How Much Does a Marketing Agency Cost? The Fee Models That Decide Whose Interests Win

The model decides the behaviour

Almost every marketing agency is paid under one of four arrangements: a retainer for reserved capacity, a percentage of the media budget, a fixed price for a defined project, or a fee linked to results. Buyers spend their attention on the number. The number decides what the relationship costs this month. The model decides what it is rewarded for in every month that follows.

That distinction matters because agencies are subject to ordinary commercial incentives, and nobody has to behave badly for a fee model to shape the advice it produces. A structure that pays more when the budget grows will, across many clients and many years, produce more recommendations to grow the budget. The pressure is gentle, constant, and mostly invisible to the person receiving the advice.

The fee model is signed in week one, before the buyer has seen what it rewards, and it then governs the advice for years.

None of the four models is dishonest and none of them is neutral. Each places its cost where the buyer is unlikely to look: in unused capacity, in inflated budgets, in work that stops too early, or in the measurement that decides whether a bonus is owed.

What an agency fee actually buys

The confusion in most quotes comes from failing to say what the fee is payment for. A fee buys capacity, output, outcome, or access to media buying.

• Capacity. The agency reserves availability for the client across the month. The client pays for the reservation whether or not the month was busy.
• Output. The agency delivers defined items: a strategy, a website, a campaign. The price stands whether they took two weeks or six.
• Outcome. The agency is paid according to what the work produced, so the client pays for whatever the measurement system counts as a result.
• Access to buying. In advertising work, part of the fee covers the operation of media budgets, priced by reference to the budget itself and not to the effort involved.

Media budgets are a separate matter from the fee that manages them, and adding the two together into one figure makes the agency fee look small or the total look large. They should be listed separately on every quote. What media itself costs is covered in the inefficient channels wasting your budget.

The retainer and the price of reserved capacity

A retainer is a recurring fee that reserves agency capacity for a client. Buyers understand it least accurately, because the word suggests a subscription to a quantity of work and it is closer to a standing reservation.

The consequence causes most retainer disputes. A quiet month still produces a full invoice. To the buyer this reads as being charged for nothing, and to the supplier the capacity was held and declined to others. Both readings are correct, which is why the retainer works well for steady demand and badly for needs that arrive in bursts.

The more consequential detail is how the retainer is measured. Where it is priced and reported in hours, the arrangement rewards filling the hours, because a task completed in half the expected time consumes half the budget. No supplier intends to work slowly. The structure removes the reward for working quickly, and over a long relationship that shows up as activity reports that grow while results stay flat. Four details decide whether a retainer is a fair one.

• What the fee reserves. A defined scope of responsibility is a better answer than a quantity of hours, because responsibility is measured by whether something is working.
• The treatment of unused capacity. Some arrangements allow a carry-over and many do not. Either is defensible provided it is stated in the contract and not discovered later.
• The rate above the scope. This is the rate the client will actually pay in every busy month.
• What the reporting counts. Reports built around tasks completed describe the supplier's month. Reports built around enquiries and cost per customer describe the client's.

Continuous programmes are normally arranged this way, including search visibility and paid media management.

Percentage of media spend and the incentive it creates

The percentage model charges a share of the advertising budget the agency operates. It descends from the commission that funded advertising agencies for most of the twentieth century, when a media owner paid the agency a standardised fifteen per cent on the money placed with it. Creative and planning work carried no separate charge, so the buyer saw the service as free while paying for it inside the media rate. Management fees today commonly sit in the region of ten to twenty per cent of spend at smaller budgets, tapering into low single figures at very large ones.

The mechanism is rarely stated aloud. The fee rises when the budget rises, whether or not the additional budget earns anything. Doubling a monthly advertising budget doubles the fee even if the second half of the spend produces no additional customers. Nothing about that is fraudulent. It is a payment structure in which the supplier's revenue and the client's efficiency point in different directions. Once that is visible, several behaviours buyers usually attribute to incompetence can be recognised as ordinary commercial incentives at work.

Under a percentage of spend, the only recommendation that costs the agency money is the recommendation to spend less.

• Reluctance to pause spend. An underperforming campaign attracts suggestions to adjust, test and refine, and not to stop. Every one of those suggestions keeps the budget live.
• Broad targeting left running. Wide audiences absorb budget reliably and are difficult to prove wrong quickly. The default settings that do this are covered in the settings spending your budget.
• Platform-reported conversions accepted without challenge. A fee tied to spend has no reason to interrogate a number that justifies the spend, which is the subject of why every platform claims the same sale.
• Silence about cheaper alternatives. Owned channels, email and search visibility reduce the media budget. They also reduce the fee, so they enter the conversation late.

