On this page
- A voice profile is a lookup table, not a mood board
- Why the corporate tone-of-voice template fails on a person
- The Personeur Nine-Panel Voice Profile
- What a real entry looks like, panel by panel
- Measure sentence physics instead of describing them
- The story bank runs out unless you size it first
- How long the profile should be, and the maths behind the cap
- What expires, and how often each panel needs refreshing
- Fill it in this order
A personal voice profile is a lookup table a writer consults before each draft, not a document anyone reads for inspiration. It holds nine panels: the identity line, the idiolect list, the banned list, measured sentence physics, the opinion register split by what you will sign, the story bank, openers and closers, forbidden territory, and the correction rules. The writer fills it from recordings. You correct it.
A voice profile is a lookup table, not a mood board
The voice profile is the document a writer opens immediately before drafting and searches for a specific answer. Does this person say clients or customers. Do they open a disagreement with a concession. Is there a story about a pricing mistake with a date on it. Everything in the document exists to answer a question of that shape, and anything that cannot be looked up does not belong in it.
That single design constraint separates a working profile from the version most engagements produce, which is an essay about the executive's personality. An essay is read once, admired, and never opened again. A lookup table gets opened ten times a month, which is the only way a document changes what appears on a feed.
The profile is built from recordings, not from a form. A writer fills it after marking a transcript, and the executive corrects it. Sending a blank template to the executive to complete produces adjectives, because people describe their own speech the way they wish it sounded. The capture method that feeds this document is set out in how a ghostwriter captures a voice.
A brand tone-of-voice guide governs a company across many writers and many channels, and its job is consistency. A personal voice profile governs one writer producing one person's first-person speech, and its job is recognisability. Consistency is achieved by rules. Recognisability is achieved by specifics. They are not the same document with a different cover.
Why the corporate tone-of-voice template fails on a person
The corporate template fails because every one of its fields was designed to make many writers converge, and a personal feed needs the opposite. Convergence is what makes an executive's posts read like everyone else's executive posts. The table below maps each corporate field to the damage it does on a personal feed and to the panel that replaces it.
| Corporate field | What it does on a personal feed | The replacement panel |
|---|---|---|
| Three brand adjectives, usually bold, human and expert | Every executive is assigned the same three, so the writer produces the same feed for all of them | Panel 2, the idiolect list, sourced with timestamps |
| Do and do not columns for tone | Written by committee about a company, so it quietly bans jokes, hedges and admissions | Panel 3, the banned list, taken from what this person actually refuses |
| A reading age or grade level target | Flattens the one long sentence that was carrying the argument | Panel 4, sentence physics, measured from real speech |
| Messaging pillars | Produces themes with no position inside them, which is how status updates happen | Panel 5, the opinion register |
| Approved proof points and boilerplate | Company proof does not work in first person, because nobody thanks a market share figure | Panel 6, the story bank |
| Sample copy written in the brand voice | Marketing register, which is the wrong register for a person talking to peers | Panel 7, openers and closers from real conversation |
| A legal appendix | Arrives at approval, after the draft is written and the time is spent | Panel 8, forbidden territory, mapped before drafting |
Corporate tone-of-voice fields, what they do to a personal feed, and the panel that replaces each one.
One field survives the translation intact. The approval path matters just as much for a person as for a company, and it belongs in panel nine, next to the correction rules, because those two together decide whether anything ever improves.
The Personeur Nine-Panel Voice Profile
Nine panels, each with a defined field list and a stated size. Sizes matter as much as contents, because a panel with no ceiling grows until nobody reads it.
Panels one to five plus nine are what a writer re-reads. Panels six to eight are searched on demand. Splitting the document this way is not tidiness, it is the reason the profile keeps getting opened in month nine.
What a real entry looks like, panel by panel
Entries fail in a predictable direction, which is upward into abstraction. Below is a composite illustration, invented for this page and not a real client, showing the difference between an entry a writer can use and the version that usually gets written.
Idiolect entries, usable against unusable
| The version that gets written | The version that works |
|---|---|
| Speaks plainly and avoids jargon | Says the honest version is before delivering any correction, twice at 14:20 and 31:05 |
| Uses humour | Undercuts his own claims with a short aside in brackets, roughly once every third paragraph |
| Warm and approachable | Calls any company under fifty people a shop, including his own |
| Prefers simple language | Never says journey, says the last two years, three separate times |
Composite illustration. The left column is what gets written. The right column is what a writer can act on.
A story bank entry with every field filled
- Date and place: March 2023, a customer's warehouse two hours outside the city, first site visit of that account.
- Stake: a contract worth about a fifth of that year's revenue, already proposed and priced.
- The turn: the operations lead walked him through the real process, which looked nothing like the process described in the proposal he had already sent.
