The DESC feedback model: the script, and when it fits

There's a specific kind of feedback conversation that most models handle badly: the one where you don't just want the other person to understand something — you want them to do something differently, and you've already tried hinting. A colleague keeps rewriting your section of the deck without telling you. Your manager keeps adding scope on Thursday for a Friday deadline. You know exactly what you want to change, and you end up describing the problem politely and hoping the request is implied. It isn't. DESC is the model built for that gap: it's the one with an explicit ask, and a stated consequence, baked into the structure.

What DESC stands for

The model comes out of assertiveness training — Sharon and Gordon Bower's work in the 1970s — and it later became a standard tool in healthcare communication, where the AHRQ's TeamSTEPPS program teaches it for exactly the situations where a nurse needs to raise something with a surgeon and the stakes of not being heard are enormous. That origin explains its shape: it's designed for the person who has to make a request, not for the person handing down an assessment.

1. 1

   D — Describe the behaviour

   State the specific, observable facts. Not "you keep undermining me" but "the last three decks came back with my section rewritten." No interpretation yet — just the thing that happened, in language the other person would recognise.
2. 2

   E — Express the effect

   Say what it does. This is the one step that's explicitly allowed to be personal: "I end up unsure which version is going to the client, and I've stopped putting real work into that section." Owning it as your experience is what keeps it from sounding like a charge.
3. 3

   S — Specify what you want

   The step the other models don't have. Name the concrete change: "I'd like you to leave comments instead of edits, and I'll turn them around same day." Specific, doable, and small enough to agree to on the spot.
4. 4

   C — Consequences

   Say what improves if it changes: "then I can actually own that section and you stop having to redo it." Positive framing is the default — the version that spells out what happens if it doesn't change is a different, much heavier conversation.

The script, in full

Here's the whole thing in one piece, so you can hear how short it actually is:

"The last three decks came back with my section rewritten" (describe). "I end up unsure which version is going to the client, and honestly I've stopped putting real effort into that part" (express). "Could you leave comments instead of editing directly? I'll turn them around the same day" (specify). "Then I can actually own it, and you stop spending your evening redoing someone else's slides" (consequences).

Four sentences. No preamble, no sandwich, no build-up. The compactness is deliberate — the longer you circle before the ask, the more it reads as a complaint looking for somewhere to land.

How it differs from SBI, COIN and STAR

Put simply: SBI, COIN and STAR are structures for delivering an observation. DESC is a structure for making a request. (The whole family in one table, sorted by what each one ends on.) (AID sits between the two — it states a desired outcome rather than asking for one, which is why it works downward and grates upward.) That's why it doesn't really compete with them — it covers a moment they leave uncovered.

The twist: DESC works better upward than downward

This is the part that gets missed, and it's the most useful thing to know about the model.

The "Consequences" step lands completely differently depending on who's saying it. From a peer, or from someone with less positional power, it's just an honest description of what improves — "then I can own that section". From a manager to their direct report, the same sentence structure hears as a warning, because the person saying it controls raises, projects and reviews. "If this keeps happening, then…" isn't information from a boss; it's a consequence in the disciplinary sense, whether you meant it that way or not.

So the model that reads as reasonable and clear between peers reads as a formal escalation coming downward. That inverts the usual advice — most feedback frameworks are taught manager-to-report, and DESC is at its best in the two directions managers find hardest: sideways to a peer, where you have no authority to lean on, and upward to your own manager, which is precisely the situation assertiveness training was invented for. Amy Edmondson's research on speaking up — much of it done in hospitals, the same setting that adopted DESC — keeps finding that people withhold exactly these requests when the power gradient is steep. A script helps, because the hardest part isn't the wording, it's starting.

When to reach for it, and when not to

Use DESC when:

• You already know what you want to change, and the other person can't be expected to infer it
• You're going sideways or upward, where you can't rely on authority to carry the point
• Hinting has already failed once — a specific ask is the natural escalation from an implied one
• The thing is small and concrete enough that a same-day agreement is realistic

Don't use DESC when:

• You don't yet know what you're asking for. If your read is "something's off but I'm not sure what", the Specify step will produce a demand you haven't thought through. Diagnose first — sometimes the recurrence itself means your diagnosis is wrong.
• You're the manager and this is routine coaching. The consequence step over-weights an ordinary note. Use SBI, keep it light, and save the structure for when it genuinely is a formal conversation.
• You're actually delivering a warning. Then say so plainly rather than dressing it as a communication model. Ambiguity there is unkind.
• The relationship is the problem. No script fixes a colleague who doesn't believe you're on their side; the model was never the lever.

Where Mirorly fits

Notice the assumption DESC runs on: that you know what to ask for. That's fine when the problem is someone else's behaviour, and it quietly breaks when the behaviour is yours. Nobody is going to run a clean DESC on their own manager's blind spots very often — the whole reason assertiveness training exists is that people don't say these things upward, and your team is no exception. Which means the requests you most need to hear are the ones least likely to be made. Mirorly is the structural workaround: your team answers specific behavioural questions about how you work, aggregated with an anonymity threshold so honesty is safe, set next to your own answers so you can see where your read is off. The core leadership behaviours template is where to start — and once you know what to change, feedforward is the low-friction way to collect suggestions for changing it.

Common questions

The one-line summary

DESC is the feedback model with a request built into it — describe the behaviour, express the effect, specify the change you want, and name what improves if you get it — which makes it the right tool when hinting has already failed and you know exactly what you're asking for; but the consequence step means it reads as reasonable sideways and upward, and as a warning coming down from a manager, so it belongs in the conversations where you have no authority to lean on rather than the ones where you do.