The buyer's protections are structural. A percentage that falls as spend increases removes most of the pressure at the top of the range, and a floor with a ceiling bounds an otherwise open-ended arrangement. A written statement that the agency receives nothing from media owners in cash, rebates or trading credits settles the remaining question. The ISBA and PwC studies of the programmatic supply chain found advertiser money that could not be attributed to any identified party, at fifteen per cent of spend in 2020 and three per cent when the work was repeated in 2023. That improvement followed advertisers demanding the audit, which is the argument for asking in writing.

Project fees and the point where work stops

A project fee buys a defined piece of work for a fixed price. It is the cleanest model to compare between suppliers, because the deliverable is written down and two quotes for it can be placed next to each other.

What a project fee rewards is completion. The commercial event that matters is the handover, and everything after it is unpaid. The relationship therefore ends at precisely the moment optimisation would begin. A website launches, and the first performance data arrives two months later with nobody contracted to act on it. That data is the only evidence of how real customers respond, and discarding it is both normal and avoidable.

A sensible post-project arrangement has four elements, all of them cheap to agree in advance and expensive to negotiate afterwards.

• A defined review point. A fixed date some weeks after launch, priced into the project, at which performance is examined against what was expected.
• An optimisation allowance. Work reserved for the changes the data recommends, so that acting on evidence does not require a fresh commercial conversation.
• Ownership settled in writing. Files, accounts, source code and analytics named as the client's property, so that any future supplier can continue the work.
• A documented handover. Enough written explanation that the business can operate what it has bought without further paid support.

Where the endpoint is imaginary, this model produces a series of small projects with a negotiation in front of each, which costs more than a retainer would have. Production work has its own version of the problem, examined in why the shoot is the cheapest part.

Performance fees and the attribution underneath them

A performance fee pays the agency according to results: a cost per acquisition, a share of revenue, a bonus above a baseline, or a reduced retainer with an upside attached. Buyers find it the most reassuring model, because it appears to remove the risk and to align both parties perfectly.

It aligns them to the measurement, which is a different thing. A performance fee is only as honest as the attribution model underneath it, and that model is normally supplied, configured and reported by the party being paid. Which channel deserves credit for a sale has no objective answer, and whoever defines the rule also defines the invoice.

• Attribution window length. A thirty-day window and a seven-day window credit very different numbers of sales to the same campaign, and the choice is made once.
• Existing demand counted as new. Customers who searched for the business by name appear in platform reports as conversions the advertising produced.
• Quality outside the definition. A cost per lead target is met by producing leads. Whether they were qualified is a separate measurement the fee has no reason to fund.
• The lag problem. Brand building, public relations and search visibility pay back over quarters. A fee measured monthly moves effort towards whatever pays back inside the month.

None of this makes performance pricing unfair. It makes the measurement layer the thing to negotiate, and independent analytics held in accounts the client owns converts the deal into a verifiable arrangement, which is the purpose of separate analytics and performance reporting.

One further point is usually missed. A performance fee transfers risk to the agency, and risk is not transferred for free. A supplier carrying it will price for the case where the work succeeds, so a successful performance arrangement can cost more in total than a retainer would have. That is a fair trade for a business that wanted certainty, and it is not a discount.

Why quoted ranges span more than an order of magnitude

Published guidance on agency costs presents ranges so wide that they carry almost no information, with the same service appearing at a figure and at more than ten times that figure in one paragraph. The spread is real and it has structural causes.

One service name covers different products, so a single label can describe a monthly report, a small ongoing programme, or a full operating function for a national business. Scope also goes unstated: two quotes for the same named service can differ on strategy, production, implementation and reporting, any of which may sit inside or outside the price. Advertising budget is sometimes inside the number and sometimes outside it, which makes two figures incomparable while they look identical. The least discussed cause is that most published pricing guidance is written by suppliers, and a range that comfortably includes the writer's own price is the range that gets published.

The conclusion is simple. A headline figure without a defined scope cannot be compared with a competing quote, and it cannot be compared with the same supplier's figure six months later. The one figure that does compare is the annual total fee expressed as a percentage of the money the agency will direct.

Making four quotes comparable

Any of the four models can be converted into that annual figure. Five questions do almost all of the work, and they are reasonable to ask of every supplier, including us.

