A version of this question comes up in almost every feedback workshop we run. It usually sounds something like this:
“I genuinely try to be kind about it. I start with something positive, I keep my tone light, I say things like ‘tiny thing, not a big deal.’ And it doesn’t matter. The moment people sense feedback coming, they start justifying, over-explaining, or going quiet in that hurt way that’s worse than arguing. One person cried over a comment about meeting notes. I’ve caught myself sitting on feedback for weeks because I dread the reaction, and I know that’s the worst option of all. What am I doing wrong?”
Quite a lot, as it happens, but almost certainly not the things this manager suspects. The instinct to blame the tone, or the timing, or the sensitivity of the team, misses where defensiveness actually comes from.
The short answer: defensiveness is rarely triggered by feedback itself. It is triggered by ambiguity and by anything that sounds like a verdict on the person rather than a description of behaviour. Softening tactics increase ambiguity, so they make the reaction worse, not better. What works is the opposite of cushioning: one specific, observable behaviour, its concrete impact, and a genuine question, delivered privately and soon. And since the first thirty seconds of a feedback conversation are partly physiology, the goal is not to prevent the flinch. It is to not treat the flinch as the outcome.
Let’s look at how this plays out in a real conversation, because the problem is easier to see than to explain.
The autopsy of a feedback conversation
Here is the conversation our letter writer is probably having, reconstructed line by line. Watch what each well-intentioned move actually does.
“Hey, do you have a sec? Nothing serious!” The threat alarm is now ringing. Everyone who has ever worked in an office knows that “nothing serious” predicts the opposite. The person has three seconds to imagine what this is about, and imagination is always worse than information. Ambiguity has entered the room before the feedback has.
“First of all, you’ve been doing really great lately. Honestly. The client work especially.” This is the top slice of the compliment sandwich, and the recipient has seen this bread before. Two things happen simultaneously: they stop hearing the praise, because it is obviously packaging, and they start bracing, because packaging means the real content is coming. The manager has just taught them that compliments are a warning signal, which quietly poisons every future compliment too.
“It’s just, a few people felt the tone of your project update was a bit off.” Three problems in fourteen words. “A few people” is an anonymous crowd the recipient cannot ask, answer, or picture, so their mind supplies the whole team. “Felt” makes it unfalsifiable. And “tone was a bit off” describes nothing observable at all: it is a verdict in search of evidence. There is no behaviour here to fix, only an impression to defend against. So the person does the only thing available: they defend.
“Well, I only wrote it that way because the deadline moved twice and nobody told me until Thursday, and also…” This is the over-explanation the letter complains about, and notice that it is a completely rational response to what was said. When you cannot fix a behaviour, you contest the impression. The manager hears defensiveness. The employee is doing evidence discovery.
“Look, honestly, it’s fine. Forget I said anything. It’s really minor.” And here the conversation completes its damage. The feedback has been retracted, the issue survives intact, and both people have learned a lesson. The employee learned that discomfort makes feedback disappear. The manager learned that giving feedback is painful and pointless. Which is how managers end up, in the letter writer’s words, sitting on feedback for weeks.
Nothing in this conversation was unkind. That is precisely the point: the problem was never kindness. It was that at no moment did the employee learn what happened, what it caused, or what to do about it.
Why softening backfires: the mechanics
There is decent science underneath this. Marcus Buckingham and Ashley Goodall, in their much-cited Harvard Business Review piece The Feedback Fallacy, summarise research showing that critical evaluation triggers a threat response that measurably impairs thinking and learning. In plain terms: the moment a person perceives an attack on their standing, the part of their brain you are trying to reach goes partially offline for a while.
Here is the part most managers get backwards: vagueness intensifies that threat response, because the brain treats ambiguous social evaluation as danger and fills the gaps pessimistically. “The tone was a bit off” gives the threat system nothing to bound it, so it expands to “they think I’m unprofessional.” A specific description (“in Tuesday’s update, the line about the design team read as blaming them”) is smaller, colder, and finite. Specificity is not harsher. Specificity is what makes feedback survivable.
