A post on , , for someone thinking .
What they believe now: They should feel , and are asked to .
Click the coloured words to change them.
Select the nearest and finish it in their words.
Audience
Puts it on the month's shoot list.
Checks the finished brief before it goes anywhere.
Name the mechanism: what in the reader's position makes them act. A description of the idea as interesting, relatable or strong does not count. For example, the reader has already paid for something that is not working, and that sunk cost pushes them toward finding a fix rather than accepting the loss.
Consult these when naming the mechanism. A mechanism outside them passes if it accounts for how the reader behaves.
Worth doing when the same reference keeps coming back, or when a post built from it beat its own account.
For a video, the opening frame and one at the payoff. For a carousel, the first two panels.A structure is only claimed when it appears in posts that beat their own account and does not appear as often in posts that did not. Below five rows nothing is claimed.
One row per post, each result read against its own account. Patterns are only read above five rows.
Choose a client to see their month.
Completed once at onboarding, displayed at every point of writing, and revised as the work informs it.
To start writing you need the name, the business model and why they're looking. Everything else can be filled in as it becomes known.
If no model fits on two or more of the questions, choose Custom and write the answers in.
New ones start as Written by us. Change it to where it came from.
The shape
A moment, not a person. Something has happened, the reader has noticed it, and they have not decided what to do yet. If it contains an age range or an income bracket, it is a demographic.
Where they come from
DMs, reviews, comments, and what people say in the first two minutes of a sales call. Harvested rather than invented, which is why they are written in the words used there.
The test
Could the person say it out loud, in those words, to a friend? If not, it is a category, and a category gives you nothing to write a hook against.
No worked example here on purpose. The client's own DMs are the source.
Checking.
Through your team hub login. Your reads and the daily limit are counted against this address. There is nothing to set here.
Estimated from published rates, in US dollars. Every read is paid for, so the count in the corner is there to be watched rather than ignored. The daily limit stops one person running away with it.
The first four are judgement. The rest are craft, and the rules are shown as you work.
An in-house system is a strong proof point, because most agencies work with whatever the platform gives them. It earns its place when it shows the reader something they could not otherwise see, and the reader comes away thinking these people look harder at an account than anyone else would.
It stops working the moment it sounds like the account runs itself. The team is the subject of the sentence and the tool is what they use: we check this every week against the account's own history, never the system checks your account. Same fact, opposite meaning.
Where the tool is itself the product being sold, the reader is judging whether it fixes a problem they have, so the product can be the subject. That is set per client on the Clients tab, because the same feature reads differently depending on which brand is speaking.
A brief carries direction: the reader, what the post is for, the visual hook, the structure, and where the words are going. It goes to the client as a concept, and what they approve is the direction rather than the wording.
Copy is written after that approval, against the approved concept, with the client's tone and the constructions to avoid in view. Writing final copy inside the brief means writing it twice, because the concept can still change.
The internal reasoning never goes into the concept row. That field is client-facing.
The order matters for production:
Handing a brief over before that risks making work that gets rejected on direction rather than on craft.
Take any sentence in the post and ask whether a competitor could say it word for word. If they could, it says nothing, however true it is.
Claim words fail it when they carry the argument: when the word is the reason given to choose them or to believe the line. Compassionate, world class, expert, trusted, exceptional, state of the art, specialised equipment, dedicated team. The same word shown as a badge or stated as a fact is fine. Stacking them is not, because a reader processes a run of qualifiers as marketing noise and stops reading.
What passes it is anything a competitor would have to change: a named piece of equipment, a named person, a number, a length of time, a condition, a procedure, a thing that happened. Specificity is what credentials cannot buy.
Run it on the hook, the caption opening and any claim. One failed sentence is usually the whole reason a post reads as generic.
Run it on the device too. If a competitor could use the same device without changing anything, the device is not yours either.
Run once on the finished brief. A failure names the reason and never rewrites the work.
Two reads, one at each end of a concept. The first proposes and you judge; the second tests the finished brief. Neither writes the copy. Why it will work comes with each opening, and accepting it unchanged is recorded as such.
Each stage has a message in the reader's own words, and the post is judged against it. Select a stage.
Awareness, consideration and conversion run in sequence. Reassurance supports conversion by removing the reason not to act, and retention speaks to people already inside, so neither is a narrower stage of the same audience.
Familiarity runs alongside all five and is carried by consistency of look and voice. Profile assets sit outside them, carry no ask, and are counted separately.
