What a good brief looks like

We have run one hundred and six projects through this studio. I went back through the eleven that lost money and looked for the moment things turned. In nine of them, the moment was before the kickoff call. The brief did not contain the information the project needed, we started anyway because the client was pleasant and the work looked fun, and the missing information came back as scope in week seven.

A brief is not a wish list and it is not a specification. It is the smallest document that makes a decision possible, and it has to answer four questions.

1. Who decides

Not who is the client. Who, by name, can look at a page and say “yes, ship it”, without asking anyone else.

• If the answer is a committee, the project does not have a decider, it has a queue, and every review round is the length of the slowest calendar in the group.
• If the answer is a name but that person has never been in a meeting with us, the answer is fictional and will be discovered in week five.
• If the answer is “we will decide together”, that is a lovely sentiment and it means the first genuine disagreement has no resolution procedure.

The fix is small and it works: one named decider, an agreed review cadence, and a written rule that feedback arrives consolidated from that person rather than as five separate emails that contradict each other. We put it in the brief in one sentence, and we read it out loud at kickoff so nobody can be surprised later.

2. What already exists

Half of every estimate is really an inventory question. A brief that says “redesign the site” is describing an unknown quantity of work; a brief with an inventory is describing a project.

Thing	Who provides it	Needed by	If it is late
Final copy, all pages	Client	Week 3	Launch moves, day for day
Photography and logo files	Client	Week 2	We use placeholders, one revision covered
Brand guidelines, if any	Client	Week 1	We propose, client approves in week 2
Analytics and DNS access	Client IT	Week 8	Launch moves
Design, build, deployment	Studio	Per plan	We eat it

That last column is the one that changes behaviour. Not as a threat: it is written in the same sentence as our own penalty, which is that anything on our line is our problem and our cost. People accept a rule that visibly binds both sides, and a schedule that only slips in one direction is not a schedule, it is a hope.

3. What counts as done

“Done” is the most expensive undefined word in this industry. Ours is a list a client can check without us in the room:

1. Every page in the agreed sitemap exists, with final content, in both languages.
2. The site meets the performance budget on the agreed test device.
3. Keyboard navigation works end to end, and the automated accessibility pass is clean on all templates.
4. Analytics, redirects from the old URLs, and the sitemap are live and verified.
5. The client’s team has done the CMS walkthrough and edited one real page unsupervised.

Point five is worth more than the other four. A site that the client cannot edit is a site that dies in eighteen months, and the walkthrough is where you find out that the person who will actually maintain it was never in any of the meetings.

A deliverable that only the studio can verify is not a deliverable. It is a promise with extra steps.

4. What happens when we disagree

Every project has one. Usually it is the homepage, usually in week six, usually because two people wanted different things and both heard yes.

We write the procedure in advance, in three lines: the decider decides; if the decision changes agreed scope, we quote the change before doing it; if the change is refused, the previously agreed version ships. That is it. It has never once felt heavy in practice, and the projects that had it were the ones where nobody had to raise their voice.

The three questions that reveal a budget

Clients often will not name a number, and the reason is reasonable: they think it becomes the price. So we ask three questions that produce the same information without anyone losing face.

• “What is the cost of doing nothing for another year?” The answer tells you the size of the problem in the client’s own units, which is the only unit that matters.
• “What else is this money competing with internally?” If it is competing with a hire, you are in the tens of thousands. If it is competing with a conference stand, you are in the low thousands, and you should say so kindly and quickly.
• “What did the last version cost, and what went wrong with it?” You learn the order of magnitude and the trauma at the same time, and the second half is usually more useful than the first.

Three answers, no number named, and you know within a factor of two whether the conversation is worth continuing. That is a kindness to both parties: the worst outcome in a first meeting is not a lost project, it is a project you take on at a number that guarantees resentment by week nine.

Content is the deadline

If I could put one sentence on the first page of every brief it would be this one, because it is the single most reliable predictor of a project running late, and it is almost never the thing anyone worries about at the start: the design will be finished long before the words are, the build will be finished before the words are, the client will genuinely intend to write the words, and then the person who was going to write them will be pulled onto something urgent in another department, and the site will sit at ninety five percent complete for six weeks while everyone waits for a page of copy that nobody has been given time to write, and at the end of those six weeks somebody will suggest launching with placeholder text, and that is how sites end up shipping with a headline that says “About us” over three sentences from the 2019 brochure. Put content dates in the brief, give them owners, and treat a missed copy deadline exactly like a missed technical one.

The one page we send back

After a first call we send a single page: the four answers as we understood them, the inventory table, the definition of done, and a price range with what moves it up or down. It takes forty minutes to write and it has never failed to be useful. Either the client corrects it, which is the cheapest correction the project will ever get, or they sign it, and we begin a project that already knows what it is. The GOV.UK Service Manual says something similar at considerably greater length, and it has been right for a decade.