Brené Brown compresses the whole principle into six words: “Clear is kind. Unclear is unkind.”
McKinsey’s learning designers make a related point in their piece on building feedback culture: organisations keep teaching managers about feedback when the missing ingredient is practised capability in real conversations. Knowing the sandwich is bad does not, by itself, produce the sentence that replaces it. So let’s produce it.
The same conversation, rewritten
“Do you have ten minutes this afternoon? I want to talk through one thing from Tuesday’s project update. Nothing about your overall work, one specific thing.” Advance notice, a bounded topic, and an explicit fence around what this is not. The threat system now has edges.
“In the update, the paragraph about the delay says the design team ‘failed to deliver again.’ Two people from design were on that thread.” One observable behaviour. A sentence that was actually written, in a document both people can look at. No crowd, no “tone,” no character.
“My concern is the working relationship: we need design’s goodwill for the next two releases, and that sentence makes their lead less likely to flag problems to us early.” Impact, concrete and forward-looking. Not “it looked bad” but what it costs.
“How did it look from your side?” And then the question, asked as a real question. Maybe the deadline did move twice and nobody told them until Thursday. Now that context arrives as information inside a dialogue instead of ammunition inside a defence.
Notice the conversation got shorter, not longer. Cushioning takes many words. Description takes few.
And when they get defensive anyway?
They sometimes will, because the flinch is physiological and no phrasing removes it entirely. The difference between managers who build feedback cultures and managers who abandon them is what happens in the next sixty seconds.
Do not retract. Retraction teaches that defensiveness works. Instead, slow down and take the time pressure off:
“You don’t have to agree with me in the next ten seconds. Sit with it, and if it looks different to you tomorrow, tell me. The only thing I’m asking is that the next update doesn’t single out a team by name.”
And if the reaction was strong, come back the next day, briefly, without reopening the case:
“Thanks for hearing me out yesterday. For what it’s worth, my view of your work hasn’t moved a millimetre. This was about one sentence, not about you.”
That follow-up costs thirty seconds and does more for psychological safety than any amount of pre-cushioning, because it proves the thing the sandwich only gestures at: that feedback and regard can coexist.
One more thing, taken straight from the pattern in our letter. If everyone on a team reacts defensively, to benign feedback, from multiple people, the explanation is usually not eight coincidentally fragile individuals. It is history: a culture, a previous manager, or sometimes the current one, that taught people mistakes are dangerous. That is worth investigating honestly before concluding the team is the problem, and it is a diagnosis worth sitting with for a moment rather than rushing past.
Frequently asked questions
So is the compliment sandwich always wrong? As a delivery vehicle for criticism, yes, retire it. People decode it instantly, it contaminates your genuine praise, and it buries the one thing they need to hear. Praise generously and often, on its own, when nothing else is attached. That is what keeps the account balanced, not bundling.
Should I give feedback immediately or save it for the 1:1? Soon and private beats scheduled and stale. Small course corrections work best within a day or two, while the situation is still vivid. Save the 1:1 for patterns and development, not for a queue of stored-up corrections; nothing makes people dread a recurring meeting faster.
What if I’ve inherited a team that a previous manager taught to fear feedback? Name it once, openly: “I get the sense feedback here used to mean trouble. From me it means one specific thing I saw and what it caused, nothing more.” Then be boringly consistent for a few months. Trust in feedback is rebuilt by sample size, not speeches.
The behaviour-impact-question structure in this article is one of the core tools managers practise, on their own real feedback situations, in our 90-minute lab Feedback That Lands (And Sticks). Reading about feedback is a start. The reaction-proof version comes from practice.
Photo by RDNE Stock project: https://www.pexels.com/photo/a-man-sitting-in-an-office-7821494/
Photo by RDNE Stock project: https://www.pexels.com/photo/man-in-white-long-sleeve-shirt-sitting-beside-brown-wooden-table-7821499/
