The retainer was profitable in month one. By month six it was quietly "under review", which is the polite phrase agencies use for the moment the finance meeting stopped being fun. Nothing had changed on paper. Same fee. Same deliverables. And yet the account was quietly losing money every week, and the founder running it didn't notice until the quarterly P&L was already printed.
This post is about how that happens. More usefully, it's about the two costs that cause it. Costs that never appear on any scope document. The monthly measurement that catches them before the P&L does. And the re-scope conversation that saves the account instead of ending it.
The advice that misses
Type "how to prevent a retainer becoming unprofitable" into a search engine and page one is roughly the same everywhere. Track hours per client. Flag the account when time-spent hits 75% of the budget. Have a monthly reconciliation call. Charge for out-of-scope work.
All fine advice. It's aimed at scope creep in the visible sense, the sort where a client asks for an extra landing page and you build one. For that failure mode it works.
What it doesn't catch is the failure mode most retainers actually die from. Which isn't new tasks the scope document didn't cover. It's old tasks the scope document doesn't describe accurately, done more expensively than the fee assumed, in ways that will never show up on a timesheet because nobody thinks to log them.
The two costs doing the killing are response latency and unbilled thinking time. Both invisible to the industry-standard tools. Both compound weekly. Both fixed by the same re-scope conversation script, the one below, which no vendor is going to publish because the vendor sells the dashboard that doesn't see them.
Cost one: response latency
Response latency is what it costs you when a client's questions get answered too quickly.
Every retainer starts with an unspoken assumption about how fast the client can expect a reply. If you never make it explicit, the client sets the default themselves. And the default they set is: whenever they email, whenever they message, whenever they call.
In month one, when there are three other clients and the account is fresh, replying inside the hour isn't expensive. By month six, when there are eleven other clients and this one has grown a habit of pinging you at 4pm on a Wednesday about a screenshot they took at 3:55pm, the same hour-inside response is expensive in a very specific way.
Here's how. It fragments the working day of your most-billable people. A senior producer who could have done four hours of concentrated work in an afternoon can instead do six twenty-minute stretches broken up by two-minute Slack replies. The difference between those two afternoons is the difference between an account that runs on margin and one that runs on goodwill.
The client hasn't asked for anything the scope document doesn't describe. They've asked for a screenshot in a Slack message. The cost of the reply isn't the two minutes it took to send. It's the twenty minutes on either side of it, which is the well-documented cost of interrupted work.
Almost no agency measures this. Not because they don't know it's happening. The producers know exactly when it's happening. The tooling to measure it doesn't exist and, honestly, the client would find the measurement embarrassing. So the cost stays invisible, and the account owner reports on hours logged and deliverables shipped while the margin drips out of a hole nobody has named.
The other thing worth saying about response latency is that the client is not the villain of the story. The client has been trained by the agency's own habits. A reply inside the hour in week two is a promise the client has no reason to disbelieve, and by week eight it is the working assumption on both sides. What looks like a client behaviour is actually a reinforcement schedule the agency set up itself. Which is useful, because it means the fix is on the agency side. You can rewrite the schedule. You can't rewrite the client.
Cost two: unbilled thinking time
Unbilled thinking time is what it costs you to do the work that happens before the work you're billing for.
Every reasonable retainer includes deliverables. A monthly report. A set of creatives. A campaign build. A strategy note. The scope document is written around those. What the scope document doesn't say is that most of them take about half as long to produce as they take to think through, and the thinking-through is what actually eats the account manager's brain.
Take the "monthly performance review" that appears on almost every marketing retainer. The document itself, once you know what you want to say, takes maybe two hours to write. The thinking-through takes six. Reading the last four weeks of dashboards. Running the diagnostics. Working out what actually changed versus what's noise. Framing it in a way the client will trust.
Nobody logs those six hours anywhere. On the timesheet, the account is booked for "monthly report, 2 hours". On the payroll, the account manager has spent Tuesday, Wednesday morning, and part of Thursday morning thinking about this client. The retainer was priced for the timesheet. The account is being delivered by the payroll.
