Feedback that lands: simple scripts for hard conversations

Hard feedback doesn't land because you found the perfect model — it lands because you strip out the two things that make people stop listening: vagueness and the feeling of being attacked. A simple script does exactly that: name the situation, describe the observable behaviour (not your verdict on the person), state its impact, then make one concrete forward ask — with enough safety that they can actually hear it. SBI — Situation, Behaviour, Impact (Center for Creative Leadership) — is the workhorse spine; the real value is having ready-made phrasings for the talks that recur in an agency: a slipped deadline, an unhappy client, a coasting senior.

Here's the part most managers get wrong: people want the hard feedback. In a survey of over 2,500 employees, 57% preferred corrective feedback over praise, and 72% said their performance would improve if managers gave it (Zenger & Folkman, Harvard Business Review, 2014). The gap isn't appetite — it's delivery and avoidance. And delivery is not a detail: across 607 studies, feedback lifted performance only modestly on average (d ≈ .41) and over a third of interventions made performance worse (Kluger & DeNisi, Psychological Bulletin, 1996). A script is how you land on the right side of that line.

Why does honest feedback so often fail to land?

Because most feedback fails on delivery, not honesty — and in an agency, the person delivering it usually isn't a trained manager. Your leads are billable practitioners: a senior designer, a principal consultant, an engagement manager who was staffing the work last week. They avoid the hard conversation for lack of the words, so it surfaces late — at review time, as a surprise.

When it does happen, it fails in predictable ways. It's too vague ("be more proactive") so the person can't act on it. It's a verdict on the person ("you're careless") so they defend their identity instead of hearing the point. It's saved up until it's a pile, so it feels like an ambush. And it comes with no forward ask, so nothing changes. Feedback isn't automatically useful — remember that over a third of feedback interventions actually reduced performance (Kluger & DeNisi, Psychological Bulletin, 1996). A script fixes the delivery so honesty gets a fair hearing.

What makes a feedback script actually work? (the anatomy)

A script that lands has four moving parts and one condition. The four parts turn a judgement into something specific and actionable; the condition makes it safe enough to hear.

1. Observation, not judgment. Describe what you saw, not who they are. "The staging build went to the client with the old logo" — not "you're sloppy." Separating the observation from the evaluation is the whole trick; the label triggers defence, the fact invites a conversation (Rosenberg, Nonviolent Communication, 2015).
2. One specific behaviour. Anchor to a single, recent, observable thing — a moment they can picture. Not a pattern, not three examples at once. One behaviour is arguable-with; a character summary isn't.
3. Its impact. Say what the behaviour caused — for the client, the team, the engagement. Impact is what makes the behaviour matter: "the client emailed the partner, and we spent Friday rebuilding trust."
4. A concrete forward ask. End with one clear request for next time, not a complaint about last time. "Before the next client send, ping me for a two-minute check." Feedback only helps when there's a next action.
5. Enough safety to hear it (the condition). People take in feedback only when they don't feel their standing is on the line. Psychological safety — the shared sense that honesty is safe — is what lets the message through rather than the threat response (Edmondson, Administrative Science Quarterly, 1999). Private setting, calm tone, clear intent to help.

How do you use SBI to open a hard conversation? (with wording)

Use SBI as the opening spine: Situation, Behaviour, Impact, then your ask. It's the Center for Creative Leadership's model, and it works because it front-loads facts before any request — so the person hears an account, not an accusation (CCL, SBI model). Say it roughly like this:

1. Situation — pin the when and where: "In yesterday's client review, when you presented the roadmap…"
2. Behaviour — the observable thing, no adjective: "…you answered the two budget questions with 'we'll figure it out.'"
3. Impact — the effect it had: "The client went quiet, and afterwards the partner asked me whether we're in control of scope."
4. Ask (forward) — one request: "Next review, let's pre-agree the budget answers so you can be specific in the room. Can we prep those Thursday?"

If you're not sure why they did it, extend to SBII — add an inquiry into intent: "What was going on for you in that moment?" (CCL, SBII). That turns a one-way verdict into a two-way conversation and often surfaces a cause you'd have misjudged.

What are the ready-made scripts for the conversations you dread?

Here are the recurring agency conversations, each as a short script you can adapt in the moment. They all follow the same spine — situation, observed behaviour, impact, forward ask.

1. The missed deadline. "On the Acme sprint, the build landed Tuesday instead of Friday, and QA lost the weekend buffer. What got in the way — and what would let you flag a slip earlier next time?"
2. The client complaint about a consultant. "The client mentioned the last two status calls ran long and light on decisions. I want your read, and I want us to tighten the next agenda together." Debrief the sting privately first; feed only the durable signal back.
3. The coasting partner-track senior. "Your delivery is solid, but I haven't seen you take a piece beyond what's asked in the last two engagements — mentoring, shaping scope, owning the client. That's the gap between senior and principal here. Which do you want to take on next?" Name the standard, not a character flaw.
4. Defensiveness in the room. "I might have the wrong read — here's what I saw, and I'd like your version." You keep the observation; you hand them the interpretation.
5. Peer conflict on a team. "On the pitch, the two of you kept re-doing each other's slides the night before. It cost us sleep and coherence. Can we agree who owns final say on which sections?" Behaviour and impact, not blame.
6. Real praise (yes, script it too). "In the workshop, the way you reframed the client's problem in one slide unlocked the whole room — that's exactly the senior move we want more of." Specific, behaviour-anchored praise is feedback too.

