The Giving Feedback Playbook
Free. No email required.
Giving feedback well means handing someone something they can actually use, not something they never asked for and can't do anything with. Most feedback doesn't work that way in practice: it's too vague to act on, or it turns up months after the moment it was about, once there's a rating attached to it.
Receiving feedback is its own playbook: see The Receiving Feedback Playbook. So is asking for it: The Soliciting Feedback Playbook.
One boundary worth naming: formal performance ratings and calibration are a different scope. This playbook covers informal, in-the-moment feedback only.
1. The 60-Second Version
Ask permission first, with the context attached. Describe the specific behavior you saw, not your read on it. Explain the impact. Leave the next move to the other person, unless they're new enough to the work that a couple of starting ideas actually help. Positive feedback follows the same shape, shorter: behavior, impact, keep it up. If the conversation goes sideways, or looks like it's about to, jump straight to “When It Goes Sideways” below.
2. The Default Play
Ask permission first, with the context attached: “Hey, there was something in that meeting. Do you mind if I give you some feedback?” Manager Tools is the primary lineage this playbook draws on for giving feedback, and asking first is its opening move. A yes means go ahead. A no means “OK, no problem,” and letting it go for now rather than forcing it.
From there, Manager Tools' sequence continues: describe the specific behavior you observed, not your interpretation of it (no “you seemed distracted,” only what you actually saw or heard). Explain the impact it had, then close by asking what the person wants to do differently rather than prescribing it yourself. Leadership Explored, episode 7, walks through this same sequence. The Center for Creative Leadership's Situation-Behavior-Impact model covers the same ground and starts by naming the situation itself, a specific moment or context, not something as vague as “last week.” From there it adds one more optional move beyond Manager Tools: once impact is on the table, asking about intent (“What were you hoping to accomplish?”). Handle it carefully. It can tip into the other person explaining or arguing their case instead of hearing what was said. If that starts to happen, redirect: “Let's come back to what happened first, then we can talk about why.” It should never sound like “why did you do that,” which reads as an accusation rather than a question. The same shape works for positive feedback, just shorter: behavior and impact, then “keep it up” instead of an open question.
Timeliness matters as much as structure. Gallup's 2021 research, based on more than 13,000 employees, found that people who report meaningful feedback in the past week are far more likely to be fully engaged at work, and people who get feedback daily are 3.6 times more likely to be motivated toward outstanding work than people who only get it once a year.
Calibrate for experience. Someone new to a domain may not yet know what to do differently, so offering a couple of starting ideas can help (Leadership Explored, episode 7). Someone more experienced usually does better with the room left fully open.
All of this assumes feedback moving down or across, to a report or a peer. Going up is different enough to call out on its own. The rule: read the culture, then read the relationship. If you're new, or an outsider to the group, the culture check is most of what you have — don't give corrective feedback up unless you're genuinely sure the culture actually supports it, and Amy Edmondson's psychological safety research backs up why that check matters. With a strong, established relationship, the relationship read counts for as much: research on the specific manager-report relationship (Leader-Member Exchange) finds it's often the sharper predictor than the org's ambient culture. The two reads can disagree in both directions — sometimes the culture says no and the relationship says yes, sometimes the culture says yes and the relationship says no — and when they conflict, go with the read you have real evidence for. When it does check out, the shape stays the same: ask permission with the specific context attached, something like “Do you have a couple minutes? There's something from [specific, recent situation] I'd like to share, if that's alright.” Use the same SBI structure as anywhere else in this playbook, then close with an explicit opt-out: “Would it be useful if I flagged this kind of thing, or would you rather I not?”
Going down carries its own version of this, in the opposite direction: the same words carry more weight coming from the seat of power, not less, and the giver has to actively correct for it rather than assume good intent will read through on its own. Practitioner writing calls the pattern executive amplification, an offhand remark from someone with real authority hardening into a standing rule nobody meant to set; the same mood-and-affect spread already named for going up runs downward too (Sy, Côté & Saavedra, 2005). My own rule for keeping that from happening: be more careful, more specific, and lean harder on the SBI discipline than usual, because feedback from a boss can't be generalized or read as a comment on someone's personality. Say the quiet part out loud, since tone alone won't reliably carry it: some of what you're saying really is just feedback, take it or leave it; some of it is a performance conversation, or closer still to an order, something you actually expect them to follow, and the receiver has no way to tell which unless you say so. How bluntly any of it should land in the first place moves with the same variable named above, the relationship and the team's own hierarchy norms, doing double duty now for tone as well as for whether to speak up at all.
Positive feedback up is safer by default, with one boundary that peer-reviewed research on workplace flattery keeps finding: generic, unearned praise toward leaders measurably backfires, breeding overconfidence in the person receiving it and resentment in the person giving it. The standard that keeps it safe is the same one already in this playbook: specific and tied to an actual behavior, the same shape as the SBI example above, aimed upward. “In [specific meeting], you [specific behavior]. That's what got us to [outcome]. That was really useful, thank you.” Positive feedback up is always fine, as long as it stays that specific.
