Feedback is a recording.

The gap you can hand over.

The most useful thing I do as a music teacher takes about 30 seconds. I record a student playing a passage, and I play it back.

I don’t tell him it rushed. I don’t tell him he was playing out of tune. I hand him the recording and let him hear the gap between what he meant to do and what actually came out. The tape does the hard part. My only job is to make sure he can’t unhear it.

That is what feedback is supposed to be. A recording you hand someone, clean enough that they can hear themselves. Most of the feedback given in most organizations is nothing like that.

Let me show you what it looks like when it goes wrong. I’m going to use a leader named Marcus, and I want to be straight about who he is: Marcus isn’t a client, and he isn’t one person. He’s a composite, stitched together from a lot of senior leaders I’ve coached who try to give feedback and miss. Every detail is something I’ve seen. He just lets me put it all in one person.

Marcus runs a team of about eighty. He has a director who’s quietly torching morale: talks over people in reviews, rewrites their work without telling them, and takes the credit. Two of Marcus’s strongest people have gone quiet. One is interviewing elsewhere. Marcus has known some version of this for months.

He finally sits the director down. Here is what comes out.

“I want to start by saying how much I value what you bring. The energy, the standards, all of it. There’s maybe one small thing, and it’s really minor, around giving folks a little more room sometimes. But honestly, overall, you’re crushing it. Keep it up.”

The director walks out believing he just got a glowing review. And why wouldn’t he?

Nothing changes.

Of course nothing changes. Marcus didn’t hand him a recording. He handed him a compliment with a whisper folded inside it, then apologized for the whisper.

Clear is kind.

Marcus wasn’t being lazy. He was being kind, or trying to be.

That’s the trap. The compliment, criticism, compliment sandwich, the softening, the “one small thing”: all of it comes from a decent instinct. Don’t hurt the person. Brené Brown put the correction better than anyone.

Clear is kind. Unclear is unkind.

Vague feedback feels gentle in the moment but does real damage over time. You send someone back to work with a warm feeling and no idea what to change. That isn’t kindness. It’s abdication with a smile on it.

Feedback misses for three reasons, and they compound.

Vague. “Be more strategic.” “Show up with more presence.” “Improve your communication.” A verdict with no evidence attached. The person nods, agrees, and has nothing to do differently on Monday, because you never told them what you saw.

Blind. This is the uncomfortable one. Most feedback is built from impression, not observation. You get a feeling about someone, it hardens into a story, and you deliver the story as if it were fact. But you can’t hand over a recording you never made. If you can’t name the specific moment and the specific thing they did, you weren’t watching closely enough to give the feedback yet. You were reacting.

Sandwiched. Compliment, criticism, compliment. I like to call this the shit sandwich. No one wants to eat a shit sandwich. It feels like the kind option. It’s the least kind one on the table, and it fails on its own terms. The real message gets buried in the middle, the softest part of the exchange, exactly where attention is lowest. The person keeps the bread and drops the meat, because that’s what all of us do when someone hands us praise and criticism in the same breath. And over time it poisons your praise: people learn that every compliment from you is the runway for a “…but,” so they start bracing during the good part.

That’s what happened to Marcus’s director. The compliments were load-bearing and the critique was a footnote, so he kept the compliments and threw the footnote away. Anyone would.

Go back and read Marcus’s script again and you’ll find all three, stacked. “A little more room sometimes” is the vagueness. The missing meeting, the missing moment, the missing behavior, that’s the blindness; he was delivering a months-old impression, not a recording. And the compliment on both ends is the shit sandwich. That’s what compounding means. Each miss covers for the others, and the whole thing passes for kindness on the way out the door.

While we’re clearing house, the whole “feedback is a gift” thing goes on the pile too. My industry loves this line. It’s on the slide right after the trust fall, and somewhere, right now, it’s on a mug. I get the appeal: it makes the hardest conversation in management sound like a birthday. But hold the metaphor for ten seconds and it comes apart in your hands. What kind of gift is this? You didn’t ask for it and you can’t return it. There’s no receipt. It arrives wrapped in someone’s good intentions, you’re expected to pay for it on the spot, in gratitude, and if you wince while unwrapping it, congratulations, the problem is now you. Meanwhile the gift itself never has to be any good. It’s the thought that counts, remember? Which is how a whole industry handed managers permission to give thoughtless feedback and feel generous doing it. Nobody needs your thought. Nobody needs your gift. And the warm little glow that comes with handing it over never leaves your side of the table. They need the recording. It’s the only thing in the whole exchange that was ever for them.

