How to Give Feedback That Actually Changes Behaviour

The feedback sandwich was popularised in the 1980s by management consultants who had, probably, never tried to use it on a peer. The idea — wrap a critical message in two positive ones, so the recipient swallows the bad news more easily — sounds humane and turns out to be ineffective. The recipient learns, within about a year of being on a team, that any positive comment is structural padding for the critical one that's about to follow. The positive comments are discounted. The critical comment gets delivered in a context that reads as manipulation rather than honesty. Trust erodes in both directions.

Ten years after the sandwich's dominance, Kim Scott's Radical Candor (2017) introduced a more sophisticated model that mostly hit its target. She argued feedback should operate on two axes: caring personally about the other person, and challenging them directly about their work. The combination of both is radical candor; the absence of challenge is "ruinous empathy"; the absence of caring is "obnoxious aggression." The model is genuinely useful and has become a minor industry.

The problem is that most attempts to use Radical Candor in real teams produce diluted versions that default to either ruinous empathy (still too polite) or brief flashes of obnoxious aggression dressed up as candor. The theory is right. The execution is hard. Here's what the execution actually looks like when it works.

The Three Specific Elements Good Feedback Contains

Almost all the feedback literature converges on a simple structure when stripped of marketing. Effective feedback has three parts:

1. The specific observation

What specifically did the person do, not a generalisation about who they are. "In the Q3 review, you spoke over Maria three times in the first 20 minutes" — specific behaviour, specific context. "You have a problem with interrupting women" — generalisation, dressed as observation, and actively wrong (it's an extrapolation from the specific).

The rule: before delivering feedback, write down the specific behaviour and the specific context. If you can't, the feedback isn't ready. You're generalising from an impression, and the recipient will legitimately push back on the generalisation.

2. The specific impact

Why the behaviour mattered, concretely. Not "it was unprofessional" or "it wasn't great." Rather: "Maria stopped contributing for the rest of that meeting, which I noticed because she had the best information on the pricing question and we ended up missing it." Concrete consequences, ideally ones the person can observe themselves.

The impact piece does the work of making the feedback worth acting on. Without it, the recipient is hearing criticism of their behaviour with no clear cost — the behaviour might be changed, or might not, depending on their mood. With a specific impact, the behaviour is connected to a real cost they should care about.

3. The specific ask or question

What you want the person to do or think about, stated clearly. Not "think about your communication style" — too vague. Rather: "next time you're in a room with Maria, I'd like you to actively invite her in on questions where she has the domain expertise." Or, if you're not sure what to ask for: "I'd like to hear how you think about this, because I could be missing something."

The question version is often the stronger one for peer feedback — it invites the person into a conversation rather than presenting a verdict. For direct-report feedback, the specific ask is usually stronger, because the point is to change behaviour, not just to explore views.

Why Most Feedback Fails — the Common Errors

1. Delivered too late

The single biggest failure mode. The behaviour happened in March; the feedback is delivered in the annual review in November. The recipient can't recall the specific incident, can't connect the feedback to the context, and experiences it as a vague criticism months after any learning opportunity has passed. The performance management system has converted real-time feedback into nothing.

The fix: feedback within 48 hours of the observation, in the vast majority of cases. If you can't deliver it that fast, the feedback wasn't worth delivering. The 48-hour rule is the single most consequential discipline in how an organisation's feedback culture actually operates.

2. Delivered through intermediaries

"Several people have mentioned that you..." is almost always a worse start than "I noticed that you..." The intermediary framing protects the speaker from direct accountability, diffuses the feedback across an anonymous group, and prevents the recipient from having a specific conversation with the specific observer. It also often isn't true — "several people" is frequently one specific person whose observation is being laundered through the "several."

The discipline: deliver feedback first-person when possible. If multiple people have observed the same thing, one of them should deliver it directly, owning the observation. The intermediary version is bureaucracy, not feedback.

3. Focused on personality instead of behaviour

"You're too aggressive" is personality feedback. It's also unactionable — aggression isn't a behaviour, it's an interpretation. "In the last three planning meetings, you've interrupted peers who were in the middle of their thoughts" is behavioural feedback. Same underlying concern, dramatically more useful.