How do you handle the reaction — defensiveness, "that's not fair", tears?

Stay with the facts and let the emotion have room without softening the message. A reaction is a sign the feedback mattered, not a sign to retreat. Keep three moves ready (Stone & Heen, Thanks for the Feedback, 2014; Patterson et al., Crucial Conversations).

1. "That's not fair." Don't argue the verdict — return to the specific behaviour: "Tell me where my facts are off — I'm describing yesterday's send, not you as a person." If they're right about a fact, correct it; that builds trust, it doesn't cost it.
2. Silence. Let it sit. Then: "Take a minute. I'd rather you think about this than agree quickly." Silence is processing, not refusal.
3. Tears or heat. Acknowledge the person, hold the point: "I can see this is hard, and I still think it's important — take the time you need, and let's pick up the forward plan tomorrow." You don't withdraw the message; you give it room to land.

The one thing not to do is dilute the message to end the discomfort. That's how the actual issue reappears at review time as a surprise.

Which words quietly sabotage feedback — and what to say instead?

Certain phrasings undo a good script before it starts. Swap them out:

1. "You always / you never…" — a global label the person can disprove with one exception, and then they're off the hook. Replace with the single specific instance: "On Tuesday's send…"
2. "You're a bit…" / hedging. Softeners blur the message until the person misses that there was one. Say the behaviour plainly and kindly.
3. "No offence, but…" — signals offence is coming and raises the shield. Drop it; lead with the observation.
4. The compliment sandwich. Burying the point between two positives teaches people to wait for the "but" and discount the praise. Keep praise and correction as separate, honest conversations — disguised critique lands as neither (Buckingham & Goodall, "The Feedback Fallacy," Harvard Business Review, 2019).
5. "Everyone thinks…" — anonymous crowds feel like an ambush. Own it: "Here's what I saw."

How do you make this work across rotating leads and client projects?

This is where scripts stop being a nicety and become infrastructure. In an agency or boutique, feedback is public-facing and cross-cutting: a client complains about a consultant, a lead has to correct someone they'll staff again next month, a partner-track senior is quietly coasting across three engagements. People are rated per engagement by different leads — and a single lead's read is unreliable, since about 62% of rating variance is the idiosyncratic rater rather than the person (Scullen, Mount & Goff, Journal of Applied Psychology, 2000). Shared scripts are how you keep feedback consistent when the reviewers rotate.

1. Give every lead the same spine. When SBI is the house standard, "feedback" means the same thing whoever staffed the work — specific behaviour, impact, forward ask. That consistency is what makes per-engagement notes comparable at calibration.
2. Deliver it in the moment, per engagement. A short script costs two minutes at an engagement debrief. Capture it while the work is fresh, so nothing has to be reconstructed — or sprung — eleven months later.
3. Keep the moment separate from the rating. Run these scripts as developmental, in-flight feedback; keep the grade and pay to the calibrated cycle. Feedback and appraisal aren't the same job, and mixing them makes candor expensive (DeNisi & Murphy, Journal of Applied Psychology, 2017).
4. Debrief client feedback as a team. Client-facing work throws off tough external feedback; separate the signal from the sting together, then deliver the durable part with a script — never as a raw forward of the client's words.

A quick script self-check

Scored four or fewer? Book a call and we'll help you build a house feedback spine your leads will actually use.

FAQ

Is SBI the best feedback model?

It's the most useful default because it's simple and it front-loads facts before the ask, which cuts defensiveness — Situation, Behaviour, Impact, then a forward request (Center for Creative Leadership). The "best model" isn't the point, though; landing is. Any model works if it strips out vagueness and the feeling of attack, and fails if it doesn't. SBI is the spine we'd start every lead on, then extend to SBII (adding an inquiry into intent) when the cause is unclear.

Do people actually want critical feedback?

Yes — more than praise, when it's specific and useful. In a survey of 2,500+ employees, 57% preferred corrective feedback over praise and 72% said it would improve their performance (Zenger & Folkman, Harvard Business Review, 2014). What they don't want is vague or personal criticism. The appetite is there; a script meets it.

Isn't the compliment sandwich a kind way to soften bad news?

It backfires. Burying the point between two positives teaches people to wait for the "but," discount the praise, and miss the message. Keep praise and correction as separate, honest conversations; disguised critique lands as neither (Buckingham & Goodall, "The Feedback Fallacy," Harvard Business Review, 2019).

What do I say when someone reacts badly?

Return to the specific behaviour and let the emotion have room without withdrawing the message. To "that's not fair," say you're describing yesterday's action, not them as a person; to silence, give them a minute; to tears, acknowledge the difficulty and still hold the forward plan (Stone & Heen, Thanks for the Feedback, 2014). Diluting the message to end the discomfort is what makes it reappear later as a surprise.

Why do we need shared scripts if our leads are experienced?

Because your leads rotate across engagements, and a single reviewer's read is unreliable — about 62% of rating variance is the rater, not the person (Scullen, Mount & Goff, Journal of Applied Psychology, 2000). A shared spine makes "feedback" mean the same thing whoever staffed the work, so per-engagement notes are comparable when you calibrate. Experience varies; a house standard doesn't.