Clear is kind isn’t permission to be harsh. It’s permission to be clear. And a clean recording, delivered well, is the kindest thing you can put on the table for someone whose blind spot is costing them.

Vague feedback isn’t kindness. It’s abdication with a smile on it.

What you actually say.

So what does a clean recording sound like? SBI. Situation, Behavior, Impact. The Center for Creative Leadership built it decades ago, and it has stuck around because it works. Three parts. That’s the recording itself. There’s a conversation to build around it, and we’ll get there. But nothing in that conversation works until the recording is clean.

That’s it. Notice what it quietly forces. Specific, because a Situation is a moment, not a pattern. Observed, because a Behavior is something you watched, not something you heard about. And worth raising, because an Impact is a consequence, not a preference. If the behavior didn’t cost anything or create anything, you don’t have feedback yet. You have a pet peeve. The model won’t let you be vague, secondhand, or petty. It only works if you actually looked.

That Behavior line matters more than it seems. The moment you trade the observable thing for a judgment, you hand the other person something to argue with. Tell someone “you were angry in that meeting” and watch what happens: “I wasn’t angry, I was passionate.” Tell them “you were dismissive” and you’ll get “I was trying to keep us on time.” Now you’re fighting about the label, the feedback is dead, and you’ve both lost the thing that actually happened.

But “you pounded your fist on the table” is not up for debate. “You cut Priya off twice” is not up for debate. It happened. The camera saw it. A judgment puts the defenses up. A fact keeps them down, and that is the whole mechanical point of this format: to get under the defenses so the person can actually hear what you’re telling them.

The same discipline applies to Impact, and this is where SBI usually breaks. Impact is what you observed happen, not what you assume they intended, and not a verdict on how everyone in the room felt. There’s a world of difference between “you clearly didn’t care what anyone else thought” and “the room went quiet after you cut in.” The first is a charge, and the moment you file it, the conversation becomes a trial. The second is a recording. Same trap in the other direction: “you made the whole team feel unsafe” is a mind-read they can wave off, but “the two people who’d been talking didn’t say another word” is just what happened. You don’t know their intent, and you can’t feel other people’s feelings for them. Stay on what you saw land, and let them account for the intent themselves.

Before we go further, one objection deserves a hearing, because it’s a fair one. Buckingham and Goodall made it best in their feedback fallacy argument: people are unreliable judges of other people. Your recording, the argument goes, is really a recording of you. Your filters, your history with this person, whatever mood you walked in with. True. And it changes less than you’d think, because I’m not asking you to hand over the truth. I’m asking you to hand over your tape, honestly labeled as yours. One mic in the room, placed where you were standing. That’s why the behavior has to be something a camera would catch, why the impact stays on what you saw land, and why the recording opens a conversation instead of closing one. One honest mic is enough.

Here’s what assigning intent sounds like, and it tends to show up in the conversations that matter most. A VP, furious after a botched launch, to a peer who runs engineering:

Maybe that’s even true. It doesn’t matter, because the other VP will spend the next hour defending his character instead of looking at what happened. Same anger, same stakes, handed over as a recording:

One of those starts a fight. The other starts a fix.

Now set the whole model next to what Marcus said to his director.

Same feedback. One version is impossible to hear. The other is impossible to miss.

Now turn the model around, because this is where most leaders waste it completely. Praise is a recording too, and “great job” is a recording of nothing.

The first is gone by lunch. The second, the person can repeat on purpose next quarter, because you told them exactly what worked and what it did. That is the entire point.

Everything so far has been about the words. The room matters just as much, and for the same reason: a recording only works if the person can actually hear it. Corrective feedback is a closed-door instrument. Always.

Deliver it in front of an audience, and it stops being a recording and becomes a sentence handed down in open court, and the person can’t hear the impact because they’re too busy managing the twelve people watching them take it. Praise in public, correct in private. And even praise sometimes lands better one to one, when public recognition would single someone out or quietly become a comparison everyone else is now running. Read the person and the room.

Clear, then a conversation.

So far I’ve made SBI sound like a speech. It isn’t one. Clear feedback is a conversation, and the recording is what makes that conversation honest.

Think about what I actually do with a student. I don’t hand him the tape and walk out. We listen back together, and then we get to work: the rushing in that one passage, here’s what to try. That’s where the lesson starts. Feedback is the same. The clean SBI is the middle of the exchange, not the whole of it, and there are three moves around it that most leaders skip.

