not explained by whether the feedback was positive or negative.

Their explanation is the one every manager should know. Feedback works when it keeps attention on the task. It backfires as attention moves up and away from the task towards the self: from "what do I change in this report" to "what does she think of me". Once that happens, the energy that should go into the work goes into protecting self-image instead.

The CIPD's own evidence review on performance feedback (Cioca and Gifford, 2022) reaches a similar conclusion for practitioners. Feedback focused on the task helps keep attention on task-related thoughts. And for feedback to improve performance, people need to react to it positively, which depends mainly on whether they see it as fair and useful.

Put those together and the relationship risk looks different. It is not "I told them something they did not want to hear". It is "they left the room thinking the conversation was about who they are".

Why do formal reviews make this worse?

Because the annual review is, by design, a conversation about the person. It summarises a year, it usually carries a rating, and it often sits next to decisions about pay or progression. Each of those features pulls attention away from any specific piece of work and towards a judgement of the individual, which is exactly where Kluger and DeNisi found feedback starts to fail.

Gallup's analysis of performance reviews (Sutton and Wigert, 2019) reports that only 14% of employees strongly agree their performance reviews inspire them to improve, and that nearly half say they get feedback from their manager a few times a year or less. Saving everything up for one meeting raises the stakes of that meeting. A small, specific point that would have taken two minutes in March becomes part of a verdict in December.

One caution before anyone rolls out weekly check-ins as the answer. The CIPD review found that making feedback more frequent does not usually make it more effective, and that quality matters much more than amount or timing. Regular conversations help because they keep each one small. They do not help if every one of them is still a judgement.

What is wrong with softening the message?

This is our reading of the evidence rather than a finding in itself, but it follows directly from it. The feedback sandwich, praise then concern then praise, is built to protect the person giving the feedback. For the person receiving it, it does two unhelpful things. It makes the actual point harder to find, so they leave unsure what to change. And it turns the conversation into an evaluation of them as a whole, good bits and bad bits, which is the shift from task to self that the research warns about.

Managers soften because they are anxious about the reaction, not because they think the sandwich works. That is worth taking seriously. The problem to solve is the manager's confidence in the moment, not their choice of bread.

What does a performance conversation that protects the relationship look like?

Take a team lead whose weekly client report has arrived late twice. "You're not very reliable at the moment" is about the person. "The last two Friday reports reached the client on Monday, and they asked about it. What's getting in the way?" is about the work. Same concern, same honesty, very different conversation. Four habits make the difference:

• Name the work, not the person. A specific piece of work, a specific moment, and its effect. No labels and no summaries of character.
• Ask before you conclude. Their view of what happened often changes what the right next step is, and asking signals that this is a conversation, not a sentence being passed.
• Stay in the room for the reaction. The minute after the point lands, when the other person goes quiet, pushes back or gets upset, is where relationships are decided. Most managers either retreat or push harder. Acknowledging the reaction and coming back to the work is the skill.
• End on the next attempt, not the judgement. Agree what will be different next Friday. A conversation that ends with a plan feels like support. One that ends with an assessment feels like a mark.

None of this is new as advice. The gap is that knowing it and doing it calmly, with a real person reacting in front of you, are different things. We have written before about the cost of the conversation that never happened. This is the other half: the conversation that does happen, and whether it leaves the working relationship stronger or weaker.

What should People and learning leaders do differently?

Train the reaction, not the script. Most programmes teach a model for structuring feedback. Very few give managers repeated practice at the part that actually goes wrong: the defensive reply, the silence, the tears, the "that's not fair". That moment is where managers fall back on softening or avoidance, so it is where the practice should go.

Separate the development conversation from the rating moment. If the only time a manager discusses performance is the meeting that sets a rating, every point will be heard as part of a verdict. Encourage short, specific conversations about the work during the year, and keep them visibly apart from pay and grading decisions.

Measure whether the conversations are useful, not just whether they happened. A completion rate on check-ins tells you very little. The CIPD's survey question is a better test: do people agree their manager gives them useful feedback? Add it to your own pulse survey and track it by team. It is an early warning that arrives while you can still act.

Use perspective-taking. The simplest intervention in the Abi-Esber studies was asking givers to imagine being on the receiving end. It costs nothing to build into manager briefings: before you decide not to say it, ask whether you would want to know.

Where does practice fit in?

This is the gap HumanVantage was built for. Managers rehearse the performance conversations they tend to soften or put off in a private video role-play with an AI colleague who reacts like a real person would, including getting defensive or going quiet. They get private feedback, try again, and walk into the real conversation having already been through the hard minute once. Nobody is watching, nothing is scored for anyone else, and nothing is reported upwards; the point is the manager's confidence and the team's sense that it is safe to talk honestly. As one pilot manager put it: "I didn't expect to get as much resistance - they were quite sassy!" In our pilots of over 100 managers across several organisations, 82% said they would use it before a real one-to-one.

Honest performance conversations do not damage relationships. Verdicts do, and so does a year of silence followed by one. The fix is not gentler words. It is smaller, more specific conversations about the work, and managers who have practised what to do when the other person reacts.