The gap it fills
Most companies have two things: a positioning statement written in a workshop, and a website written by whoever wrote the website. Between those two there is a gap, and everything falls into it. The positioning statement is too abstract to write a page from. The page is too specific to check a deck against. So the deck drifts, then the ads drift, then the sales team improvises, and eighteen months later four parts of the company are describing four different products.
A message architecture is the thing in the middle. It is short, it is structural, and its job is to be the document that any piece of copy can be checked against. Not a tone of voice guide, not a brand book, and not a list of features. A structure of claims.
The shape of the artefact
It has four parts and it fits on one page. If it does not fit on one page it is a strategy document, and strategy documents do not get used by the person writing an ad on a Tuesday.
- The audience. Who this is written for, specifically enough to exclude people. Not mid-market companies, but the platform engineering lead at a company running more than one production cluster.
- The primary claim. One sentence, the thing you want a buyer to repeat to a colleague when you are not in the room. It has to survive being said out loud by somebody who does not work for you.
- Three supporting claims. The three things that would have to be true for the primary claim to hold. Three, because two looks thin and four is nobody's memory. Each one is a claim, not a feature.
- The proof each claim needs. Underneath every supporting claim, the specific evidence that entitles you to say it. This is the part that gets skipped and it is the part that does the work.
The proof column is the whole point
Almost every messaging document you have seen stops at the claims. That is why almost every messaging document you have seen is ignored. Claims are free. Anybody can write that a product is fast, is safe, is loved by developers. The proof column is what turns the artefact from a wish list into a plan.
Filling it in has an uncomfortable side effect, which is the reason it is valuable: you find out which of your claims you are not currently allowed to make. A supporting claim with an empty proof cell is not a message, it is a task. Either you go and get the evidence, which is a real piece of work with an owner and a date, or you drop the claim from the copy until you have it.
This is the single fastest way to improve a B2B homepage. Not better words for the claims you already make, but deleting the two claims you cannot evidence and expanding the one you can.
The fifth section: what you are not saying
Underneath the four parts, one more list: the things this message deliberately gives up. A cheaper option, a broader audience, a use case you could serve but will not lead with, a claim a competitor makes that you are choosing not to match.
This exists because messages do not usually fail by being wrong. They fail by widening. Every quarter somebody has a reasonable request to add one more audience, one more use case, one more claim, and each request is individually sensible. Two years of individually sensible additions produce a homepage that says nothing. A written list of exclusions gives whoever owns the message something to point at when the request arrives, which is the difference between holding a position and having opinions about one.
How the homepage inherits from it
Once the architecture exists, the homepage stops being a design question. The primary claim is the hero. Each supporting claim gets a section, in the order that matches how a buyer's objections arrive rather than how your product is organised internally. Each section carries its own proof, next to the claim it supports, rather than pooled in a testimonial band at the bottom where it proves nothing in particular.
The order matters more than the wording. Buyers do not read a homepage as a brochure. They read it as a sequence of questions: what is this, who is it for, why should I believe you, what does it cost me to find out. A page that answers those in order feels obvious. A page that answers them in a different order feels like work, and the reader does not usually tell you that, they just leave.
Who has to sign it
One person, and they have to be senior enough to defend it when a colleague disagrees. A message architecture agreed by a committee and owned by nobody lasts until the first person who was not in the room writes something.
Ideally the same person owns the exclusions list, because that is where the pressure lands. The claims are easy to agree to. Saying no to a new audience in month seven is where the document either does its job or quietly stops existing.
A test you can run today
Ask four people in your company, separately and in writing, to answer three questions: what does the product do, who is it for, and what does it replace. Do not let them look at the website first.
If the four answers are recognisably the same answer, you have a message and your problem is execution. If they are four different companies, no amount of design or copy will help until somebody writes the architecture and signs it. That test takes an afternoon and it is the cheapest diagnostic in marketing.