Leadership
Key points

Key points

  1. 1Feedback does not reliably improve performance on its own, and research finds that a sizable share of feedback efforts make performance worse.
  2. 2Feedback changes behavior when it describes a specific, observable action and its impact, without judging the person's character.
  3. 3The missing step in most feedback is a forward plan: an agreed goal, a concrete next behavior and a follow-up date.
  4. 4Frequent, small, private conversations beat saved-up criticism, and specific praise is as important as correction.

Why most feedback changes nothing

Most managers assume that telling someone about a problem will fix it. The research is less encouraging. A review published in the Journal of Evaluation in Clinical Practice and archived on PubMed Central (read on 2026-10-17) summarizes a large meta-analysis by Kluger and DeNisi covering 131 studies. Feedback had a positive effect on average, but more than one third of the feedback interventions were associated with worse performance.

The same review explains why. Under the theory those researchers developed, people change when they notice a gap between what they did and a goal they care about. Feedback that shifts attention toward the person's self-image instead of the task tends to backfire. The review also reports (read on 2026-10-17) that in one analysis of 85 interventions, only a small fraction combined feedback with performance goals and an action plan, even though control theory suggests all three are needed.

In everyday terms, feedback fails for three reasons:

  1. It is vague. "Be more proactive" gives the person nothing to do differently.
  2. It is late. Feedback saved for a quarterly review arrives after the details are forgotten.
  3. It feels like an attack. The person spends the conversation defending themselves instead of listening.

Deep End Strategies names the same three failures as timing, specificity and delivery, and notes that each is fixable with a little structure.

What makes feedback actionable

The PubMed Central review describes a model of "actionable feedback" drawn from studies of high-performing medical centers. Feedback in those settings shared four traits:

TraitWhat it means at work
TimelyGiven close to the event, ideally within days, not saved for a review.
IndividualizedAbout this person's actions, not a general memo to the whole team.
Non-punitiveFramed as information to improve with, not as blame or a threat.
CustomizableThe recipient has input into the content and the way it is delivered.

The University of Waterloo's Centre for Teaching Excellence adds a useful distinction. Feedback is formative: descriptive, constructive and nonjudgmental, meant to help someone improve before the next attempt. Evaluation is summative: it judges an outcome against a standard. Mixing the two is a common reason feedback lands badly. If the person thinks they are being graded, they will defend the grade rather than consider the behavior.

Waterloo's practical checklist for giving feedback:

  • Limit feedback to the most important issues. Too much at once is overwhelming.
  • Concentrate on the behavior, not the person.
  • Be specific and include an example.
  • Be realistic. Only comment on what the person can change.
  • Own the feedback by saying "I" instead of implying everyone agrees.
  • Be timely, and follow up afterward.

Describe the behavior with Situation Behavior Impact

The most widely used structure for the content of feedback is Situation-Behavior-Impact, developed by the Center for Creative Leadership. It has three parts:

  1. Situation. Name the specific moment. "This morning at the 11 am team meeting" works. "Last week" invites confusion.
  2. Behavior. Describe what the person actually did, as a camera would record it. "You interrupted me while I was presenting the budget", not "you were rude".
  3. Impact. Explain the effect on you, the team, the customer or the work. "I lost my place, and we ran out of time before the decision."

CCL reports that its research on SBI found the method reduces anxiety for the person giving feedback and defensiveness in the person receiving it, and that employees rate managers as more effective when they give feedback more often.

CCL extends the model with a fourth step: ask about intent. A question such as "What were you hoping to accomplish there?" turns a one-way message into a two-way conversation. People usually intend to do the right thing. Asking reveals whether the gap is skill, information, priorities or something else entirely. Untools gives a worked example: a manager tells a report that in yesterday's product review they shared no customer learnings, explains that this made the manager worry they had stopped talking to customers, and then asks why.