Multiply that gap across ten clients and it isn't a gap anymore. It's the entire margin.
There's a second-order version of this that's worth naming. The account manager who has been spending six unlogged hours a month thinking about the client also becomes the person the client trusts. So the next month, when the client asks for a piece of ad-hoc thinking that wasn't in the scope, they ask that account manager, not the wider team. Which means the six hours becomes eight, then nine. The account gets better outcomes and the retainer gets worse economics. Both things are true at the same time, which is why the trend is easy to miss from the top.
The compound
The two costs aren't additive. They're multiplicative.
A client who has trained you to reply within the hour is also a client whose account manager can't get any thinking time done during the working day. So the thinking-time cost, the six unlogged hours per month, moves to evenings and weekends. Where it's done tired, badly, by people who will eventually resign. The retainer looks fine on paper for another quarter. The turnover in the team is where the cost actually lands.
Meanwhile the client's own perception of value has been quietly recalibrated by the response latency. A retainer whose deliverables ship monthly but whose replies arrive within the hour trains the client to see the retainer as an on-call resource. They stop remembering the deliverables. They start remembering the availability. When it's time to renew, the negotiation isn't about the deliverables you've shipped. It's about the availability they've come to expect. Next month's fee gets priced against the deliverables. Next month's cost gets delivered against the availability.
The monthly measurement that actually catches this
You don't need a timesheet. The people who fill in timesheets have been trained to fill them in against the scope document, so timesheets tell you exactly what the scope document already told you.
What you need is a two-column note per client. Twenty minutes at month-end. Done by whoever owns the account. It looks like this.
Column one: shipped this month. The list of things that came off the account this month. Deliverables, reports, campaigns, strategy notes. Anything the client received. Plain language, one line each.
Column two: everything else this month. Everything the account team did that didn't appear in column one. Every meeting. Every Slack conversation over ten messages long. Every email chain over five replies. Every ad-hoc analysis. Every time somebody redid a piece of work because the first version got sent back. Every 4pm-on-Wednesday screenshot request. Plain language, one line each.
You're not measuring hours. You're counting entries. If column two has more entries than column one, the retainer is running on unbilled thinking time and response latency, and the numbers on the P&L are lagging what's already happened.
Why this works when timesheets don't: it doesn't ask anyone to remember how long something took. It only asks whether it happened. That's a question a human being can honestly answer at month-end. How long something took is a question they can't.
The three warning signs before the P&L catches up
Do the two-column note for three months and the pattern will announce itself before the finance meeting does. The warnings, in the order they show up:
Sign one: the account manager stops proposing anything. In month one, the account team was full of "we should try X". By month six, when the same team is asked what to try next, they say "we should keep doing what we're doing". This isn't agreement with the strategy. It's the operational reality that they don't have any thinking-time left. A retainer whose team has stopped proposing new work is a retainer delivering only from muscle memory, which is the most expensive way to deliver anything.
Sign two: the client's Slack messages start with "quick one". The word "quick" in a client message is almost never accurate. What "quick one" actually means is: this is a small enough thing that I don't feel bad asking, but a specific enough thing that I need a real answer. The client isn't being manipulative. They've simply learned that "quick one" gets a response inside the hour. Once the pattern establishes itself, you've become a shared inbox for their operations without noticing.
Sign three: deliverables start slipping and nobody knows why. The monthly report is a day late, then two, then three. The team isn't idle. The team is exhausted. Nobody is doing less work. They're doing less work that's on the scope document, because everything not on the scope document has to happen first, in real time, at 4pm on a Wednesday.
When you see all three signs on the same account, the P&L reconciliation for that account is going to be ugly. You have between four and eight weeks. Have the conversation now.
The re-scope conversation
The reason this account is losing money isn't that the fee is wrong. It's that the delivery model is wrong. The conversation you need to have is the one that fixes the delivery model. The reason most agencies never have it is they think the conversation is a request for more money. It isn't. It's a request for the account to survive.
