The short answer
A paid pilot is the better first step when the question is whether this agency can produce something a UK practice can read and accept, because that answer arrives inside a few weeks. A retainer is better where the only question worth asking has a measurement window longer than any pilot. Scoping a pilot to prove an outcome it cannot show is the costly mistake.
Key takeaways
- A pilot and a retainer are different purchases, not one purchase at two lengths. One ends in a thing, the other in a month.
- A pilot settles a question only where the answer appears inside its own window and somebody other than the agency can see it.
- The client acceptance standard a practice already applies asks whether a supplier has adequate resources and expertise for the work. A first engagement is that test, reversed.
- A pilot billed as a cheaper month of the retainer has bought the retainer, because nothing exists to accept or refuse.
- Where the real question has a longer window than any pilot, the honest answer is a retainer with a named review date, not a pilot that cannot fail.
Before an accountancy practice takes on a new client, ICAEW asks it to be able to demonstrate four things. Two of them read differently when you turn them around: that the firm has adequate resources and expertise to provide the services required, and that the terms of engagement are clear and unequivocal. Every practice applies that test to the clients it accepts. Very few apply it to the supplier they are about to pay.
The pilot question is usually argued as a question about money, which is why the published answers disappoint: they compare cash flow for the agency rather than evidence for the buyer. The useful version is narrower. A first engagement is a test, so what matters is what this firm needs proved, and whether the shape under consideration can prove it before the money runs out.
A pilot and a retainer buy different things
A pilot buys a deliverable somebody can inspect. A retainer buys capacity somebody at the firm has to direct. That is what decides which one can answer a question, and it is not a difference of length or price.
A pilot ends in an object: a page that exists, or a document a partner can read on a train. It has an acceptance point, which means it can be refused. A retainer ends in a month, and a month cannot be refused. Continuous work belongs in a retainer, and most marketing is continuous, which is why agencies prefer one and why that preference is not a warning sign.
Scope the pilot to what the pilot can show
A pilot settles a question only where the answer appears inside its own window and somebody other than the agency can confirm it. Those two tests rule out more than most UK firms expect.
| What the firm wants proved | Visible inside eight weeks | Who can confirm it apart from the agency | The shape that tests it |
|---|---|---|---|
| Whether this team is workable | Yes, usually inside a fortnight | Your own staff, from how the first requests are handled | Pilot |
| Whether the work itself is any good | Yes, if the engagement ends in something readable | A partner outside the project, or a peer elsewhere | Pilot |
| Whether they understand a regulated practice | Yes, from the first draft of anything client facing | Whoever signs off compliance | Pilot |
| Whether enquiries of the right kind increase | No, for almost every channel | Your own enquiry log, read over a longer period | Retainer with a named review date |
| Whether the agency is still attentive in month nine | No, and nothing tests it in advance | Nobody, before the fact | Neither. This is what notice and break clauses are for |
The test your own profession already runs
ICAEW asks a firm accepting a client to demonstrate that it has adequate resources and expertise to provide the services required, and that the terms of engagement are clear and unequivocal. Those conditions sit in the Practice Assurance standard on client acceptance, which also treats an engagement letter as best practice on appointment, and says clarity about an engagement's scope and complexity is what avoids misunderstandings later.
Two of the institute's helpsheets point the same way. Its guidance on fee information records that a firm must tell every client in writing how its fees are calculated. Its guidance on not having issued an engagement letter is blunter about why: a clearly scoped letter is the practice's defence when a client later says it expected work nobody agreed to.
An authorised firm has the same discipline imposed from outside. The Financial Conduct Authority's client agreements chapter requires the terms of an agreement for designated investment business to reach a retail client in good time, before that client is bound or before the services are provided, whichever comes first.
So the mirror is worth stating plainly. A pilot is the buyer's version of an engagement letter: a defined scope with a stated basis for the fee, and an end somebody can judge. None of these rules binds a marketing agency. They bind the practice towards its own clients, which is the point: it is a standard the reader already knows how to apply. No agency retainer is improper for failing to meet it.
