Structure difficult feedback conversations with iCOIN
iCOIN - A framework for feedback conversations
Most of us have a conversation we've been putting off.
Someone on the team is consistently late to meetings. Work quality has been slipping. There was an incident last week that hasn't been addressed.
We know something needs to be said, and we also know, from experience, how these conversations tend to go. Defensiveness. Justifications. A tense silence. And then not much changes.
This isn't usually because the feedback is wrong. It's because high-stakes conversations are genuinely hard to navigate, and almost nobody has ever had the chance to practise them in a structured way before they happen for real. iCOIN gives that structure: a framework that makes these conversations easier to start, harder to derail, and more likely to actually land.
What is iCOIN?
iCOIN is a framework for structuring high-stakes 1:1 feedback conversations.
It stands for Intention, Context, Observation, Impact, Next Steps.
It's adapted from Anna Carroll's COIN model, which itself draws on Nonviolent Communication (Marshall Rosenberg).
The key adaptation is the i: a preparation step that happens before the conversation begins, not during it.
The framework is deliberately simple. It doesn't require a skilled communicator or a trained coach.
It requires slowing down before opening your mouth — which turns out to be most of the work.
When to reach for it
- When a conversation has been postponed because it's not clear how to start it
- When the same issue has come up before and the conversation didn't stick
- When there's a sense that the other person is likely to get defensive or shut down
- When there's a feeling of frustration or charge going into the conversation, without it being entirely clear why
- When the goal is feedback that actually changes something, not feedback that gets acknowledged and forgotten
How to run it
Before the conversation:
- Clarify the intention. Ask honestly: Why am I really having this conversation? If the answer is "to relieve frustration" or "to put this on record," that's worth knowing. The goal of this step is to arrive at an intention that genuinely serves both people, something like: I want this person to succeed, and there's something they need to hear to get there. If that intention isn't reachable yet, the conversation probably isn't ready to happen. This step also functions as a safety valve mid-conversation: if things go sideways, it's possible to return to it explicitly. "Let me come back to why I wanted to have this conversation."
- Name the topic when asking for the meeting. "Can we talk about our meeting with the dev team yesterday morning?" gives the other person a chance to prepare and signals that this is about something specific. Vague requests ("can we have a quick chat?") raise alarm in milliseconds. People spend the waiting time catastrophising, and arrive on the defensive before a word has been said.
During the conversation:
- Context. Set the scene factually. What are we talking about? When and where did it happen? Brief and neutral — orienting, not prosecuting.
- Observation. Describe what was noticed, not what was concluded. "I" language tends to be more helpful than "you" language: "I didn't receive a response to my email" rather than "you never replied to my email." Avoid "always" and "never" — it only takes one counter-example to shift the conversation into a debate about exceptions. One observation, not a list. A catalogue of grievances lands like a prosecution, and the other person stops listening.
- Impact. Explain why it matters, and where possible, frame it in terms the other person is likely to care about. The most effective version isn't "this affected me" but "here's why this matters for your work, your reputation, or the team." The more the impact lands in their world rather than ours, the more likely it is to be heard.
- Next Steps. Invite rather than demand. "I need you to start coming in on time" sounds reasonable, but it skips something important. We don't yet know what's driving the behaviour. The person who's late every morning might have a family member in hospital. They won't say so if the verdict has already been handed down. Ask first: "What do you think we should do about this?" or "What would help?" A concrete proposal can always follow, but the question comes first.
Why it works
Emotion regulation (James Gross, 1998) tells us that intervening early in an emotional response is far more effective than trying to manage it once it's running. By the time we're in the room, it's often too late to regulate well. The intention step is a pre-conversation reset: it shifts our internal frame from threat to purpose before we open our mouths. A dysregulated giver almost always produces a dysregulated receiver.
The Fundamental Attribution Error is a well-documented cognitive bias: when we observe someone else's behaviour, we tend to attribute it to their character. When we observe our own, we tend to attribute it to circumstances. "You never reply to my emails" is a character judgement. "I didn't receive a response to my email" describes a situation. The observation step is designed to keep us on the right side of that distinction, because character judgements produce defensiveness and situational descriptions produce conversation.
Self-determination theory (Deci and Ryan) shows that people are significantly more likely to follow through on solutions they helped generate. When we hand down a verdict ("I need you to start coming in on time"), we get compliance at best. When we ask a genuine question ("What do you think we should do about this?"), we get ownership. The next steps structure is built around that difference.
What to watch for
The vague opener. "Would you be open to some feedback?" or "I wonder if we could have a little chat?" feel considerate, but often they're not. Unspecified high-stakes conversations risk raising threat responses immediately and people end up spending the waiting time catastrophising. Name the topic when asking, don't be afraid to be direct.
Observation drift. The observation step is where it's easiest to slip from facts into judgements without noticing. "I noticed you seemed disengaged in the meeting" is a judgement. "I noticed you didn't contribute during the discussion" is closer to an observation. The difference matters more than it seems.
The laundry list. Arriving with multiple incidents feels thorough. It lands as a prosecution. One well-chosen observation, delivered clearly, has more impact than five.
The reasonable-sounding demand. "I need you to start coming in on time" closes the conversation just before the most important part. Stay curious longer than feels comfortable. The question always comes before the proposal.
Go deeper
- Anna Carroll, The COIN Conversation Model
- Marshall Rosenberg, Nonviolent Communication
- James Gross, The Emerging Field of Emotion Regulation (1998)
- Adam Grant, Stop Serving the Compliment Sandwich
- Edward Deci & Richard Ryan, Self-Determination Theory