There are two fundamentally different retainer types and confusing them causes most retainer problems. A capacity retainer sells an agreed volume of work per month - hours or deliverables - and its risk is silent overage. An outcome retainer sells ongoing responsibility for a result regardless of hours, and its risk is unbounded effort. Whichever you sell, four things must be written down before the first invoice: what the allocation actually is, whether unused capacity rolls over, what happens when it is exceeded, and how either side ends it. Retainers fail from ambiguity far more often than from bad pricing.
A retainer is the most commercially attractive thing an agency can sell. Recurring revenue, predictable capacity, a lower cost of sale, and a client relationship measured in years rather than projects.
It is also the easiest arrangement to run badly, and the failure is quiet. Nobody notices a retainer going wrong in month two. They notice in month nine, when the account has become the one everyone dreads, the margin has vanished, and the client genuinely believes they are getting what they bought.
This guide covers the two retainer types, how to structure each, the four clauses that prevent almost every dispute, how to price them, and how to tell when a retainer has quietly gone bad.
The two types, and why the distinction matters
Capacity retainers
The client buys an agreed volume of work per month. "40 hours of design support." "Four articles and two email campaigns." "Up to 60 hours of development."
The unit of sale is your capacity. What gets done inside it is flexible; how much gets done is not.
Best for: ongoing support work where the client's needs vary month to month but the total volume is roughly stable. Design support, development maintenance, content production.
The risk: silent overage. This is covered in detail below because it is the single most common way retainers fail.
Outcome retainers
The client buys ongoing responsibility for a result. "You own our SEO." "You keep the platform running and improving." The hours are your problem, and if you find a way to deliver the outcome in half the time, you keep the difference.
Best for: work where the outcome is measurable, you have delivered it before, and you have enough control to be genuinely accountable for it.
The risk: unbounded effort. Without a scope boundary, "responsibility for the outcome" expands to whatever the client thinks is needed to achieve it - which, in a bad month, is everything.
Why conflating them is expensive
Most retainer disputes trace back to the two sides holding different models in their heads. The agency sold 40 hours. The client bought a partner who handles design. Both descriptions were used in the sales conversation, and nobody noticed they were different arrangements.
Then the client sends a request in week four that would take 15 hours. Under the capacity model, that is next month's work. Under the outcome model, it is simply the job. Neither party is being unreasonable; they are answering different questions.
Fix this by naming the model explicitly in the first line of the agreement. Not "monthly retainer" - "capacity retainer: 40 hours per calendar month" or "outcome retainer: ongoing responsibility for X, scoped as follows."
The four clauses that prevent almost every dispute
Teamwork's overview of agency pricing models and Scoro's guide both cover retainer mechanics. The four items below are the ones most contracts leave implicit, and each one has a predictable failure attached.
1. What the allocation actually is
Be specific enough that both sides could count it independently.
Weak: "ongoing design support." Strong: "up to 40 hours per calendar month of design and production work, tracked and reported weekly."
If you sell deliverables rather than hours, define the deliverable tightly - "four articles of up to 1,500 words, including two rounds of revisions each" rather than "four articles." The revision rounds are where deliverable-based retainers leak.
2. Whether unused capacity rolls over
This clause is skipped more often than any other, and its absence causes the ugliest arguments.
If you say nothing, the client will reasonably assume unused hours accumulate. Three quiet months later they will expect 120 hours in one month, which will break your capacity plan and force you to either refuse - looking like you are reneging - or deliver it at a catastrophic margin.
Three workable answers, all fine, none of them silence:
- No rollover. Cleanest. The client is buying reserved capacity, and reserved capacity has a cost whether used or not. Say exactly that, because it is true and it is defensible.
- Capped rollover. Unused hours carry to the next month only, then expire. A reasonable middle ground.
- Rollover into a bank with a ceiling. Accumulates up to, say, 1.5× the monthly allocation, and drawing on it needs two weeks' notice so you can staff it.
3. What happens at overage
Two acceptable answers:
- Stop and ask. Work pauses at the allocation, and further work needs written approval as a change order. Protects margin, occasionally frustrates clients.
- Bill the overage. Work continues at a stated overage rate. Smoother, but only if the client sees burn before the invoice arrives.
The unacceptable answer is the default one: absorb it and hope. That is not a policy, it is a slow transfer of your margin to the client, and it teaches them that the allocation is notional.
4. How it ends
Retainers end. Write down the notice period - 30 or 60 days is standard - what happens to work in progress, and who owns what on exit. An agency without an exit clause is one difficult conversation away from either working a month for free or ending a relationship badly enough to lose the referral.
Our guide to client offboarding covers the operational side of ending well.
Pricing a retainer
Start from your real cost floor. A retainer is the model where an underpriced rate does the most damage, because you have committed to repeating it every month for a year.