1. What is included. Ask for the scope as a written list of responsibilities, so that strategy, production, implementation and reporting each appear on one side of the line.
2. What is billed separately. Establish which costs arrive on later invoices: media budget, software licences, photography and video, translation, and revisions beyond an agreed number.
3. What happens at the ceiling. Ask what the fee becomes when spend doubles, when the retainer scope is exceeded, or when the project needs another round.
4. Who owns the output. Accounts, data, source files and analytics history should belong to the client from the first day. This answer determines what the arrangement costs to leave.
5. What the minimum term is. A twelve-month minimum and a monthly rolling term are different products at the same headline price.

The conversion is short arithmetic, and the figures below are illustrative. A retainer becomes twelve monthly fees plus the over-scope rate multiplied by the months that exceed the scope. A percentage of spend becomes the rate applied to the annual media budget the agency will direct, so twelve per cent on ten thousand a month is fourteen thousand four hundred a year. A project fee becomes the fee plus the review point and the optimisation allowance, divided across the period until the next project. A performance deal becomes any base fee plus the agreed rate applied to the annual result volume the business expects, and not the volume the proposal assumes. Two quotes that looked incomparable then resolve into a difference of scope, and the cheaper one usually excluded more.

Most agency quotes differ in what they exclude rather than in what they charge.

The question that reveals the proposal

One question does more than the five above combined. The question to put to any agency at proposal stage is what happens to its fee if the recommended activity is reduced.

The answer describes the commercial relationship precisely. A supplier whose fee falls when advertising spend falls has an interest in the spend, and a supplier whose fee is unchanged has an interest in the outcome. Neither answer is disqualifying, because a fee that ignores workload will eventually ignore the work. What matters is that the answer arrives without discomfort.

The follow-up is equally short, and it asks the supplier to name something it would advise this business not to buy. A proposal that recommends every service the supplier happens to sell is answering a question about the supplier. A proposal that names an exclusion, and explains the reasoning, has done actual thinking. Our own position is set out on how we work.

Reluctance here is informative, and it should be read calmly. Most agencies have never been asked and need a moment to answer honestly, which is different from evasion.

Where each model is fairest

The retainer is fairest for continuous work that compounds, such as search visibility and content, because a project fee cannot buy continuity. A percentage of spend is fairest at large and variable media budgets, provided it tapers and carries a floor and a ceiling. A project fee is fairest where the work has a real endpoint, including positioning, identity and a website build. A performance fee is fairest where attribution is clean, meaning a short purchase cycle and independent measurement the client owns.

A hybrid suits most real situations, because a modest base fee plus an element linked to either spend or performance distributes risk without asking either party to carry all of it. The instruction is to match the model to the shape of the work rather than to a preference about fairness.

Doing it internally versus hiring an agency

The point at which an agency earns its fee is a question of arithmetic. It arrives when the cost of the gap exceeds the cost of closing it, and the gap is usually one of four things: money spent inefficiently at a scale where a percentage improvement exceeds the fee, a capability the business cannot practically acquire, a deadline internal capacity cannot meet, or an expensive mistake being repeated because nobody internally can see it.

Below that point, several things a business can do for itself would otherwise appear as billable lines. Customer research is the clearest, because how buyers describe their own problem is the most valuable input in marketing, and gathering it directly costs time instead of a discovery fee. Listings, reviews and customer email are operational work that agencies charge a management fee to run, and written answers to genuine enquiries produce content that would otherwise be commissioned. The tactical detail is covered in the free online guides that fail.

The test can be run on paper before anyone is hired. Estimate what a realistic improvement would be worth over a year, and compare it with the annual fee calculated using the five questions above. Where the answer is close, the honest recommendation is to wait.

Key takeaways

• Four models cover almost every agency arrangement, and the model decides the behaviour more than the number does.
• A retainer buys reserved capacity, so a quiet month is still a full invoice, and hourly measurement rewards filling the hours.
• A percentage of media spend rises with the budget whether or not the extra budget earns anything, which explains reluctance to pause and unchallenged platform reporting.
• A project fee rewards finishing, and ends where optimisation would begin, unless a review is priced in.
• A performance fee is only as honest as the attribution beneath it, which the party being paid normally supplies.
• Five questions make any quote comparable: inclusions, separate billing, the ceiling, ownership, and minimum term.

There is no correct answer to what a marketing agency costs, and there is a correct way to ask. A buyer who writes down the scope and converts every quote to an annual figure will pay a fair price under any of the four models. A buyer who compares headline numbers will overpay under all of them.

For a written scope with the fee model stated and the reasoning explained, request a proposal and Reachford will set out what sits inside the price, what does not, and which model fits the work described.