My ghostwriter's posts don't sound like me: locate the failure before you fix it

Two different failures produce the same sentence

This does not sound like me is the most common sentence in executive ghostwriting, and it describes two unrelated problems. In the first, the writer has not captured how you speak, so the prose is competent and belongs to nobody. In the second, the voice is right and the substance is empty, so you are reading your own unformed thinking in your own cadence, which is a genuinely unpleasant experience and produces the same complaint.

The two need opposite responses. A capture failure is fixed by giving the writer more raw material and a corrected profile. A point-of-view failure gets worse with more raw material, because the writer will faithfully reproduce more vagueness. Nearly every published piece of advice on this subject assumes the second problem does not exist, and tells the client to give better feedback.

Feedback cannot repair a missing position. The diagnostic below tells you which failure you have before you spend another month editing drafts, and it runs in about thirty minutes on a single bad draft.

The Personeur Four-Probe Diagnostic

Four probes, run in order, on one draft you disliked. Each returns a single letter: C for capture failure, P for point-of-view failure. Do not skip to the fix, because the two fixes cancel each other out.

Probe two is the one to run if you only have five minutes. It is the closest thing this work has to a decisive test, because it separates a missing input from a missed input, and those two look identical on the page.

Reading the four probes together

Count the letters and read the row that matches. Mixed results are the most common outcome and they have a specific order of operations, because voice cannot be judged on a draft that argues nothing.

Four-probe results, the diagnosis, and what to do first.

The last row is the one most often misread. Work the arithmetic before you accuse anyone. Assume conversational speech runs near 140 words a minute and that you hold about seventy percent of the talk time, both stated assumptions you can check against your own transcript. Eight minutes of hurried voice notes gives a writer roughly 780 words of your material. Ten posts at 200 words is 2,000 published words. Around sixty percent of what you are reading had to come from somewhere other than you, and it will sound exactly like that.

If it is capture: what to send the writer this week

Send raw material and specific corrections, not adjectives. A capture failure is repaired by evidence of how you speak, and the fastest evidence is speech that already exists rather than a new interview.

One thing not to send is a list of writers you admire. It moves the target from your speech to somebody else's and produces a second stranger, this time an imitation of a person you follow rather than an imitation of the average executive.

If it is point of view: the fix is not a writing fix

A point-of-view failure is repaired by deciding what you think, in specific enough terms that someone could argue with it. No writer, price or process substitutes for this, and the honest version of the situation is that you are paying a production service to solve an editorial problem.

The mechanism is worth understanding, because it explains why the drafts feel worse than badly written ones. A skilled writer with a good profile will reproduce your register accurately. If the underlying material is a set of unobjectionable observations, you will read your own cadence saying nothing, and recognising yourself in something empty is more uncomfortable than reading a stranger. The complaint that comes out is still that it does not sound like you.

Turning an observation into a position. Composite illustration, not client work.

Three questions produce more usable positions than any brainstorming session. What do you believe that most of your competent peers do not. What did you get wrong in the last two years, with the date and the cost attached. What do clients ask you that has an answer you have never written down. Answer those on a recording and the next draft has something to be in the voice of.

The correction ledger, which is what makes a fix stick

A correction survives to the next draft only if it is written as a rule and filed against a panel of the profile. Corrections given as edits fix one sentence. Corrections written as classes fix a category and stop the same note recurring for the rest of the engagement.

The correction ledger. Composite illustration of five rows from a first quarter.

The cost of skipping the ledger is measurable. Assume ten drafts a month and six corrections per draft, which is sixty corrections a month. A correction that never becomes a rule returns on the next draft. One recurring correction takes about ninety seconds of your attention each time, so across ten drafts a month for twelve months it costs three hours of executive time for a single word. Five recurring corrections cost fifteen hours a year, and they are the same fifteen hours that make people conclude the engagement is not working.

Two operating rules make the ledger work. Every correction gets generalised at the moment it is given, by the writer, in writing, and read back to you. And a rule that recurs twice after being filed is no longer a writing problem, it is a process problem, which is the threshold used in the final section.

The third cause: the draft was fine until three people edited it

Compare the writer's first draft against what actually published before you conclude anything. In organisations with a communications team, a legal reviewer or a chief of staff in the chain, the flattening frequently happens after the writer's involvement ends, and the writer is then corrected for someone else's edits.

The test takes ten minutes. Put the writer's draft next to the published post and count the sentences that changed. If more than a quarter of them were altered by reviewers rather than by you, the feedback the writer receives is describing a text they did not produce.

Reading the difference between the writer's draft and the published post.

The structural fix is to review once, early, on the writer's draft rather than on your approved version, so every correction lands in the same place and reaches the person who can generalise it. Where the chain has to include compliance or legal, the sequence matters even more, which is the subject of the piece on how an approval workflow should run.

Watch for the version of this that involves nobody but you. Executives edit their own drafts toward safety late at night, publish the safer version, and then report that the drafts are flat. The diff test catches that too, and the answer is uncomfortable rather than complicated.

When to conclude it is the writer

Change supplier when two or more of the four thresholds below are true after draft ten, and not before. One threshold alone is a conversation, because most of these have a repairable cause that has nothing to do with talent.

Before you move, count your own recorded minutes across the last three months and be honest about the total. A capture failure caused by an executive who supplies eight minutes a month reproduces perfectly with the next writer, at the cost of another onboarding period and another three months of unrepresentative drafts. The time arithmetic is set out in what your own time in the process costs.

If you do move, structure the next engagement so this diagnosis happens in week three rather than month four. A properly designed trial produces a blind attribution check, a filled voice profile and a correction ledger before the retainer starts, which is the whole argument in the piece on designing a trial period.