In this guide
- 01Describe an observation
- 02Explain the supported effect
- 03Hear the other account
- 04Agree and revisit action
A useful feedback conversation makes a specific part of the work discussable. Describe what you observed, explain the effect you can reasonably establish, ask for the other person’s account and agree what should happen next. Keep your interpretation open to correction.
Use feedback when you have relevant observations and a legitimate reason to discuss them. If you are passing on an unverified judgement about someone’s character, first establish what actually happened. If the problem is a shared process or missing resource, the conversation should address that context too.
Prepare an observation rather than a verdict
The Center for Creative Leadership’s Situation-Behavior-Impact model distinguishes the situation, observed behaviour and its impact. Its extension adds inquiry into intent. This is a model-owner account of a conversational structure, rather than independent evidence that following it guarantees a good outcome.
Before speaking, write what you directly observed and what you inferred. “The proof was sent without the agreed second check” describes an event to investigate. “You do not care about quality” assigns a motive and a broad personal judgement. The first can be corrected with evidence; the second is likely to move the discussion away from the work.
Check whether your expectation was actually agreed and feasible. A person may have followed a different instruction, lacked access to the checking system or faced a conflicting deadline. These possibilities do not remove responsibility, but they matter when deciding what should change.
Separate observation from interpretation
Consider a hypothetical print studio. A colleague notices that a proof reached production without a particular alignment check. They want to discuss the missed step with the person preparing the job. The correct purpose is to understand the event and make the next handoff reliable, rather than obtain an admission of a personality flaw.

Choose an appropriate setting and explain the purpose. In the fictional conversation, an opening could be: “On this job, I received the proof without the second-check mark we had agreed. I could not tell whether alignment had been checked, so I stopped the setup to clarify it. Can we go through what happened?” This is an invented example, not a quotation from an actual workplace.
The effect should be one you can support. If you know that you paused setup, say that. If you only suspect that the event delayed the whole day, label that uncertainty or check the record first. Do not enlarge an impact to make the feedback sound more consequential.
Invite the other person’s account before deciding on a remedy. Use active listening to check what you understood. A paraphrase should leave room for a different sequence, an overlooked constraint or a misunderstanding of the agreement.
Keep the conversation connected to the task
Kluger and DeNisi’s meta-analysis of feedback interventions found variable effects, including interventions that reduced performance. It challenges the assumption that more feedback is always better. It does not establish the likely effect of this particular workplace conversation.
Focus on the task, its conditions and a useful next action. Avoid turning a specific event into a judgement about the person’s overall worth. Positive feedback also benefits from specificity: identify the contribution and why it helped, rather than offering a vague label that the person cannot use.
Ask whether the conversation is addressing the right level. A repeated checking problem might concern unclear responsibility, a difficult system or a workload conflict. The person may need practice, but training will not fix a missing decision right. Competency mapping can help separate a task capability from an organizational obstacle.
Use a preparation record if the conversation is difficult:
| Field | What to prepare |
|---|---|
| Situation | A recognizable occasion or piece of work |
| Observation | What you directly saw, heard or received |
| Expectation | The relevant agreement or requirement |
| Effect | What happened and what remains uncertain |
| Inquiry | What you need to understand from the other person |
| Possible next step | A practical action to discuss, not impose prematurely |
The table is an original planning aid. It is not a performance assessment scale and does not replace formal procedures where those are required.
Make it possible to correct the account
The CCL listening guidance includes clarification and summarizing. Applied to feedback, that means testing your interpretation before treating it as settled. Ask what you have missed and show that a factual correction can change the conversation.
Edmondson’s research on psychological safety and learning behaviour examines the interpersonal conditions for raising questions and mistakes. It supports attention to that context without proving that one phrasing removes the risk of disagreeing with a manager. Be especially clear about how the conversation relates to evaluation or other decisions.
Do not demand agreement on every interpretation. Two people may agree on the event and still understand its effect differently. Identify the shared facts, the remaining disagreement and what evidence or decision is needed. A forced expression of agreement can leave the practical issue untouched.
If the other person raises a concern about your own behaviour, listen to it as a relevant contribution. It may need a separate conversation, but it should not be dismissed automatically as deflection. Explain how both issues will receive attention.
Agree and revisit a practical next step
In the hypothetical studio, the colleagues might discover that the check happened but its status was recorded in a place the receiving person could not see. The next step would differ from the remedy for an unperformed check. No outcome is asserted here; the example shows why inquiry should precede the agreement.

Agree who will act, what support is needed and how the next relevant piece of work will be reviewed. A practical agreement might change the check location, clarify the receiving role or arrange supervised practice. It should address the mechanism discussed rather than end with an instruction to be more careful.
Use interface management if the problem crosses roles. Use learning transfer if a new task needs practice during actual work. A conversation can initiate those changes, but it cannot substitute for them.
Set a proportionate follow-up. Ask whether the agreement was usable and whether it created another problem. Record what matters to the work and treat personal information appropriately. Avoid turning every informal conversation into a permanent character record.
Common failures and evidence limits
Feedback can fail when it surprises someone with a judgement, mixes several unrelated complaints or invites a response that the speaker has no intention of considering. It can also fail through vagueness: a pleasant conversation about doing better leaves no shared understanding of the next action.
The models and research here do not establish a universal script, optimal frequency or guaranteed effect. The print-studio case and example language are hypothetical. This guide concerns ordinary work conversations, while some situations require a formal management or organizational route.
Before your next conversation, identify one observable event, one supported effect and one question you need answered. Leave with a specific agreement or a clear unresolved issue and its owner. Review the work that follows, rather than treating the delivery of feedback as the result.
Sources:
- Center for Creative Leadership (2025). Use Situation-Behavior-Impact to Understand Intent.
- Avraham N. Kluger and Angelo DeNisi (1996). The effects of feedback interventions on performance.
- Center for Creative Leadership (2026). Active Listening Techniques: Best Practices for Leaders.
- Amy Edmondson (1999). Psychological Safety and Learning Behavior in Work Teams.



