Agency or freelancer: you are choosing which continuity to risk

The cost comparison is the least decisive thing about this choice

Nearly every published comparison of ghostwriting agencies and freelancers is a price comparison with a conclusion already attached, and it is usually written by one of the two parties. Price is real, and it is also the only variable in the arrangement you can renegotiate later. The variables you cannot renegotiate are the ones almost nobody writes down.

All of them concern what happens when the arrangement is disturbed. Somebody gets ill. A larger client appears. A writer is promoted off your account. The engagement ends and you need the material. Those four moments decide whether you got what you paid for, and none of them appears on a pricing page.

Structure is also the second question rather than the first. Before deciding which operating model you want, the harder problem is judging whether the samples are even theirs, because an unverified portfolio makes both models look identical from outside.

You are choosing which kind of continuity to risk

A solo writer gives you voice continuity and risks output continuity. An agency gives you output continuity and risks voice continuity. That sentence resolves most of this decision. Everything after it is a method for pricing each risk in your own numbers.

The mechanism behind the freelancer risk is that one person is a single point of failure with a nervous system. They cannot cover their own illness, they cannot be in two client crises at once, and a strong month for their business is a thin month for yours. Nothing about this is a character flaw. It is arithmetic about how many hours exist.

The mechanism behind the agency risk is that writers are allocated, and allocation changes. Somebody is promoted, somebody leaves, a larger account arrives and the staffing shifts underneath you. The output keeps arriving, which is exactly why the change is easy to miss until the drafts have stopped sounding like you.

The same six disruptions, seen from each operating model.

Pricing both risks in posts per year

Both risks convert into a single unit, which is posts you did not publish or published below standard. The figures below are assumptions rather than findings, and you should replace each one with your own before drawing a conclusion. The point is the method, not my numbers.

Start with a cadence of two posts a week, which is 104 posts a year. That cadence is already unusual. Around 3% of members post more than once a week, according to aggregate 2026 LinkedIn statistics reports, so a dark fortnight is more visible in this cohort than it would be anywhere else on the platform.

Illustrative arithmetic. Replace every assumption in the first column with your own.

Eight against five looks like a verdict, and it is not one, because the two failures are not the same kind of failure. A missing post costs reach, which is recoverable next week. A published post that does not sound like you costs credibility with the small number of readers who actually know you, and those are the readers the whole exercise exists to reach.

Now run the same table with cover terms in place. A freelancer with a named substitute editor drops the outage from four weeks to about one. An agency with a named writer locked for twelve months drops the reassignment to zero for that period. Both terms are ordinary to ask for, and both are far easier to obtain before signing than after.

The Personeur Handover Test

Ask these five questions on the first call, in this order, of either kind of supplier. Each one names a specific moment of disruption rather than asking about process in general, which is why they produce concrete answers instead of reassurance.

Score it simply. Five concrete answers means you are talking to somebody who has run this before. Three or fewer means you are the experiment, which is sometimes acceptable at the right price, provided you know it.

What the senior person who sold you is actually for

The senior fade is a real pattern and it is rarely a deliberate switch. The person who sold you is usually the person with the judgment, and judgment is the scarcest resource in any writing business. The useful question is not whether they attend every call. It is what they remain contractually attached to.

A senior person can hold three distinct jobs on your account, and buyers tend to assume they hold all three when in practice they hold one. The positioning owner decides what you are arguing this quarter. The editorial reviewer reads every draft before you do. The escalation contact appears when something has gone wrong. Ask which of the three, and ask for the name.

A freelancer has no senior fade because there is nobody available to fade. That is the model's advantage and its entire risk in one sentence: the person holding the judgment is also the person who might have the flu.

Ask which job the senior person keeps in month six. The answer is either a name and a role, or it is a description of the team.The test that separates the two answers

Who holds the voice profile when it ends

This is the most valuable term in either agreement and almost nobody raises it. A mature engagement produces assets worth more than the posts themselves, and whoever holds those assets controls the cost of your next engagement. The posts are the output. The assets are the machine that made them.

The switching cost follows directly from that list. If you do not hold those assets, changing supplier means repeating onboarding. Assume onboarding is one month of interviews plus three drafts that land below standard. Two supplier changes in three years is therefore two months of rebuilding, which is a real number you can compare against any price difference on the table.

The two models fail here in opposite ways. Agencies more often have these assets documented, and more often keep them, because documentation is how a business survives staff turnover. Freelancers more often hold them in their head, which is a different problem: there may be nothing to hand over even in perfect good faith. Ask for the assets to be delivered quarterly rather than at the end, which removes the question from the exit conversation entirely.

Which one fits your situation

The choice is decided by which disruption you can least afford, not by which model is better. Find the row that describes you, then treat the third column as the term that makes that choice safe rather than merely cheaper.

A decision table by situation, with the contract term that de-risks each choice.

If two rows describe you, take the one whose failure would be hardest to explain to your own board. That is usually a better tie-breaker than preference, because it forces the question of which risk you would have to defend rather than which arrangement you would enjoy. If the trial row is yours, the design of that trial matters more than the choice of supplier, and a trial period with a defined kill point is a different piece of work from a discounted first month.

The hybrid most buyers actually want, and why it is rare

Asked to describe the ideal arrangement, most buyers describe the same thing: a senior editor who owns the positioning, a named writer who does the drafting, and a substitute who has already read the archive. That is agency structure at freelance intimacy. It exists, and it is priced as senior time rather than as production.

The reason it is uncommon is arithmetic rather than unwillingness. A senior editor who holds four accounts properly cannot hold forty, because reading every draft and running the positioning takes hours that do not compress. Any studio offering that shape at production pricing is either subsidising it from elsewhere or describing something else.

You can detect the real version in one question: how many accounts does the senior editor personally read for, this month. A number under about eight is credible. A number in the dozens means the editorial review is a spot check, which may be fine, but you should be paying production pricing for it rather than senior pricing.

What neither model fixes

Neither structure solves the input problem, and input is the variable with the largest effect on the result. An engagement that receives sixty recorded minutes of your thinking each month will outperform one that receives twenty, whatever the invoice says at the top. Choose the structure that makes your own time most likely to be given, because that is the one variable you control completely.

Neither model settles the question of whether an executive should be publishing under their own name at all, which is a separate matter with its own arguments and a better answer than the usual defensiveness. That question is worth reading on whether ghostwriting is dishonest before you sign either kind of agreement, because the answer changes what you ask a supplier to do.