Step 1: Price the capacity at your standard rate. 40 hours at your true blended rate. Our guide to pricing agency services covers deriving that rate from genuine available hours rather than a notional 40-hour week.
Step 2: Decide the discount, deliberately. Retainers usually carry one, and there are two honest justifications: lower sales cost, and easier capacity planning because the work is predictable. Both are real. 10-15% is common and defensible.
What is not defensible is a discount you cannot explain. If the number came from wanting to win the deal, it is not a retainer discount, it is a rate cut that now recurs monthly.
Step 3: Check the floor. Discounted rate still above break-even cost with a real margin? If not, the retainer will lose money every single month, reliably, which is worse than a bad project.
Step 4: Set the overage rate above the retainer rate. If overage is billed at the same discounted rate, the client has no reason to stay within the allocation - the discount was for predictability, and overage is by definition unpredictable. Standard rate for overage is normal and easy to justify.
Managing a retainer so it does not rot
Pricing is the easy half. Retainers go wrong operationally, and always the same way.
Track burn where both sides can see it
The single highest-leverage practice. A retainer only works as a commercial instrument if both sides can see how much of it is left, mid-month, without asking.
A client who can see they are at 80% on the 18th self-regulates. The same client, told on the 3rd of the following month that they went 30% over, experiences a surprise bill - and they are right to be annoyed, because the information existed and nobody showed them.
This is why a client portal matters more for retainers than for projects. The status question on a retainer is not "is it done", it is "how much have we used", and that is a question a portal answers continuously.
Review quarterly, not annually
Retainers drift. Work that was 40 hours a month in January is 55 by June because the account grew, and nobody re-priced. A short quarterly review - actual hours against allocation, what changed, whether the allocation is still right - catches drift while it is a conversation rather than a renegotiation.
Watch for the three rot signals
Consistent overage. Three months over allocation is not a busy patch, it is a mispriced retainer. Re-price it.
Consistent underuse. Also a problem. A client using 40% of their allocation is deciding whether to cancel, whether or not they have said so. Underuse is the leading indicator of churn, and the response is to proactively surface unused capacity and propose work - not to quietly enjoy the margin.
Scope migration. The work is within the hours but no longer what was scoped. A design retainer that has become 60% project management is still "on budget" and is no longer the engagement you priced, staffed, or wanted.
Report without being asked
Most retainer clients cannot easily articulate what they got for the money, which makes renewal a matter of feeling rather than evidence. A monthly summary - hours used, work delivered, what is queued - answers the question before it is asked and makes renewal a formality.
The trap is that building these by hand is a real part-time job across a dozen accounts, which is exactly why it stops happening in month four. It has to be a by-product of the work rather than a deliverable in its own right.
Selling the first retainer
Most agencies wait for the client to ask, which means most retainers never happen. The conversion is a conversation you initiate, and timing matters more than pitch.
The moment to raise it is immediately after a successful delivery, while the relationship is at its high point and the client is actively thinking about what comes next. Not three months later, when momentum has gone and you are effectively cold-selling to someone who already knows you.
Frame it as continuity, not a new purchase. "The site is live. The things that will matter over the next six months are the iterations - the pages that need testing, the content that needs building, the fixes nobody can predict. We can handle those ad hoc, or we can reserve capacity so they get done in days rather than whenever there's a gap." That is a real choice between two real options, not a pitch.
Price the alternative honestly. Ad-hoc work is genuinely more expensive per hour and genuinely slower, because it has to be squeezed between committed projects. Saying so is not a sales tactic, it is the actual economics, and clients respect hearing it plainly.
Start smaller than you want. A 20-hour retainer that consistently runs at capacity and grows to 40 is a far better outcome than a 40-hour retainer that runs at 50% and gets cancelled at renewal. Underuse is the strongest churn predictor there is.
Converting a project client to a retainer
The mechanics that make conversion work:
- Do it inside the project's final phase, not after the invoice. There is a natural conversation about what happens next, and it is much easier to have while you are still in the room.
- Bring evidence. "Over the last four months you sent 38 requests outside the original scope. That's roughly 22 hours a month." Data from the project makes the allocation obvious rather than arbitrary - and if you have been tracking time properly, you already have it.
- Name the first month's work. A retainer starting with an empty queue feels like a subscription to nothing. Walking in with a list of things you already know need doing makes month one concrete.
- Agree the reporting rhythm at the start. When they will see burn, when they will get a summary. Set once, this removes most of the friction that appears in month three.
The metrics that tell you a retainer is healthy
Four numbers, reviewed monthly, catch almost everything:
Utilisation of the allocation. Hours used ÷ hours sold. Healthy is 85-100%. Consistently over means mispriced; consistently under means churn risk.