Four pilots that prove nothing
A pilot that cannot come out as a no is not a test, it is a discount. Each of these is bought in good faith.
A cheaper month of the retainer, same scope, called a trial.
Proves that a month of the retainer feels like a month of the retainer. Nothing is delivered, so nothing can be accepted or refused.
An audit with no decision attached to it.
Proves the agency can write a document, which was never the question. An audit the firm was always going to be given is part of the sale, not of the delivery.
A six week search pilot judged on position movement.
Proves nothing either way: the window is shorter than the channel takes to respond, so any result is noise read as signal.
A pilot the agency scoped and then reported on itself.
Proves the agency can produce a favourable report. The reporting does not have to be dishonest for this to be worthless, only unexamined.
Before the first invoice
Not sure a pilot would prove what your firm needs proved?
Free to your firm. The agency pays us, and only if the relationship works, which is why we would rather talk you out of an engagement nobody could judge than place one.
Get a recommendationWhen a retainer is the better first step
Where the only question that matters has a window longer than any pilot, a pilot is a more expensive way to learn less. A retainer entered deliberately is not the weaker option. It wants three things a pilot gets for free: a scope that names units rather than service categories, a review date fixed in the document, and a notice period the practice can live through. The term is where this is lost, and how long a first agency contract should be is a separate decision from how good the agency is.
When not to start either one
Where the firm cannot supply its own half of the work, both shapes bill and neither produces evidence.
Two situations are worth waiting out. A practice that cannot name the person who accepts work on its behalf has no acceptance point to offer, so a pilot degrades into a retainer on the first disagreement. And a UK practice inside its busiest filing weeks cannot answer a supplier inside a day, which moves every date in the engagement. What the first weeks demand from the firm is set out in what an agency needs from you in the first month.
We are paid by the agency, so a page that persuades a partner group to postpone an engagement costs us money. What we look at before recommending anybody is set out for accountancy practices, recruitment agencies, tax advisory and financial advice firms.
Frequently asked questions
What partners ask before a first engagement is signed.
How long should a paid pilot be?
Long enough for the deliverable to exist and be read, which for a UK practice buying a first piece of work is usually four to eight weeks. A pilot with no deliverable has no natural length, which is why it settles at three months: that is what a quarter looks like.
Should a pilot be cheaper per month than the retainer?
Expect it not to be. A defined piece of work carries setup that a retainer spreads over a year, so a pilot priced well below the retainer is usually the retainer with the deliverable removed.
What has to be written down before a pilot starts?
The deliverable, the date it is due, who accepts it, and what happens to the work if the engagement stops there. The last one is the clause firms forget, and it decides whether a failed pilot leaves anything behind.
Is it reasonable to ask an agency for a pilot?
Yes, and the answer is itself informative. An agency confident in its work usually welcomes a scoped first piece, because it shortens its own sales cycle. Hear a refusal in full before holding it against anybody.
Does a pilot protect a firm legally in a way a retainer does not?
Not by itself. What protects a firm is a written acceptance point, and a pilot makes one cheap to agree because the deliverable is already defined. A retainer can carry the same clause, but somebody has to draft it.
Sources and useful reading
- ICAEW, Practice Assurance guidance for sole practitioners and new or smaller firms, Standard 2, for the conditions a firm should demonstrate on accepting a client.
- Fee information helpsheet, for telling clients in writing the basis on which fees are calculated.
- What if you have not issued an engagement letter?, for the scoped letter as a defence where a client expected work nobody agreed to.
- Financial Conduct Authority Handbook, COBS 8.1, Client agreements, for the duty to give a retail client the terms of the agreement in good time.
This article is commercial decision support and is not legal advice. The guidance cited governs its own members and authorised firms in dealings with their own clients. None of it binds a marketing or web agency, and nothing here says an agency retainer is improper. The eight week window is this page's own scoping test rather than a published finding. The four sources were located on 1 October 2026, and nothing here describes any named agency or engagement.