Here's the shape of the conversation, in three moves. Each move is one sentence long, in normal working language, said by the account owner and heard by whoever holds the budget on the client side. This isn't a script to be recited. It's the sequence and the specificity.
Move one: name the pattern, not the client's behaviour. The account owner opens with a sentence like: "This retainer was built around monthly deliverables, but the work has moved toward real-time back-and-forth, and I want to reset how we run it before the next quarter, so the delivery model matches how we're actually using each other." Notice what this doesn't say. It doesn't blame the client for asking too much. It doesn't say the retainer is unprofitable. It names the pattern and asks to re-scope. Both neutral facts.
Move two: propose the two structural changes, in a form the client can accept without losing anything. The two changes are almost always the same. First: response windows. Slack replies within one working day rather than one hour, urgent things flagged as urgent, and one weekly office-hours block for real-time questions. Second: strategy time. A named half-day per month for the account team to think through what's next, protected from the ongoing work. Neither change costs the client anything in absolute terms. Both save the account.
Move three: offer the trade. Where a fee increase is warranted, this is where you name it, and you name it in exchange for something specific. A new deliverable, an expanded reporting cadence, a stated ambition. Where a fee increase isn't warranted, this is where you propose the delivery-model reset without a fee change and offer to review the fee in the following quarter if the model has settled. The important thing is that the trade is specific. "We need to raise the fee" is a fight. "We are moving to a model where we can protect strategy time; here is what that adds" is a conversation.
Have the conversation with whoever holds the budget. In person or on video. Never by email. Fifteen minutes on a calendar invite booked two days in advance is enough. If you can't get the meeting, you've already found out something about the account that's worth acting on separately.
What to do if the conversation fails
Some accounts won't accept the re-scope. When that happens you have three options, and the one you pick is a business decision, not an operational one.
Option one: raise the fee to match the delivery model. If the client wants to keep responding in the hour and doesn't want a named half-day for strategy, the price for that is a fee that assumes the higher cost of delivery. Model it honestly. If the current fee is X and the account is being delivered at cost 1.3X, the fee that makes it profitable at the current delivery model is closer to 1.6X than 1.3X, because you also need margin. Present the number with the reasoning. Be prepared for the client to leave. That's a fair outcome. The mechanics of the announcement itself, which accounts to raise on first, which to raise on last, and the meeting move that decides whether the increase actually holds, are in how to raise your prices without losing the accounts you actually want to keep; the order matters more than the wording.
Option two: reduce the deliverables to match the fee. If the fee can't go up and the delivery model can't change, the honest response is to reduce what you ship each month so the account's actual cost matches the actual revenue. Drop the monthly report to a quarterly one. Drop the strategy note. Whatever you drop, name it, and get the client's written acknowledgement that this is now the scope. Unpopular. Honest.
Option three: end the retainer at renewal. If neither of the above works, the account is unprofitable at any delivery model the client is willing to accept, and the right move is to let it end at renewal, on good terms, with a handover that protects the client. Don't fire the client mid-quarter. Time it to a natural boundary and offer to introduce them to two competitors who could serve them well. This costs you nothing that was not already lost, and it saves the reputation on both sides.
The one option that's not on the list is "continue and hope". The account isn't going to fix itself. Every month you don't act, the two-column note gets more lopsided and the account manager gets closer to a resignation letter.
Why this is not a dashboard problem
The temptation, having read the above, is to buy a piece of software that measures response latency and thinking time. There are many such pieces of software. They won't solve the problem, for two reasons.
The first is that measurement without a conversation is just data. If the account owner sees a red number on a dashboard once a month and doesn't have a script for what to do about it, the red number becomes background. The sort of thing you learn to notice without acting on. The two-column note works because it's done by the person who has authority to act, in the same twenty minutes it's done in. There's no lag between measurement and action.
