Most marketing plans are documents nobody opens twice. They get written during a planning week, contain a competitive analysis and a persona deck and a quarterly calendar, and then reality diverges from them within a month.
The problem is not that planning is useless. It is that most plans specify the wrong layer. They specify outputs — thirty posts, four channels, two ebooks — which are exactly the things that should change as you learn. They leave unspecified the things that should stay fixed: what you claim, who for, and how you will know if it worked.
Here is the inversion. Six fields, one page.
1. The claim
One sentence: what you assert that a reasonable competitor would not.
Not a value proposition. A position — something with an implied opponent. "We help teams grow" is not a claim, because nobody argues the reverse. "Generated content plateaus without a feedback loop" is a claim, because plenty of people are selling the opposite.
If you cannot name who disagrees, you do not have one yet. Everything downstream is weaker without it, because a claim is what makes content worth reading rather than merely accurate.
2. The reader
Not a persona with a stock photo and a name. One sentence describing someone in a specific situation.
"Solo founder who has been posting for three months and cannot tell which posts are working" is useful — it tells you what they already know, what they are frustrated by, and what would be condescending to explain.
"Marketing managers aged 25–40" is not useful. It cannot tell you anything about how to write a sentence.
3. The surfaces
Which platforms, in what order, and why that order.
Two rules. Start with one — the contracts are learned rather than looked up, and learning several at once means learning none. Add the second only when the first has a working voice, and pick one whose norms are far from the first, because two similar platforms will not teach you the difference between your voice and the platform's conventions.
Write down for each: what the platform rewards, what gets you downranked, and what "good" looks like numerically. That last one you cannot fill in on day one — leave it blank and fill it from your own data, not from a benchmark article.
4. The cadence
The rate you can sustain for six months while everything else in your life continues.
Whatever number you first thought of, halve it. Consistency beats volume on every platform in this dataset, and the failure mode is universal: sprint for five weeks, miss one, feel bad, stop. A cadence you hit is worth several times one you miss.
5. The verdict rule
The field that separates plans that compound from plans that do not, and the one almost nobody writes down.
Decide in advance what counts as a win, and how long you will wait before judging. Both parts matter.
Pick the metric your platform actually distributes on. On Xiaohongshu that is saves, not likes. On TikTok it is watch-through. On Reddit it is upvote ratio and comment quality. Using the wrong metric means optimizing for something the algorithm does not reward, which is worse than not measuring.
Then set the window. Long tail is real on several platforms — a post can surface weeks after publishing — so a twenty-four-hour verdict systematically mislabels your best work as failure. Pick a window, write it down, and do not judge before it closes.
Write the rule as an if-then: "A post is a win if saves-per-impression clears X within seventy-two hours." Fill in X from your own median after your first dozen posts. Not from an industry benchmark — those are averages across accounts nothing like yours.
6. The feedback step
What physically happens to the numbers after the window closes.
This is where plans die. Not from lack of ambition — from having no mechanism. Results come back, you look at a dashboard, you feel vaguely informed, nothing persists. Repeat forty times and you have a year of output and no accumulated knowledge.
Specify the mechanic: where results get recorded, in what structure, and — critically — how they reach the next thing you write. If the answer is "I'll remember," the answer is no. You will remember your three most recent posts and nothing before that, and you will misremember at least one of them.
The minimum viable version is a spreadsheet: date, topic, tone, the verdict metric, win/loss. Review it before writing the next batch. That is genuinely enough to beat most people, because most people have nothing.
Why this order
The fields are ordered by how often they should change.
The claim and the reader should change rarely — those are strategy, and if they change monthly you do not have one. Surfaces change quarterly. Cadence changes when your life changes. The verdict rule changes when you have enough data to sharpen it.
The thing everyone writes plans about — topics, calendars, specific posts — is not on the page at all. That is downstream output, it should change constantly in response to what the feedback step tells you, and fixing it in a document is what makes plans go stale.
Fill in the six fields. Delete the thirty-page version.
The one-page version
CLAIM — one sentence, has an implied opponent
READER — one person, one situation
SURFACES — platform 1, then platform 2, and why that order
CADENCE — n per week, sustainable for 6 months
VERDICT — a post wins if [metric] clears [X] within [window]
FEEDBACK — results go to [where], read before [next batch]If your plan does not fit here, the excess is probably output specification — and output specification is what you should be deriving from feedback rather than deciding in advance.
clownfish101 automates fields five and six: deterministic verdict rules read your real numbers — no model picks a number — and the results feed back as the next batch's examples and counterexamples. See the loop →