Why Guides Fire Marketing Agencies

Nobody publishes data on why small businesses fire marketing agencies. I went looking, and it does not exist, which means every page you will read confidently listing the top five reasons is describing somebody's impression rather than a measurement. So this is written from mechanism instead, and the first thing worth saying is that engagements die in a small number of recognisable ways, each with a warning sign that shows up months before anybody gets fired. The uncomfortable half of the list is that two of the six are the guide's own doing, and one of those is firing a shop that was actually working. If you are trying to decide whether the problem is the agency or the business underneath it, the plateau diagnosis is the first thing to rule out.

Why is there no data on this?

Because nobody collects it. There is no register of small-business agency relationships and no survey of why they end. The nearest public thing is consumer complaint data, and it does not measure this at all, which is worth understanding rather than glossing over.

The FTC's Consumer Sentinel Network is the largest public collection of marketplace complaints in the country. In 2024 it received 6.5 million consumer reports sorted into 29 top categories. That is a lot of data about the marketplace, and none of it answers this question, because it is consumer complaint data rather than a record of business-to-business service relationships.

The FTC is admirably direct about its own limits, and the caveat is worth borrowing. The data book, it says, is based on unverified reports filed by consumers, and is not based on a consumer survey. That is exactly the right way to read any complaint statistic anybody shows you: complaints measure who complained, not what happened.

So when a page tells you that a specific percentage of businesses leave their agency for a specific reason, the useful question is where that came from. Usually the answer is a vendor survey of its own customers, which measures something, but not the thing being claimed.

Rio Anglers, TX

Death one: no attributable trips, and no agreed checkpoint

The most common ending and the most avoidable. Six months in, the guide cannot point at a single booking that came from the work, the agency cannot either, and neither party ever wrote down what would count as success or when it would be judged.

Notice that the failure here is not the absence of results. It is the absence of a standard. Without an agreed number and an agreed date, month seven becomes an argument about impressions and effort, and both sides can make a case, which means neither side can be satisfied.

Prevention is one sentence, written at signing: by this date we expect this many enquiries at roughly this cost, and here is what we do if we are short. Committing to it while the account is still empty turns a later argument about feelings into a comparison of two figures. That cuts both ways, and the side it protects most often is a competent shop having a slow first quarter.

Setting the figure is your own sum, weighed against a day's takings on the water, and the wider budget question is handled in the piece comparing agencies, freelancers and doing it yourself.

Death two: reports that measure motion

Impressions up, reach up, engagement up, phone quiet. This one kills slowly, because each individual report looks like progress and the trend only becomes obvious after a season. The warning sign appears in month one, if you know to look for it.

The tell is a report with no calls in it. Not a report with disappointing call numbers, which is honest, but a report that does not mention calls at all and leads with numbers that move because money was spent. Those numbers were never in doubt and cannot fail.

What makes this a fatal condition rather than an annoyance is that it removes the possibility of a real conversation. Nobody can say the work is not working, because the agreed measure keeps improving. The relationship then ends abruptly when the guide finally looks at their own book and notices nothing changed. What a monthly summary owes you is set out in the reporting piece.

Death three: custody discovered late

The relationship was fine and then the guide tried to leave, discovered the domain was registered to the agency, and everything afterwards was a negotiation instead of an exit. This one rarely causes the firing; it poisons it.

The reason it belongs on a list of why engagements die is that it converts a routine ending into a grievance. A guide who leaves cleanly rarely says much afterwards. A guide who spent six weeks getting their own domain back tells every other guide on that river, and the relationship is remembered as a bad one regardless of the work.

The whole audit takes about an hour and it is entirely preventable by the guide, at the start, at no cost. Eight assets, one check each, set out in the custody article. Doing it while everyone is friendly is the difference between an administrative request and a standoff.

Death four: the rep cliff

Sold by somebody sharp, served by somebody new. The founder ran the sales call and answered every question well, then the account moved to whoever joined most recently, and the quality of thinking dropped without anybody announcing a change.

This is not misconduct and it is not unusual. It is how agencies scale, and plenty of them do it well with proper handovers. It becomes fatal when it happens silently, because the guide experiences it as the agency suddenly getting worse for no reason.

The warning sign is specific and early: the person who sold it stops appearing on emails, and replies start arriving from a name you have not met, with no introduction. That is the moment to ask directly who owns the account now and what the handover involved. Asking early keeps it a normal conversation, and it is one of the things worth establishing before signing at all, per the pre-hire questions.

Death five: silence in your season

Response times slip, then the account goes quiet in exactly the months you are least able to chase anybody. For a seasonal business this is the most infuriating failure mode, because the neglect lands precisely when the work matters most and you have the least capacity to manage it.

The mechanism is usually mundane rather than sinister. A small agency's attention flows toward whoever is complaining loudest, and a guide who is out on the water for twelve hours a day is not complaining at all. Being a low-maintenance client is quietly penalised.

The early warning appears well before your peak: replies that took hours in month one take days by month three. Notice it in the quiet season and raise it then, because raising it in July is both harder and less likely to land.

The structural fix is to agree the seasonal shape at the start, so the busy months have a defined scope and a defined rhythm rather than depending on you chasing. That argument sits in the piece on how the agreement should be built.

Death six: fired too early

The one that is squarely the guide's fault, and the one with published numbers behind it. A campaign cancelled or repeatedly edited inside its calibration period was never given the thing the money was buying, and the guide concludes the channel does not work when what actually happened is that it never started.

Google publishes the duration. Its documentation puts calibration for an automated bid strategy at as much as three weeks, or one to two conversion cycles, and lists what sends it back to the start: standing up or reactivating a strategy, altering a setting, or adding and removing campaigns, ad groups or keywords. For a guide business whose clients click in February and fish in June, a conversion cycle can run to months rather than weeks.

