Most website briefs describe how the site should look and say almost nothing about how it should work. Then the site arrives, it looks right, and nobody can tell whether it does the job until the enquiries do not come.
Below is the brief we would hand over. Copy it, send it to whoever is building your site, and keep the right-hand column, because a requirement nobody can check is a requirement that gets quietly dropped. The reasoning behind each line is in the longer piece on redesigning a website.
Use it even if you are not working with us. It is more useful to us as something that gets passed around than as something we keep.
How do you use this?
Three rules, and they matter more than the list.
Put it in the contract, or at least in the email that agrees the work. A brief that lives in a chat thread is not a brief.
Check each line on the live site after launch, not on staging and not in a demo. Most of the failures in this list only happen in production.
Where a line does not apply to your business, delete it and say why. A brief with everything in it is as useless as a brief with nothing in it.
What does every page need?
| Requirement | How you check it |
|---|---|
| One job per page, stated in one sentence | Ask what the single next step is. If the answer has an "and" in it, the page is not finished |
| The question the page answers, answered in the first two lines | Read the first 40 words aloud. If they are a welcome rather than an answer, send it back |
| A visible last-updated date, changed only when the content changes | Fix a typo and check the date did not move |
| A named author linking to a real page about that person | Click the byline. It should go somewhere, and that somewhere should have a face and a history |
| What it costs, or the honest range and what moves it | Search your own site for a number. If there isn't one, decide that on purpose rather than by default |
| Proof sitting next to the claim it supports | Find your strongest claim. Is the evidence on the same screen, or on a testimonials page nobody opens |
| Who the page is not for | If there is nothing on the site that disqualifies anybody, nothing on the site is believable |
| Internal links inside the sentences, not only in a related-posts strip | Count the links in the body copy. Zero is a fail |
| Exactly one next step | Count the calls to action above the fold. More than one means nobody decided |
| Structured data matching what a reader actually sees | Run the page through Google's Rich Results Test and read what it claims. It must not claim FAQs that are not on the page |
What should you demand on speed?
| Requirement | How you check it |
|---|---|
| The main content is in the HTML, not assembled by JavaScript | Turn JavaScript off in your browser and load the page. If the prices or services vanish, it fails |
| Core Web Vitals passing on mobile, on a real device, not a desktop simulation | PageSpeed Insights on your own phone, on mobile data, not office wifi |
| A stated target for time to first byte, and a number you were given before launch | Ask what it is. "Fast" is not a number |
| Images sized and compressed for the slot they sit in | Look at a page's largest image file. If it is a 4MB photo in a 400px box, nobody checked |
| No render-blocking third-party scripts in the head that nobody can name | Ask for the list of third-party scripts and what each is for. Anything nobody can explain comes out |
Why this matters more than it used to: a growing share of what reads your website will not wait and will not run your JavaScript.
What should you demand on the move itself?
Only if you are replacing an existing site. This is where most of the damage happens.
| Requirement | How you check it |
|---|---|
| A redirect map written before the build starts, every old URL to a new one | Ask to see it. It is a spreadsheet. If it does not exist yet, the launch date is wrong |
| The pages that currently earn traffic identified by name, and carried across | Ask which pages those are. If nobody can name them, nobody looked |
| No ranking page shortened or "tidied" without a conversation | Compare word counts before and after on your top ten pages |
| Staging blocked from search engines, and production not blocked | Load yoursite.com/robots.txt on launch day and read it. See the plain English guide |
| Analytics and conversion tracking live and tested on launch day | Submit a real form on the live site and watch it arrive. Do not accept "the tag is installed" as the test |
| A 404 page that suggests somewhere useful | Type a nonsense URL on your own site and see what happens |
The full version is in our migration guide and there is a pre-launch checklist here.
What should you demand on measurement?
This is the section people cut, and then cannot prove the website was worth it.
| Requirement | How you check it |
|---|---|
| Every enquiry route tracked: form, phone tap, email tap, chat, booking | Do each one yourself and confirm each shows up |
| Each step of a booking or checkout tracked separately, not only the end | Start a booking, abandon it, and see whether that shows |
| The measurement date recorded somewhere written down | If you add consent management, your comparable history ends that day. Write the date down |
| A named person who looks at it monthly | Not "we'll keep an eye on it" |
We got this one wrong on our own relaunch and published what happened, because it is a better warning coming from someone who tripped over it.
What is not worth paying for?
Said plainly, because briefs are usually silent on this and it is where budget goes quietly missing.
An llms.txt file. It was tested properly and not a single crawler came for it. It was on our own recommendations list before anyone checked, which is why we say so out loud.
A visibility score with no method attached. Ask what is in it. Ask again tomorrow and see whether the number is stable.
A rebuild, when the problem is the content. The most expensive answer to "our site is not working" is usually not the right one. Two of our own case notes are projects where we did not rebuild: a South West accountancy practice, impressions up 73% in six months, and a wedding celebrant, organic clicks up 577% off one page.
Anything described as future-proof. Ask what it was tested against.
In what order?
- Decide whether you are rebuilding at all.
- Fix what machines can reach. JavaScript off, load the homepage, see what survives.
- Fix the pages that already earn, using the page table above.
- Rebuild if you still need to, with the redirect map written first.
- Build the narrow pages: glossary entries, one specific use case each.
- Measure what it brings in, and name who looks at it.
Can I use this brief if I am not working with you?
Yes. That is the point of publishing it. Copy it, change it, put your own logo on it, hand it to whoever is building your site.
My developer says some of this is not their job. Are they wrong?
Often not. Content, pricing decisions and who the page is not for are yours. Speed, redirects, schema and tracking are theirs. The useful move is to go through the list and agree out loud which lines belong to whom, before the work starts.
What if the site is already built?
Most of this still applies. Work the page table against your five most important pages, then the speed section, then measurement. The migration section is the only part you can skip.
How much of this can I check myself?
All of the "how you check it" column, with no tools beyond a browser and a phone. That was the constraint when writing it.
What is the single most common failure?
Conversion tracking that was never tested on the live site. It is invisible, it only shows up when somebody asks what the website earned, and by then the months with no data cannot be recovered.
Should I really publish my prices?
We think so, and we do. If your site says nothing, an AI assistant will answer for you and will sometimes be wrong, and you never find out, because that person never gets in touch. If you want someone to go through this with you before you commission anything, that is a short conversation. [Book a discovery call](/discovery-call).