Two rules keep SBI honest:

  • Remove the interpretation from the behavior. "You didn't care about the deadline" is a guess about motive. "The draft arrived two days after the date we agreed" is a fact.
  • Avoid "always" and "never". CCL warns that these words make people defensive, because there is usually an exception they can point to.
Instead ofSay
"You're not a team player.""In Tuesday's planning session, you rejected three proposals without asking questions. Two people stopped contributing after that."
"Your reports are sloppy.""The Q3 report had four totals that did not match the source data. Finance found them before the board did, but it cost them a day."
"Great job!""The way you summarized the customer calls in one page let us make the pricing decision in ten minutes."

Care personally and challenge directly

Structure handles the words. The relationship decides whether anyone listens. Kim Scott's Radical Candor framework describes feedback on two axes: caring personally and challenging directly. Doing both is radical candor. Caring without challenging is "ruinous empathy", where you stay quiet or sugarcoat. Challenging without caring is "obnoxious aggression". Doing neither is "manipulative insincerity": flattery to someone's face and criticism behind their back. The site stresses that these are behaviors everyone slips into, not labels for people.

Ruinous empathy is the most common failure among well-meaning managers. Radical Candor's guide to helpful feedback makes the point that retreating to abstractions to spare someone's feelings is itself a form of ruinous empathy. Its practical advice:

  • Know how the feedback helps before you give it. If you cannot say why it helps, the other person will not see it either.
  • State your intent. A short preamble such as "I'm telling you this because if I were in your place, I'd want to know" lowers defenses.
  • Show, don't tell. Describe exactly what happened and its effect, for criticism and praise alike.
  • Praise in public, criticize in private.

CCL adds a warning about the feedback sandwich, where criticism is hidden between two compliments. People learn to ignore the first compliment because they know bad news is coming, and they miss the last one because they are still processing the criticism. Be direct and kind at the same time instead.

Turn the conversation into a plan

Here is where most feedback stops and where most behavior change starts. The PubMed Central review notes that for feedback to work as control theory predicts, it must be accompanied by a clear performance goal and an action plan for reaching it. Feedback alone describes the gap. A goal and a plan tell the person how to close it.

After you have described the behavior and listened to their side, move forward:

  1. Agree on the target behavior. "In the next planning meeting, ask one clarifying question before responding to a proposal."
  2. Let them shape it. The actionable feedback model calls for customization. People commit more to a plan they helped write.
  3. Remove obstacles. If the gap is skill, arrange training. If it is information, provide it. Radical Candor notes that connecting someone with the right person often helps more than trying to coach them yourself.
  4. Set a check-in. Name the date you will look at it together.
  5. Notice the change. When the new behavior appears, say so specifically. This closes the loop.

A short script that puts the whole sequence together:

Situation:  "In yesterday's client call..."
Behavior:   "...you answered the pricing question before they finished asking it."
Impact:     "They repeated the question twice, and the call ended without a next step."
Intent:     "What was going on for you there?"
Plan:       "For the next two calls, could you let them finish and summarize
             the question back before answering? Let's compare notes on Friday."

Use praise to grow what is working

Correction is half of the job. Specific praise tells people which behaviors to repeat, and it makes criticism easier to accept because the person knows you notice what they do well.

CCL's research offers two useful data points (as published by the Center for Creative Leadership, read on 2026-10-17). Employees it surveyed said they would prefer a mix of about 65% positive and 35% negative feedback, compared with roughly 77% positive and 23% negative that they actually received. CCL suggests aiming for feedback that is about 75% positive and 25% negative overall. The point is not the exact ratio. It is that people want more honest correction than most managers give, and that praise should still be the larger share.

Praise should follow the same rules as criticism: specific, timely and tied to impact. Deep End Strategies contrasts "Great job this week", which is easy to forget, with "The way you prepared detailed notes before the board meeting made the presentation much smoother", which tells the person exactly what to repeat. Radical Candor adds that a manager is often the only person who can explain the wider effect of someone's work, because the person doing it rarely sees the full picture.