The rule: if the feedback uses a trait adjective, rewrite it as a verb with a specific object. "You're not a team player" becomes "you've declined three collaboration requests from marketing in the last month, and accepted none." The verb version is discussable; the adjective version is combat.

4. Mixed with unrelated concerns

A common failure: the manager sits down for "a chat about your performance" and raises seven separate concerns. The recipient can't process seven pieces of feedback. They hear the first two, get defensive, and stop listening. The other five are wasted.

The fix: one feedback conversation, one concern. If there are seven concerns, have seven conversations over seven days, not one conversation of 45 minutes. The sequencing respects the recipient's cognitive bandwidth.

The Thing Kim Scott Doesn't Emphasise Enough

The Radical Candor framework assumes the "caring personally" axis is roughly constant. It isn't. Caring personally has a specific prerequisite: time. You can't deliver hard feedback to someone you've met three times and expect it to land. The person has no evidence that you care about their success; your feedback reads as attack, because the trust infrastructure isn't there.

The implication: hard feedback needs a relationship. Not friendship — but demonstrated investment in the person's success over a reasonable period. New managers who try to deliver hard feedback to their teams in week two usually fail, not because the feedback is wrong, but because the relationship context isn't yet established. The feedback reads as authority asserting itself rather than as a trusted party helping.

The move: for new relationships, front-load investment in understanding the person, learning their strengths, backing them publicly. Build the account before you try to withdraw from it. Three to four months of visible investment gives you the standing to deliver hard feedback. Without it, the same words produce very different reactions.

How to Receive Feedback — the Rarely Addressed Other Half

Most feedback advice is about delivery. Almost none is about reception. This is a mistake, because reception is at least as important and harder than delivery. The people who improve fastest over their careers are disproportionately those who are unusually good at receiving feedback — even feedback that's delivered badly, partially wrong, or emotionally charged.

The specific skills of being a good feedback recipient:

• Separating the signal from the delivery. Feedback delivered poorly often contains accurate information. The work is to extract the signal from the static. A reflexive "they're being unfair" response blocks the signal. A better response: "what's the 20% of this that's true, even if the 80% is wrong?"
• Not defending in the moment. The instinct to explain, contextualise, or counter is almost universal and almost always counterproductive. The better response: "thank you, I'll think about that." Then actually think about it.
• Following up specifically. After you've thought about the feedback, come back to the giver. "I thought about what you said. Here's what I'm going to try." This closes the loop and signals that the feedback was taken seriously.
• Seeking feedback from specific sources deliberately. Don't wait for it to arrive. Ask specific people specific questions: "what's one thing I did last week that you'd have done differently?" This shifts you from feedback recipient to feedback initiator, which dramatically increases both the volume and the honesty of feedback you get.

The Organisational Pattern That Matters Most

At organisational scale, the single most leveraged feedback habit is what Andy Grove called "task-relevant maturity" awareness — calibrating the type of feedback to what the person can actually use. A junior hire with low task-relevant maturity needs specific, frequent, sometimes directive feedback. A senior professional with high task-relevant maturity on a given task needs sparse, high-signal feedback that respects their own judgement.

The common failure: giving the same kind of feedback to everyone, regardless of context. Directive feedback to senior professionals reads as micromanagement. Sparse feedback to junior people reads as abandonment. Both cost trust, cost productivity, and cost the specific learning the feedback was supposed to produce.

The fix is deliberate calibration. Before giving feedback, ask: what's this person's task-relevant maturity on this specific area? A junior engineer on a specific coding problem might be at low TRM even if she's senior overall. A senior marketer on a technical product detail might be at low TRM even if she's expert at marketing strategy. The TRM is specific to task, not to person. The feedback style should follow.

Calibrated feedback — delivered early, specifically, with explicit impact, from the right person, in the right mode for the recipient's level — is the kind that actually changes behaviour. It's also rare, because all these elements together require deliberate practice. The feedback most organisations deliver is the degraded version, lacking one or more of the elements. The feedback cultures that genuinely compound, over years, are the ones where this craft is taught, practised, and valued. They are rare. They also outperform, measurably, the organisations that don't have them.