Open it. Don’t ambush. A sentence of frame is enough: “Can I share what I saw in the roadmap review?” Sometimes it’s worth asking first: “What’s your read on how that landed?” A self-aware person will often name the thing themselves, and feedback someone says out loud beats feedback they receive. But this only works as a real question. “How do you think that went?” when you’ve already decided it went badly is a trap, and people feel it. Ask only if you mean it, and hand over your recording either way.

Then hand over the recording. Situation, behavior, impact. Clean. The way we just built it. This is the part you don’t outsource and don’t soften. Whatever they said in the opening, they still deserve to hear exactly what you saw.

Then open it up. This is where “impact, not intent” pays off. You named the impact. Now ask about the intent: “That’s what I saw on my end. What was going on for you?” You might be missing something. So might they. Either way, you’re both looking at the same recording now and talking about it, instead of arguing over whose version of the meeting is the real one.

Then agree on what’s next. SBI ends at impact, which is a diagnosis, and a diagnosis with no treatment is just bad news. The conversation has to move: what changes, and what would better actually look like. Ask before you tell: “What would you do differently next time?” People resist what they’re handed and own what they help build. Be ready to be direct if they can’t see it, but give them the first move.

So the honest shape is open, hand over the recording, ask, agree on the next rep. Clear and a conversation, in that order, because the order is the whole thing. Lead with the conversation, and you’ll talk around the point for twenty minutes. Lead with the recording and the conversation finally has something solid to stand on.

The whole arc in one piece, built around the same feedback for Marcus’s director:

Six moves. Two of them are just questions. The whole thing takes about ninety seconds, and he walks out holding a clear recording and one rep to run.

Nothing at the review should be news.

A recording has a third property, beyond specific and observed. A timestamp.

Feedback has a shelf life. Hand someone the recording while they can still feel the moment and it lands. Save it up for the quarterly review, and you’re playing them a tape of a performance they barely remember giving.

The rule I’d put above almost every other one: nothing in a formal review should be new information.

If the mid-year sit-down is the first time someone hears they’ve been steamrolling their team, that review is the invoice for six months of feedback you never gave. And by then it can’t be a recording, because it arrives with a rating and a raise attached (or not). The person can’t hear a word of it. They’re too busy doing the math on what it costs them.

Feedback is a loop, not an event. The recording gets handed over in the days after the moment, not the months. Small, specific, often. A director who hears “here’s what happened when you cut Priya off,” on the day it happened, six times over a quarter, never needs the big uncomfortable review. There’s nothing saved up to unload, because none of it was saved.

Leaders who dread giving feedback almost always have this backwards. They wait, because the conversation feels big. The waiting is exactly what makes it big. Give the recording early, and each one is small, almost casual, a normal part of how you work together. Stockpile them, and you build a monster you’ll avoid until the review forces your hand, which is the worst possible time and setting to finally press play.

The formal review still has a job. But its job is to summarize a story the person already knows, not to break news. If it’s breaking news, the loop was broken first.

And if the recordings change nothing? Say the director hears the same specific, timestamped tape six times and the behavior holds. Then the recordings have done a different job, just as important: you are no longer missing information, and neither is he. What you have now isn’t a feedback problem. It’s a performance problem, and performance problems get a different instrument. Expectations in writing. Consequences named. A timeline attached. The recordings earn you the right to escalate cleanly, because nobody at that table can claim they didn’t know. The real failure is the leader who keeps handing over recordings forever, a little softer each time, mistaking repetition for management.

Saying it is the skill.

So the model is clean. Three parts, both directions, delivered early and often. If knowing that were the same as doing it, this essay would be over.

It isn’t, and here’s where every SBI article on the internet quits. They teach the three parts and stop, as if naming them were the same as being able to say them out loud when it counts.

Watch the first time a leader tries this for real. They script a beautiful SBI, clean as anything. They walk into the room. They get two sentences in, the other person’s face falls or they push back or they just go quiet, and the leader feels the old pull toward the sandwich and hears themselves say, “but honestly it’s a small thing, you’re doing great overall.”

The sandwich reflex isn’t ignorance of the model. It’s what the model does under pressure.

Knowing SBI is information. Saying it cleanly when the room turns, and every instinct is telling you to soften, is a skill you build.