- The line he actually says: we had written a proposal for the company on the website, not the company in the building.
- Arguments this story can serve: why discovery beats a briefing document, why a site visit outperforms three calls, how to reprice inside a live deal without losing it.
The last item is the field nobody templates and the one that changes the economics of the whole document. A story tagged with three arguments is three posts. A story with no tags is one post, and then it sits in the bank being described as a good story until everyone forgets it.
The opinion register, filled honestly
| Will defend publicly | Believes, will not publish | Undecided |
|---|---|---|
| Discovery calls should be paid, and free ones select for the wrong buyer | Two named competitors are selling a service they cannot staff | Whether the firm should publish its pricing |
| Most operations problems are scheduling problems wearing a costume | The industry body's certification is worthless | Whether remote implementation genuinely works past fifty seats |
Composite illustration of panel five. The middle column is the one that keeps an executive out of trouble.
A writer who has the middle column stops accidentally drafting it, which is the single most common near-miss in executive ghostwriting. A writer who has the third column knows what to ask about next month.
- A company tone-of-voice guide and a personal voice profile share almost no fields, because one describes how an organisation should sound and the other records how one specific person already sounds.
- Sentence physics should be measured from thirty consecutive sentences of unrehearsed speech rather than described in adjectives, which turns a vague instruction into a median, a ninetieth percentile and a hard ceiling.
- The story bank runs out by month three unless every entry is tagged with the arguments it can serve, because a year of posting needs roughly twice as many story slots as an executive naturally supplies.
- Keep the working profile to two pages and push the lists into appendices, because a ten-page profile costs a writer around fourteen hours a year in re-reading and will quietly stop being opened.
- The opinion register decays fastest of the nine panels and is the only one that needs a quarterly rebuild, which is what the monthly capture call is really for.
Measure sentence physics instead of describing them
Sentence physics takes twenty minutes to measure and removes an entire category of argument from the engagement. Take thirty consecutive sentences of unrehearsed speech from the transcript, after the warm-up is stripped, and count the words in each.
Thirty consecutive sentences, spoken not written, from a passage where the person was explaining something rather than answering a direct question.
Word count per sentence, sorted ascending. Ignore filler words at the start, keep everything else.
The median, the twenty-seventh value which is roughly the ninetieth percentile, and the longest. Those three are the panel.
Target the median, allow the ninetieth percentile once or twice per post, treat the longest as a ceiling that is never exceeded.
A composite illustration makes the output concrete. Suppose the sorted sample runs 4, 6, 7, 9, 9, 11, 11, 12, 13, 13, 14, 14, 15, 15, 16, and continues upward to a single 44. The median lands near 14, the ninetieth percentile near 31, the ceiling at 44. The rule written into panel four then reads: median 12 to 16, one sentence over 30 per post, never past 44, and never three consecutive sentences under 6 words because this speaker does not talk in fragments.
Text generated by a model tends to cluster tightly around a single sentence length with very little variance. A real speaker's distribution is lumpy, with short asides sitting next to a long sentence that carries an argument. Recording the spread, not just the average, is what makes a draft pass as speech.
Paragraph physics is measured the same way and matters more on a feed than in prose, because the platform truncates. Count how many sentences the person produces before changing idea. If the answer is two, the drafting rule is two, regardless of what reads well on a page.
The story bank runs out unless you size it first
Size the story bank against the publishing plan before month one, because the default arrangement empties it around month three and nobody notices until the drafts turn abstract. The arithmetic is simple and worth doing out loud.
Assume ten posts a month, which is 120 a year, and assume one post in three carries a story. That is about 40 story slots a year. Now assume the supply side: a founding capture call typically surfaces three to five properly formed stories, and a twenty-minute monthly top-up adds one or two. Call it four at the start plus roughly sixteen across eleven months, so about twenty stories against forty slots.
| Line | Assumption | Stories a year |
|---|---|---|
| Demand | 10 posts a month, one in three carries a story | 40 |
| Supply, founding call | 3 to 5 entries with date, place, stake and turn | 4 |
| Supply, monthly top-up | 1 to 2 entries a month across 11 months | 16 |
| Deficit | Demand minus supply | 20 |
| Closed by tagging | Each entry tagged with 2 arguments it can serve | 40 usable slots |
Story bank supply against demand, on stated assumptions. Adjust the assumptions and the deficit moves, but the shape stays.
There are exactly three ways to close a gap of that size. Tag each story with two or more arguments it can serve, which doubles usable slots without inventing anything. Lower the story ratio to one post in six and carry the rest on positions and observations. Or raise capture, which means more executive minutes and a different price, covered in what your own time in the process costs.
The wrong way to close it is the one that happens by default. A writer short of material starts stretching thin anecdotes into stories, and thin anecdotes are exactly the posts that read as generic. The bank running dry is not a content problem, it is a measurement problem that was never done.