Effective hourly rate. Retainer fee ÷ hours actually delivered. This is the number that reveals silent overage - a £4,000 retainer for 40 hours is £100/hr on paper and £67/hr if you actually delivered 60. Track it monthly and the erosion becomes visible long before it becomes a crisis.
Scope composition. Roughly what proportion of hours went to the work you scoped versus something else. Drift shows up here first.
Response and turnaround time. The thing retainer clients are actually buying is responsiveness. If it is slipping, renewal is at risk regardless of how good the output is.
None of these require a reporting project. They fall out of tracked time, which is the argument for getting time tracking reliable before you build a retainer book on top of it.
Retainer vs project: when to push each
Not everything should be a retainer, and pushing one on the wrong client damages both sides.
Retainer fits when: the need is genuinely ongoing, the volume is roughly stable, the client values responsiveness, and there is enough work to justify reserved capacity.
Project fits when: there is a defined outcome with an end, the work is lumpy, or the client is new enough that neither of you can predict the working relationship.
The most common mistake is converting a client to a retainer too early. A retainer is a commitment to reserve capacity for someone whose behaviour you have not yet observed. Run one project first. You will learn how they approve, how they scope, and how they communicate - and all three of those determine whether the retainer will be profitable.
Staffing a retainer book
Retainers change how you staff, and agencies that treat them as "projects that repeat" run into the same two problems.
Reserved capacity has to be genuinely reserved. If a retainer sells 40 hours a month and those hours are not held in the capacity plan, they will be consumed by whichever project is loudest that week - and the retainer client, who is paying for responsiveness, gets the leftovers. Block the hours before the month starts.
Continuity matters more than efficiency. Rotating people through a retainer to fill gaps looks efficient on a resourcing board and is expensive in practice: every rotation costs context, and the client notices immediately because they have to re-explain things. One named person with a named backup beats a pool, even if the pool utilises better on paper.
The related failure is the single point of dependency - one person who holds the entire relationship and all the context. That is fine until they take leave, and it is a genuine business risk if they resign. The cheapest insurance is a documented account brief and a second person who joins the monthly call, doing nothing, just staying current.
What to do when a retainer has already gone bad
Most agencies read a guide like this while running two or three retainers that are already unprofitable. Fixing an existing arrangement is different from structuring a new one, and it is very doable.
Start by measuring, not negotiating. Pull three months of actual hours against the allocation. You need the real number before any conversation, because "it feels like we're doing too much" is not a position and "you're averaging 61 hours against a 40-hour retainer" is.
Open with the data and no accusation. "I want to show you what we've actually been delivering, because I don't think our agreement reflects it any more." Clients are rarely aware of overage - they are not tracking it either - and most respond reasonably to evidence.
Offer two paths, both fine for you. Increase the allocation and the fee to match reality, or hold the fee and bring the work back inside the original allocation. Both are legitimate. Presenting them as a genuine choice avoids it feeling like a price rise.
Set the new arrangement up with visible burn from day one, so this cannot recur silently.
The conversation is uncomfortable once. Absorbing 50% overage indefinitely is uncomfortable every month, and it eventually ends the relationship anyway - just later, worse, and with more resentment on both sides.
Two retainer structures worth stealing
The tiered retainer. Three levels - say 20, 40 and 80 hours - at descending effective rates. This does two useful things: it gives the client an obvious upgrade path rather than a renegotiation, and it makes the middle tier feel like the sensible choice, which it usually is. It also means a growing account moves up a tier instead of quietly running 60 hours on a 40-hour agreement.
The base-plus-project retainer. A small ongoing allocation covering maintenance and responsiveness, with larger pieces of work quoted separately as fixed-fee projects. This suits clients whose steady-state need is genuinely small but who periodically want something substantial. It avoids the two failure modes at once - the retainer does not have to absorb a big project, and the big project does not have to be squeezed into an allocation designed for maintenance.
The structure to avoid is the unlimited retainer - "as much as you need for £X". It sounds generous, it wins deals, and it has no mechanism for staying profitable. The only agencies for whom it works are those with genuinely productised, tightly-bounded deliverables, and even they usually cap concurrent requests rather than total volume.
A workable default
If you want a starting structure rather than a menu:
- Capacity retainer, hours-based, stated in the first line of the agreement.
- No rollover, explained as reserved capacity - honestly, because that is what it is.
- Overage billed at standard rate, with burn visible to the client throughout the month.
- 30 days' notice either side.
- Quarterly review of actual hours against allocation.
- 10-15% discount against project rates, justified by lower sales cost and predictable capacity.
That covers the four clauses, prices honestly, and makes the two most common failures - silent overage and rollover disputes - structurally difficult rather than merely discouraged.
The agencies with healthy retainer books are not better negotiators. They have simply written down the four things above, and they let the client see the burn.