3. Where the Sources Disagree
Not every model behind this playbook agrees, and the disagreements are worth knowing rather than smoothing over.
Ask permission, or don't. Manager Tools treats the permission question as load-bearing, the gate that makes someone ready to actually hear what comes next. Radical Candor doesn't use an equivalent gate: once real care for the person is established, the ceremony can become a hedge rather than a courtesy. Both are defensible. This playbook defaults to asking, because it costs little and protects the moments when someone genuinely isn't ready.
Ask permission going up, or skip the ceremony there too. The same disagreement above has an upward-specific version, and the stakes aren't symmetric. Radical Candor's remove-the-ceremony logic carries upward the same way it carries down. Crucial Conversations pushes the opposite direction here: keep full ceremony, explicit permission and naming the shared goal, whenever a real power gap exists, because the person with less power has more to lose from guessing wrong. This playbook's default for downward feedback is to ask; upward gets the same default, and the reason matters even more the higher the stakes climb.
Ask about intent, or don't. CCL's model adds a step Manager Tools doesn't have: asking what the person intended, once impact is already on the table. It closes a real gap in the plain Manager Tools sequence, but the same question can also invite an explanation instead of hearing what was said, which is why the default play above treats it as optional.
A fixed ratio, or not. Radical Candor argues against any fixed ratio of positive to corrective feedback, calling it a formula for insincerity. That can look like it conflicts with leading generously with positive feedback, but it doesn't: a healthy ratio describes a relationship's climate over months, not a script for any single conversation.
How hard to push before someone isn't improving. This playbook's default across the escalation material below is patient and gradual: pause, name the pattern, offer real help, then climb an actual ladder in tiers rather than jump straight to the hardest read available. Radical Candor's own material takes a harder edge on sustained non-improvement, arguing that carrying someone indefinitely in a role they're not succeeding in costs them more than it protects them. This playbook keeps the slower default; the harder edge is worth knowing rather than assuming away.
4. Scripts
Permission ask:
“Hey, there was something in that meeting. Do you mind if I give you some feedback?”
Yes, give it. No, “OK, no problem,” and let it go for now.
An SBI example (composite, for illustration only, never a real exchange):
“In this morning's status update, you cut off two people mid-sentence before they'd finished their point. That meant we moved on without hearing an alternative that might have caught a scheduling conflict earlier. What do you think would help next time?”
If the person is new enough to the work that a starting idea helps, close differently instead: “One thing worth trying next time: agree on a time limit for each person's turn before the meeting starts.”
A positive example (composite, same shape as above, shorter):
“In this morning's status update, you asked a follow-up question that pulled out the detail we actually needed instead of moving on. That's exactly the kind of facilitation that keeps these updates useful. Keep it up.”
If an intent question helps, once impact is on the table: “What were you hoping to accomplish?” Never “why did you do that.”
5. When It Goes Sideways
Sometimes the conversation goes sideways before a second round ever happens: the other person pushes back or gets upset, and the moment stops being about the behavior at all. What happens next can tier or escalate, and how far it goes depends on your actual role.
- Validate without agreeing. When someone pushes back, acknowledging the feeling behind it works better than either agreeing with their conclusion or arguing against it. Marsha Linehan's research on acknowledging emotion without excusing behavior draws exactly that line. Two ways to do it in the room. Name what you don't mean before you name what you do, what Crucial Accountability calls contrasting: “I don't want you to think I'm unhappy with how we work together. Overall, I'm very satisfied. I just want to talk about how we make decisions together.” Or guess at the feeling instead of defending yourself or agreeing outright: “Are you upset because you think I don't respect you?”
- Pause and reschedule. If things are too heated to keep going, say so and set a real time to come back to it, rather than letting it just end. Gottman's flooding research gives the number: twenty minutes minimum before either person can actually hear anything again. Crucial Accountability calls the same move strategic delay: “You know what? I need to think about this in more detail. I'll get back to you later.” A softer version, closer to how it might actually come out: “I'm too fired up to keep going right now. Can we pick this up tomorrow morning, same time?”
- Name the pattern. If this is the third time you've had to raise the same thing, say that directly, as its own topic, not folded into today's instance: “Look, we've tried to have this conversation three times now, and it seems like you're not willing to take feedback on this. I want to talk about that pattern, not just what happened today.” Crucial Accountability's CPR model gives this shape a name. Content is today's specific instance. Pattern is the fact that it keeps happening. Relationship is what the pattern is starting to cost between the two of you. Keep a pattern conversation as its own conversation; folding it into a content conversation buries the one thing that actually needs saying.
