The most profitable decision I made one year was ending a retainer that represented a meaningful chunk of our monthly revenue. Within two months we had replaced it. Within four, the team's output on every remaining account had visibly improved.
I had known for at least six months. I kept it because losing revenue feels like failure and because I told myself the situation would improve. It never does. In eleven years I have never once seen a bad client relationship improve on its own.
The four kinds of client you should end
The margin destroyer. They pay a fair fee and consume three times the hours. Usually through revision cycles, unscheduled calls, and scope that expands quietly. The tell is that your team dreads the account but cannot explain why in a way that shows up on an invoice. If you do not track hours per account, you cannot see this, which is one reason I insist on the measurement discipline in agency profitability metrics.
The disrespect case. Rudeness to your team, hostile revisions, missed payments treated as normal. This one has a zero tolerance rule for me now. The damage is not to me, it is that every person on my team learns what behavior I will accept in exchange for money.
The strategic mismatch. They want something you do not do well, or they want a channel you do not believe in. You can do competent work here and still be the wrong provider. Keeping them prevents you from getting better at the thing you actually want to be known for, which undermines everything in positioning against bigger agencies.
The unwinnable engagement. They will not give you access, will not approve anything, will not provide the inputs the work requires, or have an internal political problem you cannot solve. You will be blamed for the outcome regardless. Duration does not fix it.
The test I use before deciding
Three questions, answered in writing.
Would I take this client today, knowing what I now know, at the fee they currently pay? If no, you have your answer, and the only remaining question is timing.
Is the problem structural or situational? A rough quarter caused by a stakeholder change or a company crisis is situational and worth waiting out. A relationship that has been the same for a year is structural.
Have I actually attempted a reset? This is the one people skip. Half the accounts I have considered firing were fixable with one honest conversation that I had been avoiding.
Try the reset first, and mean it
The reset conversation is not a warning and should not feel like one. It is a renegotiation.
Structure: name the specific pattern with examples, state its effect concretely, propose the new arrangement, ask if it works for them.
"Over the last four months we have averaged nine revision rounds per deliverable against a scope of two. That is why deadlines have slipped. Going forward I want to move to two included rounds with additional rounds billed hourly, or raise the retainer to reflect the actual volume. Which works better for you?"
Two things happen. Some clients say they had no idea and fix it immediately, and that account becomes one of your better ones. Others react badly, which tells you what you needed to know, and you have made the ending much easier to justify.
A price increase is a legitimate form of this. Raise the fee to what the account is genuinely worth to serve. If they accept, the problem is solved. If they leave, the problem is also solved. The psychology of why this works is in pricing psychology, and the mechanics of structuring it are in pricing strategy for services.
The actual exit, step by step
Decide the date before the conversation. Usually 30 to 60 days out, honoring your contract. Never end mid deliverable if you can avoid it.
Do it on a call, not in an email. Send a written summary after. Doing it in writing only reads as cowardice and gets forwarded.
Give one reason and make it about fit, not fault. "We have narrowed our focus and I do not think we are the right team for what you need next" is true in almost every case and does not invite debate. Listing their sins is satisfying for four seconds and permanently expensive.
Do not negotiate. If you have already tried the reset, this is not the moment for a counteroffer. Accepting one means repeating this conversation in five months with the relationship worse.
Offer a real transition. Two or three referrals to firms genuinely better suited. A documented handover. Access to every asset, organized. This costs you a day and it is the difference between an ex client who speaks well of you and one who does not.
Do not disparage them afterward. Not to prospects, not to your team beyond what is operationally necessary, not in public. Our industry is small and every conversation gets back.
What to say to your team
Be specific about why, because they are watching. If you end an account over disrespect, say so plainly. It is one of the strongest culture signals a founder can send, and it buys you enormous credibility the next time you ask people to push hard for a client who deserves it.
If you end it for margin reasons, share the actual numbers. Teams handle real constraints well and handle vague ones badly.
The math that makes this easier
Founders keep bad clients because the revenue is visible and the cost is not. Make the cost visible.
Take the hours the account consumed last quarter and multiply by the fully loaded cost of the people involved. Add the opportunity cost, which is what those hours would have produced on your best account or on your own business development. Then subtract the fee.
I have run this and found accounts that were nominally profitable and actually costing money once you counted the founder hours spent managing the relationship. That number ends the debate faster than any feelings based argument.
There is also a compounding cost that never appears anywhere: a bad account degrades the work you do for good ones, because attention is finite and stress is contagious. I have watched a single difficult client measurably slow delivery across an entire team, which is a cash flow problem as much as a morale one, and cash flow is the thing that actually kills agencies as I argued in cash flow for small business.
Build the exit into the contract
The best time to make firing a client easy is before you sign them.
A 30 day termination clause available to both parties, without cause. Defined scope with explicit revision limits. A payment terms clause with a work stoppage trigger. A clause on communication channels and response expectations. And, before you sign anything performance based, model the realistic case through a ROAS calculator rather than the optimistic one, because a deal priced on optimism is a deal you will want out of by month four.
None of this is adversarial. It is the same instinct as defining success criteria before a hire, and clients who object to reasonable mutual terms are showing you something useful for free.
The two week test
If you are reading this because a specific client came to mind on the first paragraph, here is what I would do.
Give yourself two weeks. In week one, run the numbers on hours versus fee and have the reset conversation. In week two, decide. Not "think about it," decide, with a date on a calendar.
The reason for the deadline is that this decision does not get easier with information. It gets easier with commitment. Every founder I know who ended a bad account says the same thing afterward, which is that they wish they had done it sooner, and none of them say they wish they had waited to be more certain. The certainty arrives after the decision, not before it.