The second reason is that the thing that fixes the account isn't the data. It's the willingness to have the conversation. Every account owner who has ever run a retainer knows, without a dashboard, which of their accounts are running on goodwill. The measurement isn't what surfaces it. The measurement is what gives them cover to raise it. If you're the founder, your job isn't to buy them the measurement. It's to give them cover, and to have the conversation yourself on the accounts where they can't.
Where this fits with the rest of the work
The retainer economics conversation sits next to a small number of related conversations, and understanding all of them together is more useful than understanding any of them alone.
If your problem is that you can see the account is unprofitable but you can't see the cash consequence yet, the difference between what you're measuring and what your bank balance is measuring is covered in Cash flow vs profit: the difference that sinks most small businesses. A retainer can be loss-making for six months before the cash-flow statement notices, and the reverse is also true. Don't conflate the two.
If your problem is that this isn't one drifting account but a whole book of them, and you can't tell which client the drift is actually in, the cross-sectional version of this measurement is in which client is actually profitable when everyone works on everything. How to allocate a shared team's cost across accounts without a timesheet culture.
If your problem is that you haven't built a retainer at all yet and you're pricing services one project at a time, the framework for setting a rate that protects against exactly the failure mode above is in How to price your services as a freelancer or consultant. Pricing badly at the start makes the re-scope conversation harder for years.
If your problem is that a client has already stopped paying, not is-not-quite-profitable but is-not-paying-at-all, the response is a different sequence and a different set of levers, in The client has stopped paying: here is the sequence, in order.
FAQ
How long does it usually take for a profitable retainer to turn unprofitable?
In most cases it happens between month three and month six, which is roughly when the delivery habits harden into a pattern the fee was never priced for. You won't see it on the P&L for another quarter after that. Which is why the two-column note at month-end matters more than the finance meeting.
Should I raise the fee or reduce the deliverables when an account slips?
Raise the fee when the client values what they're actually getting. The availability, the real-time replies, the informal strategy. Reduce the deliverables when they only value what's written on the scope. If you can't tell which one it is, the re-scope conversation is what tells you, because their reaction to the two structural changes is the answer.
What if the client refuses the response-window change?
That refusal is the answer to a question you were already asking. Which is whether this account can ever be profitable at the current fee. Move to option one from the post, raise the fee to match the delivery model, and if they refuse that too, time the exit to renewal.
How do I bring this up without sounding like I am asking for more money?
Open with the pattern, not the fee, and propose the two structural changes before you propose any commercial change. If the account genuinely needs a fee increase, it becomes the third move in the conversation, not the first, and by then you've already reframed the meeting from a negotiation into a working session.
Do I have this conversation with every client at once, or one at a time?
One at a time, starting with the account whose two-column note is most lopsided, because that's the one where the argument is easiest to make and the outcome is most instructive. Once you've run the conversation twice, you'll have a version of it that sounds like you rather than like a script, which is what the third client will need.
What if the account manager is the founder, is the conversation different?
The mechanics are identical, but the discipline is harder, because the founder has more tolerance for absorbing the invisible costs personally and calls that tolerance loyalty. The two-column note is more important, not less, when the person delivering the account is the person who will not resign over it.
Is a shared Slack channel with the client always a bad idea?
No, but a shared channel without a stated response window is, because the channel itself sets the expectation of an inside-the-hour reply. Keep the channel and add the window. One working day for normal messages, an "urgent" flag for genuine urgency, and a weekly office-hours block for real-time back-and-forth.
The one-sentence version
A retainer becomes unprofitable when the delivery model outgrows the scope document. The delivery model outgrows the scope document by two things, response latency and unbilled thinking time, and both are invisible to the tools that measure hours. The re-scope conversation is what turns the invisible into a decision, and the decision is what saves the account.
Do the two-column note this month. Have the conversation before you present the P&L. It's a shorter meeting, and a much cheaper one, than the one where you tell the team you're letting an account go.
Top comments (0)