So an account that is adjusted every time a week looks thin is an account permanently in its least stable state, and cancelling at week six measures the calibration period rather than the campaign. The published durations for each channel are laid out in the timeline article, and knowing them before you sign is what makes a checkpoint fair to both sides.

None of which means an agency should be given unlimited patience. It means the date you judge them on should be set with the mechanics in mind rather than by how you feel in month two.

Which of these are actually the guide's fault?

Two of six, and being honest about that improves your next hire more than anything else on this page. Firing too early is entirely the guide's error. Discovering custody late is the guide's failure to run a free audit at the start.

This is worth sitting with rather than skipping, because the instinct after a bad engagement is to conclude that agencies are the problem and to hire the next one more defensively. If two of the six failure modes were yours, defensiveness does not fix them and will actively make the next relationship worse.

The other four are genuinely the agency's: reports that measure motion, a silent handover to a new person, going quiet in your peak, and failing to establish a checkpoint they should have insisted on. A shop that has done this work before proposes the checkpoint themselves. One caution on benchmarks while you are judging any of this. Serious long-running sources such as The CMO Survey sample corporate marketing organisations, so their figures describe budgets and staffing no guide business will ever resemble. Useful for understanding how the other side thinks; useless as a bar to hold yourself to.

What are the early warnings, in order?

Month one: a report with no calls in it. Month two: no agreed checkpoint exists. Month three: reply times have doubled. Month four: a name you have not met is answering. Every one of these is visible long before anybody is angry.

The value of the ordering is that it tells you when to intervene rather than when to leave. A report with no calls in month one is a conversation, not a crisis, and a shop that adjusts the report when asked has just demonstrated something useful about how they take feedback.

Write the warnings down somewhere you will see them, because the failure mode is not missing the signs but normalising them one month at a time. Nobody notices reply times doubling; they notice a year later that the relationship went cold.

Cajun Paradise, LA

What is the point of no return?

When you stop asking. Once a guide has decided internally that the arrangement is not working and simply keeps paying while waiting for the contract to lapse, the relationship is over regardless of what happens next, and every further month is wasted money.

That state is common and it is the one worth avoiding, because it is expensive and silent. The agency does not know they are in it, so they have no chance to correct, and the guide has stopped investing the small effort that might have fixed things.

If you notice yourself there, either raise it properly with a specific list, or end it. The same disengagement shows up when a website has quietly stopped earning its keep, which is the subject of the stale-website piece. The middle position, paying and disengaging, is the worst available outcome for both parties and it is where a lot of these engagements actually live for their final six months.

How do you fire an agency properly?

Recover your assets first, then give notice. In that order, always. Once you have announced you are leaving, every access request becomes a negotiation, and the leverage you had five minutes earlier is gone.

Work the list before you send anything: domain in your name, website admin login working, ad accounts under your business account, Business Profile ownership, booking data exported, client list downloaded. Then check your notice period and what it says about work in progress and prepayments.

Then be straightforward and unemotional. Name the notice period, ask for the handover as a list of specific items with a date, and offer a transition call. Most shops behave well when the ending is professional, and the ones that do not have told you something you can pass on to the next guide who asks.

Resist the urge to explain at length. A short, factual message ages better than a grievance, and you are likely to cross paths again in a small industry. Rules and platform policies also shift, so verify the current position on anything procedural before you rely on it.

What separates the guides this does not happen to?

Three habits. The checkpoint is settled before any work begins, the custody audit happens in week one, and problems get aired in the off season instead of being banked until the calendar is full.

The checkpoint habit is the highest-value one because it makes firing rare. Most of these deaths happen because nobody defined what good looked like, so both sides drifted until somebody lost patience. A number and a date, agreed in writing at the start, prevents four of the six.

The timing habit matters nearly as much. Every conversation on this list is easier in February than in July, and guides who do their reviewing in the off season end up either fixing the relationship or replacing it cleanly, both of which beat limping through a season with a vendor they have privately given up on.

What are the common mistakes?

Giving notice before recovering the assets. Firing inside the calibration period. Never agreeing a checkpoint and then being surprised by the argument. And hiring the replacement with the same process that produced the last one.

The repeat mistake is the most costly over time. A guide who fires an agency and then hires the next one on the same basis, without a checkpoint, without a custody audit, and without asking who will actually do the work, has bought a second copy of the same experience. The list that prevents that is in the piece on what to check before hiring.

The sequencing mistake is the one that turns a clean break into a bad month. Assets first, notice second, and never the other way round.

What surprises people?

That no data exists on this at all. That two of the six causes belong to the guide. That the earliest warning usually appears in month one. And that the point of no return is a decision the guide makes silently rather than anything the agency does.

The absence of data is worth internalising because of how much confident writing exists on this topic. When somebody tells you the top reasons businesses fire agencies, they are almost certainly reporting a vendor survey or an impression, and neither is measurement.

The silent decision is the more useful insight. Engagements rarely end at the moment they fail; they end months later, after a long stretch where the guide has stopped engaging and the agency does not know it. Noticing that state in yourself is the single most actionable thing here. Other pieces on hiring and living with an outside shop are gathered under marketing help.

Should you replace them at all?

Not automatically. Run the diagnosis on your own business first, because if the constraint is capacity, concentration or price, a new agency will produce the same flat result at the same cost and you will have spent a season learning that.

The honest sequence is: establish what is actually binding, then decide whether the answer requires an agency at all. Plenty of guide operations at their ceiling need a rate rise and a second boat rather than a marketing vendor.

If you do replace them, replace the process rather than just the supplier. Same brief, same checkpoint, same custody terms, and a written answer to the question of what would make them tell you not to hire them. The mechanism for that budget conversation is in the cost article.