- Know which ladder you're on. Where this goes from here depends on your actual role, and there are really two separate ladders. Tier one is available to anyone, peer or manager: document what happened; restate the standard clearly; say what the consequence will be if it continues; address the very next infraction immediately rather than letting it slide; then start down whatever path of discipline is actually yours to start. Tier two only opens up with real authority over the person, the formal version the Society for Human Resource Management lays out: a verbal warning, then written, then a final warning or performance improvement plan, then termination. Knowing which ladder you're standing on before you're on it is most of what makes escalation feel like something you said you'd do, not a threat you're making up on the spot.
6. The Follow-Through Loop
One conversation rarely finishes the job. Giving the same feedback more than once is normal, and it doesn't need to escalate into a performance conversation by default.
- If you notice yourself repeating the same feedback, check your own delivery first: was it specific and clear, and did you actually set the expectation?
- Then figure out what's actually happening: real resistance, or a genuine misunderstanding of what you were asking for. They call for different responses, and the way to tell them apart is to ask rather than guess. Stay on the specific behavior: “I noticed you interrupted people while they were still talking. Can I help you come up with a strategy or approach to change this?” What looks like resistance on the surface often has a plainer explanation: low confidence, burnout, unclear expectations, or friction in how the work itself is set up.
- If they seem receptive, give the feedback again and give them another chance. If the same thing keeps recurring anyway, say so directly and offer to help: “I've noticed [behavior] is still happening. Is there something making this hard, like time or how the work's set up, or is this more that you're not sure it's worth doing? Either way, can I help you come up with an actual approach for it?”
- Write the agreed next step down, the same day if you can, while it's fresh for both of you.
- Set a concrete date for the next checkpoint, not “let's revisit this.” At the checkpoint, look at the behavior since the conversation, not the original incident again. A large meta-analysis of feedback research, covering more than 600 effect sizes across 131 studies, found that feedback improves performance on average, but more than a third of the time it actually made performance worse, and the throughline in the interventions that backfired was attention shifting onto the person instead of staying on the work (Kluger & DeNisi, 1996). Re-litigating the original moment is exactly that shift.
- Tie the checkpoint to a rhythm you already have, like a standing 1:1, rather than scheduling a separate meeting just for this.
- If none of this moves the needle after several rounds, there's a general threshold worth knowing, and the number isn't mine: practitioner guidance puts it at roughly four weeks of no real change after feedback has been clear and repeated, backed by real documentation and genuine coaching along the way. What makes that point defensible instead of a surprise is exactly that trail: expectations stated plainly; more than one specific conversation; help actually offered; the gaps written down as they happened. This is also the moment the pattern has started costing trust, not just recurring, worth naming directly rather than letting it pass as one more repeat. See “When It Goes Sideways” above for what that escalation actually looks like in tiers.
7. Failure Modes and Fixes
Feedback gets hoarded until the annual review. By the time it surfaces, the moment it applied to is long gone, and there's usually a rating attached now, raising the stakes right when there's the least room left to act on it. Fix: say it within days. If it's not urgent enough to say this week, ask whether it's worth saying at all.
Feedback arrives secondhand, relayed from one source, without ever checking the receiver's side. Someone passes along what one other person said, and it never gets checked against anyone else's read or the receiver's own account. Fix: triangulate before relaying it, or send it back to the person who actually needs to say it themselves, directly.
Performative empathy. The giver opens with how hard this is for them and how uncomfortable it makes them, and the moment becomes about managing the giver's feelings instead of delivering something usable. Fix: keep the framing on the receiver and the behavior.
Giver-centered feedback. The content is really about what the giver prefers, dressed up as a note on the receiver's growth. Fix: before delivering it, ask what the receiver gets out of acting on it. If the honest answer is “I'd prefer it,” it may not be feedback at all.
All correction, or all compliment. A steady diet of only what's wrong wears a person down. A steady diet of “great job” with nothing underneath teaches nothing. Fix: lead generously with specific positive feedback (the same discipline corrective feedback requires) and attach both to an actual behavior worth naming.
The suggestion that becomes a mandate. A leader's casual, unprioritized feedback gets treated as a standing directive, because nobody downstream can tell “a thought” from “an instruction” once it's left your mouth. Fix: name the register every time rather than trust people to guess it right: “This is a thought, not a directive — if I ever mean it as a directive, I'll say that plainly. Take or leave this one.” Over-labeling costs next to nothing; a misread suggestion costs a standing process nobody asked for.
Point an agent at this section to help you prepare to give this feedback: describe the situation, and ask it to draft an SBI opening in your own words.
8. AI-Era
This sits somewhere in between: further along than a pure experiment, not yet a settled recommendation.
Rehearse the hard ones. Before a difficult conversation, describe it to an AI and work through the likely reactions ahead of time (Leadership Explored, episode 8). It won't remove the reaction, but it can help you get past your own automatic response and consider the other person's side before you're in the room.
9. This Is a Fit Question, Not a Recipe
The shape above is a strong default across all three directions. Where a specific team or leader actually sits against it, and what that's costing, is what the Engineering Ways of Working Diagnostic is built to find.