The five steps of a hard conversation, and the two that usually go wrong

There are more models for difficult conversations than there are difficult conversations. SBI, COIN, DESC, NVC, SPIKES if you work in a hospital, half a dozen more with a book each. Lined up next to each other they look like a crowded market. Read them properly and they are mostly the same advice wearing different hats.

Underneath, nearly all of them say this:

Prepare the goal and the facts. Open with clear intent. Listen more than you talk. Respond to the emotion without losing the point. Close with agreed next steps.

Steady uses those five steps, with plain names you can remember under pressure: prepare, open, listen, stay on point, agree. Most conversations go wrong at the start and the end, so those two get the most attention below.

Prepare: know three things, and no more

Preparation is where most people either do nothing or do the wrong thing. The wrong thing is writing the script. A scripted conversation survives exactly as long as it takes the other person to say something you did not plan for, which is usually their first sentence.

Three things is enough, and all three have to be specific:

1. What happened, exactly. Which deadlines, which meeting, how often, and what the effect was. Not “missing deadlines”. Which ones.
2. What you need to see instead. Something you could watch happening. Not “more focus”.
3. One question to open with. “The last three deadlines moved. What is going on from where you sit?”

Write those on a card. Setting a goal before training is one of the few things the behaviour modelling research keeps finding actually transfers to the job.

Open: the first two sentences do most of the work

This is one of the two steps where conversations go wrong, usually because of politeness. You start with the weather, the project, the holiday, and the real subject arrives eleven minutes in, by which point they have worked out that something is coming and have spent eleven minutes bracing for it.

Say what this is about in your first two sentences. Then:

• Describe what they did, not what they are. “The deck went out at 6pm for a 4pm meeting” is a fact. “You are disorganised” is a verdict, and there is no reply to a verdict except defence.
• Say what you do not mean. “This is not about your ability. It is about the three deadlines.” People fill silence with the worst available interpretation, so take that one off the table early.

The largest meta-analysis of feedback ever run, Kluger and DeNisi’s 607 effects across 23,663 observations, found feedback helped on average and made things worse in more than a third of cases. Their explanation is about attention: feedback that pulls attention onto the person, rather than onto the task, is the kind that backfires. Your opening is where you choose which kind you are giving.

Listen: ask, then stop talking

You have said what you came to say. The urge now is to keep going, to soften it, to explain it a second time in gentler words. Every one of those makes your message smaller.

Ask for their side, and then leave the space alone long enough for them to use it.

There is decent evidence that being listened to properly changes the person doing the talking. In a set of studies by Guy Itzchakov and Avraham Kluger, high-quality listening lowered speakers’ social anxiety and defensiveness, and left them holding their own views less rigidly. Not agreeing with you. Less dug in. In a conversation about someone’s performance, that is the difference between a discussion and a siege.

Then say back what you heard and check you have it right. It costs one sentence and it stops half of all misunderstandings before they start.

Stay on point: stay kind, and keep your point

Nobody plans for this step and everybody meets it. They get upset, or angry, or they go completely flat and agree with everything in a voice that means the opposite.

Three moves, in order:

• Name what you can see. “This is clearly frustrating.” Matthew Lieberman’s lab found that putting a feeling into words reduced the amygdala’s response, though that was people labelling faces in a scanner, not managers in a meeting room, so treat it as a reason to try it rather than proof that it works.
• Do not fold. The most common failure here is not shouting. It is quietly withdrawing the point. “Maybe it is not that big a deal” costs you the entire conversation, and they will remember the retreat, not the point.
• Do not go vague. Watch for yourself getting woollier as they get hotter: more “sort of”, more “just”, more “I mean”. That is the same retreat in slower motion.

Agree: the other step where conversations go wrong

Openings get buried in throat-clearing and endings evaporate into “let’s keep an eye on it”. Both of you leave feeling better and nothing has been agreed.

A close is three things said out loud, by you, in plain words:

• Who does what, by when.
• A clear yes from them, not a vague one. “Sure, yeah, no worries” is not a yes.
• When you will check in.

And one more, which is cheap and appears to matter: say you believe they can do it. In David Yeager’s field experiments, critical feedback that came with high standards and an explicit statement of belief got significantly more people to act on it. That was seventh-grade students revising essays, so do not oversell it. But it is one sentence, and it is true, and the alternative is ending on the criticism.

Follow up: the step that happens afterwards

Acas, whose guide to challenging conversations is the one to read if you are in the UK, is blunt about follow-up, and the training research agrees. If nothing is checked, nothing changes, and your next conversation on the same subject starts from a worse place: you have now taught them that you raise things and then forget them.

Put the check-in in the diary before the conversation ends.

Pick the step you know you skip. For most first-time managers it is the opening, because getting to the point feels unkind. A clear first sentence only feels kind afterwards.