A kickoff has a job that isn't on the agenda. The client needs to leave feeling heard, the agency needs to leave feeling trusted, and for about an hour both sides work quite hard at giving each other that.

You can feel the moment it turns. Someone says the thing they came to say, someone else nods at it, and the meeting shifts from working out what we're making to confirming that we agree. Notes get shorter. People say "yeah, exactly." Someone uses the word aligned, and then someone else uses it, because that's how that word works.

Three weeks later the work goes up, the client says it isn't quite what they had in mind, and nobody in the chain has done anything wrong. The brief just never said what any of us thought it said.

This comes up when you're training people, and it's the clearest way to see it.

Ask five people at the same company for their elevator pitch. A founder, someone in sales, someone who's been there three months. They come back describing five different companies. Every version is true — each person has built theirs out of the part of the business they actually touch, and most of them have never had to say it out loud next to the other four.

Ask those same five what they want from a website and the same thing happens. The difference is that this time it becomes a document. Somebody has to write the brief, so somebody takes five pictures and flattens them into a paragraph.

Which is where premium but accessible comes from. Nobody chose it. Premium is the founder, thinking about margin and the brands they'd like to be shelved beside. Accessible is whoever owns the numbers, thinking about the buyer they'd lose. Both read it and see themselves in it. Neither has ever had to say their half out loud with the other one in the room.

The family is large. Modern but timeless. Simple but distinctive. Confident but not arrogant. Two people's positions stapled together, and "but" is the join.

The biggest study on this, BetterBriefs, asked 1,700-odd marketers and agency staff across 70 countries. Eighty per cent of marketers think they write good briefs. Ten per cent of agencies agree.

Agencies enjoy that number too much. The person who wrote the brief was solving a different problem to the one we imagine. They got twenty minutes with the founder, in a diary that had none. Legal took two things out. Someone senior added a line late that contradicts paragraph two and can't be removed. And somewhere behind them is an agency that found an ambiguity in a brief and used it as cover in week six, which teaches you to write documents that can't be turned against you.

Our half is simpler. The brief lands, someone reads it, and the tension is visible inside a minute. We say nothing, because raising it means opening a new relationship by telling someone their thinking is muddled, and because it's easier to start work than to start that conversation.

So the retro, when it comes, is thorough about the wrong weeks. Timeline, handoffs, who was in the review, whether feedback came in one go or dribbled over four days. All of it week three, week five, week seven.

The answer is usually week one. It's an uncomfortable place to look, because everyone was in the room and everyone felt fine.

Most clients can't tell you what they want in advance, in words, before seeing anything. A CMO will give you positioning, audience, competitive set, margin pressure, all of it sharp. What they can't do is describe in the abstract how the brand should feel, because that knowledge doesn't live in description. It lives in reaction. They know it when they see it, or when they don't.

So you ask them to point at something instead. Neither of these questions is clever. Both are mildly awkward, which is the entire reason they work.

"When you say premium, can you show me something that's too premium, and something that's not premium enough?"

Two minutes, and it turns a word into a range with edges. Edges are the only part you can design against.

With five people on the call you won't get one answer. You'll get five, and two of them won't agree at all. That's the point of asking. The disagreement was always going to turn up. The only question was whether it turned up in a room where nothing has been made yet, or in week four with work on the screen and a deadline behind it.

"These objectives pull against each other. Which one wins if they conflict?"

Harder, and you'll get some version of they all matter equally. Fair enough, because in the abstract they do. So make it concrete: if we could only do two, which goes first? People give things up when they have to choose, and not before. You also find out who's actually deciding, which no brief has ever told anyone.

Both questions make you the difficult one, and that cost is real. You're eleven minutes into a new relationship and you've told someone their homework is confusing.

The alternative costs more. BetterBriefs' follow-up with the IPA and WFA put the average at five rounds to get an idea signed off, up from three in the IPA's 2007 figure. Nobody asked for that. It accumulated one extra review at a time, and the work at the end isn't better for the wait. Five rounds turns up in the account lead's inbox at eleven at night, and in the margin on a fee scoped for three.

Against that, the fix is embarrassingly cheap. No process, no template, nothing to roll out. Two questions and about ninety seconds of discomfort at the start of something that runs four months.

They still don't get asked. A kickoff has its own gravity — everyone has decided in advance that this will go well, and asking what a word means is a strange way to open a relationship you want to keep.

So the room settles, and everyone nods, and it feels like agreement.

It's usually just quiet.