That capacity has a name, and I’ve written about it before, so I’ll keep it short here. It’s the active observer: the part of you that watches what you’re doing while you’re doing it. Dr. Theo Dawson at Lectica calls it the skill underneath every other skill, and giving feedback is one of the clearest places it either shows up or doesn’t.

Because feedback is a live performance, and every performer has tells. The pull to soften into a sandwich the second their expression changes. The flash of defensiveness when they don’t just accept it. The itch to fill the silence right after you land the impact, usually by walking your own feedback back before they’ve had a chance to sit with it. Each of those is a tell, and each can be caught in the moment once you know it’s yours.

Dr. Jonathan Reams named the failure mode when this capacity is missing. He calls it downward assimilation: you pick up the words for a skill without ever building the thing underneath. You can recite SBI. You can teach it. And you still fold into a sandwich the moment someone’s chin drops. The words are real. The skill isn’t built yet.

Knowing SBI is information. Saying it cleanly when the room turns is a skill you build.

You build it outside the room.

You don’t build this in the room. Not at first. You build it the way my student builds his ear, and the way I built mine on the euphonium: outside the moment first, then closer and closer to it.

Start outside. Before the conversation, script it. Actually write the three parts down. What’s the situation, in one specific sentence. What’s the behavior, stated as something a camera saw and not as your interpretation. What was the impact, on the work and the people, that you observed. Then read it back against one question: could I hand this over and let it speak for itself, or am I still hedging, still reaching for the bread?

That scripting isn’t a crutch. It’s the rep. It’s the thinking you do before and after, away from the heat, and it’s where every in-the-moment skill is born.

Then deliver it, and afterward run a five-minute practice. Two questions, in writing:

• Where did my tell show up today? The reach for the sandwich, the flinch when they pushed back, the rush to fill the silence.
• What did I notice right before it? The heat in your chest, the breath you take to soften, the small drop in your stomach when their face changes, the jittery shot of adrenaline while you wait for their response.

That second question is the whole game. You’re hunting for the signal that fires just before you fold, because that signal is what you’ll eventually learn to catch in real time. And it moves earlier with reps. First you notice the sandwich reflex the next morning. Then on the walk back to your desk. Then one beat after you did it. Then, one day, you feel the pull while it’s still just a pull, and you let it go by without acting on it. By then you’re catching yourself mid-performance, the way a good musician hears the wrong note coming before he plays it.

The worksheet with this essay is built for exactly this. Notice what you actually observed. Script the three parts. Check them against a short quality bar. Then, after the conversation, run the two-question rep. Keep it next to you for the next few conversations you’ve been avoiding.

This is what it looks like when it’s built.

Marcus, months later, same director. He’d been running the rep after every hard exchange and he’d found his tell: a specific tightness, right before he softened, the same feeling he got before smoothing over any conflict. This time he’d scripted it. Situation, the Tuesday planning session. Behavior, cutting off both junior analysts mid-sentence, each as they started to raise a concern. Impact, that by the end only the director was talking and the two people closest to the data had stopped offering it.

He delivered the situation and the behavior clean. Right on cue, the director’s jaw set, and Marcus felt it. The tightness. The pull toward the bread. The sentence forming, the one that would have dissolved the whole thing into “but you’re doing great.”

For the first time, he caught it while it was still just a feeling. He didn’t soften. He let the silence sit, which was the hardest part, and then he said the impact exactly as he’d written it. The director was quiet for a moment. Then: “I didn’t realize I was doing that.” Not defensive. Just true.

He doesn’t catch it every time, even now. But the tightness announces itself earlier than it used to, the sandwich shows up less. And this quarter, the director he’d nearly lost two good people over is the one asking his team what he’s missing.

One recording this week.

So this is where I’d start, if I were you. Not with more reading about feedback. With one recording.

Think of the person you’ve been meaning to say something to. You know the one. You’ve rehearsed it in the car. It’s been sitting in you for weeks, getting heavier, turning into a review you’re dreading instead of a recording you could have handed over a month ago.

A few honest questions before you do:

• When did they last hear something specific from you, good or bad, close enough to the moment to still feel it?
• If your next formal review with them happened tomorrow, would any of it be news?
• Have you ever actually practiced this, or just kept meaning to?

Pick the one conversation. Script the three parts this week, on paper. Deliver it clean, early, before it grows into something that needs a whole meeting. Then spend five minutes afterward finding your tell. That’s the practice. That’s how the formula becomes a skill, and how the skill becomes the way you actually lead.

A clean recording doesn’t ask anyone to feel bad. It just lets them hear what you already heard. Hand it over.