Make feedback a habit instead of an event

Feedback that arrives once a year carries too much weight. Deep End Strategies notes that annual reviews overweight recent events and make each conversation overwhelming. When feedback is frequent and expected, it loses its charge.

Practical habits that make feedback routine:

  • Give it within a few days. Waterloo warns that delayed feedback loses its impact and can breed resentment when the chance to improve has passed.
  • Keep each conversation small. One or two points, not a list of everything saved up over months.
  • Use the one-on-one. A regular private meeting is the natural home for both praise and correction.
  • Ask for feedback yourself. Deep End Strategies suggests ending one-on-ones with "Is there anything I could do differently to better support you?" Managers who receive feedback well model what they ask of others.
  • Write brief notes. A dated line on what was discussed and agreed makes reviews accurate and fair, instead of relying on memory.
MistakeBetter habit
Saving criticism for the reviewGive it within days, in private
Commenting on attitude or characterDescribe the observable behavior
Wrapping criticism in complimentsBe direct and kind at once
Stopping at the problemAgree on a target behavior and a date
Praising in generalitiesName the action and its effect
Never asking for feedbackAsk at the end of every one-on-one

The same discipline applies when your team works with AI agents. Feedback on an agent's output works best in the same form: the specific output, what was wrong with it, and the instruction that should change. Quanta Intelligence's leadership playbooks package these conversation structures as short checklists, but a notebook and the SBI pattern are all you need to begin.

FAQ

FAQ

What is the SBI feedback model?

SBI stands for Situation, Behavior, Impact, a structure developed by the Center for Creative Leadership. You name the specific moment, describe the observable behavior without judgment, and explain its effect. Adding a question about the person's intent turns it into a two-way conversation.

Is the feedback sandwich a good technique?

Usually not. People learn to discount the opening compliment because they expect criticism to follow, and the closing compliment gets lost while they process the bad news. Give praise and correction as separate, specific messages instead.

How soon should I give feedback after an event?

As soon as you can do it privately and calmly, ideally within a few days. Quick feedback lets the person recall the situation accurately and act before the pattern repeats. If you are angry, wait long enough to prepare what you will say, but not so long that the moment is gone.

What if the person disagrees with my feedback?

Listen first, because they may have context you lacked. Ask about their intent and their view of what happened. If you still see the issue differently, restate the specific behavior and its impact, agree on what you will both watch for, and revisit it at a set date.

How do I know if my feedback worked?

Look for the target behavior at the check-in you agreed on. If it appears, name it specifically so the change sticks. If it does not, revisit whether the plan addressed the real cause, such as missing skills, unclear priorities or an obstacle you can remove.

Sources

Sources

  1. https://pmc.ncbi.nlm.nih.gov/articles/PMC3303967/
  2. https://www.deependstrategies.com/the-managers-guide-to-giving-feedback-that-sticks/
  3. https://uwaterloo.ca/centre-for-teaching-excellence/catalogs/tip-sheets/receiving-and-giving-effective-feedback
  4. https://www.ccl.org/articles/leading-effectively-articles/closing-the-gap-between-intent-vs-impact-sbii/
  5. https://untools.co/situation-behavior-impact/
  6. https://www.radicalcandor.com/our-approach
  7. https://www.radicalcandor.com/blog/tips-for-giving-feedback
  8. https://www.ccl.org/articles/leading-effectively-articles/review-time-how-to-give-different-types-of-feedback/
About the author

About the author

Antonio Evans
Antonio Evans

Antonio Evans is the founder of Quanta Intelligence and a fractional CTO who builds HIPAA-ready AI systems for health and life-science companies. Before that he was technical co-founder of FriendsEAT for twenty years, CEO of Socialdraft, and a data engineer at IBM. Antonio lives with his dog Santino in Lisbon, Portugal.

All articles