How long the profile should be, and the maths behind the cap
Cap the working profile at two pages, roughly seven hundred words, and push every list into an appendix. The reason is a cost that never appears on an invoice.
A writer re-reads the working profile before drafting. Assume ten drafts a month. A two-page profile takes about ninety seconds to scan, which is fifteen minutes a month and three hours a year. A ten-page profile takes about seven minutes to read properly, which is seventy minutes a month and roughly fourteen hours a year. Nobody spends fourteen hours a year re-reading a reference document. They read it twice, absorb an impression of it, and then write from memory, which is the same as not having one.
| Working profile length | Time per re-read | Cost per month | Cost per year | What actually happens |
|---|---|---|---|---|
| 2 pages, about 700 words | 90 seconds | 15 minutes | 3 hours | It gets opened every time |
| 5 pages | 3.5 minutes | 35 minutes | 7 hours | Opened for the first month, skimmed after |
| 10 pages | 7 minutes | 70 minutes | 14 hours | Read twice, then written from memory |
Re-reading cost at ten drafts a month, on stated assumptions about reading speed.
Appendices do not carry this cost, because they are searched for one answer rather than read from the top. The idiolect list can run to sixty entries, the story bank to forty, and the forbidden territory panel to whatever length reality requires, as long as none of them sit inside the two pages a writer scans before drafting.
What expires, and how often each panel needs refreshing
The nine panels age at wildly different rates, and treating the profile as one document with one review date wastes effort on the stable panels while the volatile one rots. Refresh on the trigger, not on the calendar, with one exception.
| Panel | How fast it ages | What triggers a refresh |
|---|---|---|
| 1, identity line | Very slow, a year or more | A role change, a new offer, a new audience |
| 2, idiolect list | Barely moves in two years | Additions only, never a rewrite |
| 3, banned list | Slow, grows steadily | Any correction that names a specific word |
| 4, sentence physics | Slow | A new medium, such as a newsletter or video scripts |
| 5, opinion register | Fastest of all, stale within a quarter | Rebuild quarterly, and after any market event the person commented on |
| 6, story bank | Grows monthly, never expires | Every capture top-up |
| 7, openers and closers | Slow | A change in what the audience responds to |
| 8, forbidden territory | Event driven | New client, new agreement, litigation, an employer change, an expiry |
| 9, correction rules | Continuous | Every draft cycle, which is the point of the mechanism |
Refresh cadence by panel. The opinion register is the exception that runs on the calendar.
Panel five is the one that decides whether year two of an engagement is worth paying for. Opinions that were sharp in January are consensus by September, because if the position was any good other people arrived at it. A writer still drafting from a stale opinion register produces posts that are technically in the right voice and nine months late, which is the specific failure examined in drafts that do not sound like you.
of members post more than once a week. Sustained output is rare because drafting from scratch is expensive every single time, and a profile that answers the writer's questions in ninety seconds is what makes the cost survivable.
Aggregate 2026 LinkedIn statistics reportsFill it in this order
Build the profile in the order below, because early panels calibrate the judgement needed for later ones. The whole document takes a competent writer around three hours after a founding call, and it should be dated and versioned from the first draft.
- The identity line is one sentence and names a specific reader, not a segment.
- The idiolect list holds at least thirty entries and every entry carries a source or a timestamp.
- The banned list is split into what the person refuses and what writers wrongly reach for.
- Sentence physics shows three numbers taken from a counted sample of thirty spoken sentences.
- The opinion register has entries in all three columns, including at least two the person will not publish.
- Every story bank entry has a date, a place, a stake, a turn, a spoken line and at least two arguments it can serve.
- Openers and closers are quoted from real conversation rather than described.
- Every forbidden territory entry carries a reason, so it can be released when the reason expires.
- Correction rules are written as classes, dated, and filed against the panel they belong to.
- The working pages fit inside two pages and every list has moved to an appendix.
- The executive has corrected the draft rather than approved it, and the corrections are visible in the version history.
One last rule, and it is the one most often broken. The writer fills the document and the executive corrects it. A blank template sent to an executive comes back with adjectives, because people are unreliable narrators of their own speech and will delete the exact habits that make them recognisable. Correction is easy and accurate. Composition from memory is neither.
Questions people ask next
Who should own the voice profile, the writer or the executive?
How is this different from a brand tone-of-voice guide?
Can I build a voice profile for myself, without a ghostwriter?
How many entries should the idiolect list have?
What goes in the forbidden territory panel if there is no legal risk?
How often should the whole profile be rebuilt from scratch?
Send one old post.
Email any post you have published, with your LinkedIn URL. A rewrite in your voice comes back free, so you can judge the craft on your own words.
Get the free rewrite →