Chosen to produce a named feeling. Select one.
A person scrolling is not looking for this content. They are not in a buying mood, not paying attention, and not obliged to give anything. Every principle below exists to deal with that one fact. Most business content is written as though the reader has already agreed to listen.
Sixteen principles, each changing something specific in a brief. Select one.
Everything above is general. It applies to any audience anywhere, which is exactly why it cannot say what to post for one client.
The principle explains why a mechanism works. The client profile decides which one is right here: the tier, why they're looking, what the reader already believes, what the brand is allowed to claim. Applied without the profile it produces posts that are well made and about nothing in particular, and that is the failure mode to watch for as someone gets good at this.
Each action has its own trigger. Asking for one does not produce it. Select an action.
Named so the team has a vocabulary. Every decoded post is filed under one of these, which is what lets patterns be found later.
The image is what makes someone stop. It also suggests something about the subject before a word is read, and that is where the message comes from.
Four steps, in order. What makes them stop. What the image suggests, listed plainly. Which one of those things this brand can credibly own, given what it sells and what trust is built from for its kind of business. Then the words, written from that one thing.
Two mistakes. Explaining the picture instead of reading it, which throws away the feeling and replaces it with a fact nobody asked for. And using an image whose suggestions the brand cannot own, which earns views and lands nothing. Reach without enquiries usually comes from the second one.
Where nothing is ownable, the honest options are to change the message to one the image can carry, or to keep the post as familiarity and let it do that job well.
Service, care
Something small and fragile placed in a setting built for something much larger. Suggests delicacy and vulnerability. A business whose work is delicate owns that, and the message is how much care something that fragile needs.
Product, range
An empty space filled to finished in a few seconds. Suggests transformation and ease. A business that sells the outcome owns ease, and the message is how little work it took.
Expertise
A picture assembled from parts with pieces visibly missing. Suggests incompleteness the viewer had not noticed. A business whose value is seeing further owns that, and the message is what the reader cannot currently see.
Three unrelated categories, one move. The move travels; none of these is the answer for your client.
Each holds attention for a different length and suggests different things.
The same idea is written differently by channel and platform.
Briefing is layer three. It runs on stated knowledge today and reads evidence in its place as the other layers fill.
Whether the month shares a look, a voice and an emotional register, or reads as unrelated good ideas. This is what carries familiarity, which the framework names and does not yet check.
Needs the month reading finished posts rather than briefs.
The plan stays the system of record. A row opens into a brief here and the brief saves back to the row, with the internal columns never rendered to the client.
Needs this page hosted beside the portal so it can reach the worker.
Two to three weeks after publishing, the post joins its own account's rows and is read against the account's own history.
Needs the mechanism recorded at publish and the organic pull feeding rows.
Declared slots. None of these produce output yet.
Paste an account and get its follower count, its posting cadence and how a given post did against its own median, so a reference can be filtered without anyone judging by eye.
Needs the Instagram business discovery API. Public engagement and follower counts are available; views and retention are not.
Twenty rows from one competitor account in a single pass rather than one row from one link, which is what makes patterns arrive at all.
Needs the baseline above, plus a place to run it that is not a browser page.
A flag when a structure or a mechanism has been used too often for one client, and the next variant worth trying.
Needs rows with dates and the client recorded.
Same video, one change from a fixed list, and the outcome read against the previous version and the account's own history.
Needs published posts carried through to their results, which means the reporting handoff.
Read weekly. A change that alters a rule alters what the brief reads against.
Sources worth watching: Meta newsroom and the Instagram and Facebook creator accounts, TikTok newsroom, LinkedIn engineering and marketing blogs, and Google Ads announcements.
Six models, named by what the reader is deciding. Industry is not one of the questions. A model sets what trust is built from and what the reader risks, not the format: the same model fits a named person on camera and a company showing its work.
One model, three industries
Handing something you care about to a person covers a vet, a physiotherapist and a driving instructor. The same answers hold for all three.
One industry, two models
A clinic selling routine appointments and a clinic selling a course of treatment costing thousands are different models, because the decision is different.
Where it breaks
When the answers diverge on three or more lines, it is a new model. Forcing a client into the nearest one is what produces content that reads as generic.
Industry is never the test. The decision is.
Name the reader's decision first, then answer these before looking at the models. Diverging on three or more lines makes it a new model.