A message house is a single page holding what your organisation is trying to say. A roof — the one umbrella message. Three columns beneath it — the supporting messages that explain and substantiate the roof. A foundation — the proof: figures, examples, the specific things that make the claims above credible.
The structure is not the interesting part; you can draw it in a minute. What makes a message house work or fail is whether the choices inside it were real, and whether anyone can retrieve it when a journalist is on the phone.
The roof: one sentence, and it has to hurt slightly
The umbrella message is the single thing you want someone to take away. One sentence, in words a non-specialist would use.
The test is whether it excludes anything. "We are committed to excellence in everything we do" is not a message, because no organisation would claim the opposite — it makes no choice and so carries no information. A real umbrella message should make some legitimate alternative unavailable to you. If saying it costs you nothing, it is decoration.
If the opposite of your message is absurd, you have not written a message.
The good version usually feels slightly uncomfortable in the room, because it means giving up the other things you also wanted to say. That discomfort is the exercise working.
The columns: three, and they must be different in kind
Three supporting messages, four at the outside. Not because of any theory, but because nobody holds seven under questioning — including the person who wrote them.
The common failure is three columns that are the same claim in different clothes. If a journalist could collapse all three into one sentence, you have one message and two restatements. Useful columns tend to differ in kind: one about what you do, one about how or why you do it differently, one about the effect on the people it is meant to serve.
Each column should be sayable aloud without embarrassment. Read them to someone outside the sector. If they ask what it means, it is strategy language, not a message.
The foundation: proof, with dates on it
Underneath each column, the specific evidence: figures, named examples, third-party validation. This is the part that decays, and the part most often left to rot.
Put a date against every proof point and review them quarterly. A spokesperson citing a figure that turns out to be three years old has not just lost that point — they have handed the interviewer a reason to doubt everything else. Stale proof is worse than no proof, because it looks like carelessness at best.
In short
- Roof: one umbrella message that excludes something. If its opposite is absurd, rewrite it.
- Columns: three, differing in kind, each sayable aloud without embarrassment.
- Foundation: proof points with dates, reviewed quarterly.
- The test is recall, not completeness — if nobody can say it from memory, it is not working.
Why most of them die
Three failure modes, all of them about use rather than content.
It was written to be approved
A message house that has been through six rounds of review usually reads like it. Every specific has been softened, every edge sanded off, and what survives is unobjectionable and unusable. Approval and usefulness pull in opposite directions here, and someone has to defend the specifics.
It lives where nobody looks
If retrieving it requires remembering which shared drive, it will not be retrieved at 16:40 with a deadline running. It needs to be somewhere people already are, and short enough to hold in your head — which is the practical argument for one page rather than twelve.
It was never tested against a hard question
A message house built in a calm room reads fine in a calm room. Put it against the questions you would least like to be asked and the gaps appear immediately: the column that has no answer to the obvious objection, the proof point that invites a follow-up you cannot handle. See 30 questions every spokesperson should rehearse for the stress test, and media training for putting a person through it.
The recall test
Two weeks after you finish it, ask three people who were in the room to write down the umbrella message and the three columns from memory, separately, without warning.
You will not get identical wording, and you should not expect to. What you are looking for is whether all three produced the same meaning. If one of them has a fourth column that does not exist, or renders a pillar as something noticeably different, the framework has not landed — and it will not survive contact with a journalist, because the version that reaches print is whichever one the person on the phone happened to remember.
That failure is usually a wording problem rather than a strategy problem. Messages that survive recall are short, concrete and use ordinary words. Messages that decay are abstract, qualified, and contain a conjunction doing too much work.
Keeping it alive
A message house is not a document you finish. Practically, three habits keep it real: quarterly proof-point review, an explicit owner who is allowed to say no to additions, and a standing rule that anything drafted should be traceable to a column. That last one is the discipline that stops the framework and the actual output drifting apart — see keeping every channel on message.
Make the framework operational. Intorian holds your messaging pillars alongside the voice fingerprint, and scores every draft for whether it actually advances them — so message alignment is a number you can see rather than an argument about whether something feels on-brand.
See how it works