When to fire your agency and how to transition cleanly

Most owners get this backwards. They send a frustrated message, the relationship goes formal, and then they start asking for logins from someone who has no reason to hurry.

Do the boring part first. Access, exports, documentation. Then have the conversation.

What actually justifies firing an agency?

Three things, and all three are documentable.

Work in the scope is not being produced, and it continued after you asked for a dated change log. One missed month is a busy month. Two after a written request is a pattern.

You were told something that was not true. A deliverable reported as complete that was not, a metric presented in a way designed to mislead, or a claim about how search works that does not hold. That last one is checkable more often than people realize, and most claims of this kind can be traced back to a source or to nothing.

Nobody can explain the work. Not in detail, in outline. If three attempts to get a plain explanation produce three different fogs, the account is not being run by anyone.

A fourth, softer reason is legitimate: your needs changed. You have grown into work they do not do, or shrunk into work that does not need them. That is not firing, it is ending, and it should be said that way.

What does not justify it?

A bad quarter. Search results move on their own, competitors invest, algorithms change. A single down period with a written explanation is not evidence of anything.

Results slower than promised, when the promise was the problem. If you were told ninety days and the work needed nine months, you were oversold on timeline. That is a real failing, and it is worth one honest conversation before it becomes a termination.

Personality friction, on its own. Someone can be dry, slow to return calls, or bad at meetings and still be doing excellent work. Weigh the output.

Price, on its own. If the work is good and the fee has grown uncomfortable, ask to restructure. Different pricing structures move risk around, and a scope reduction is usually available for the asking.

Before you fire anyone, make one specific thirty-day ask in writing. Name the deliverables, name the date. A surprising number of engagements recover from exactly this, because the account had drifted and nobody had said so plainly.

What should I do before I send notice?

Seven things, over about a week, none of them confrontational.

Read the termination clause and compute the exact date and delivery method for notice. Some contracts require a specific address or form.

Confirm owner-level access on everything: domain registrar, hosting, site admin, Google Business Profile, analytics, Search Console, ad accounts, review platform.

Export what you own. Content files, images, form submissions, customer data, reports, and any research they produced.

Record a baseline before anything moves. Pull ninety days of Google Business Profile performance data, your current calls and form submissions, and a rank grid. That baseline is the only thing that will let you tell later whether the gap hurt you or the new vendor helped you. An hour now, and it settles an argument you cannot otherwise settle.

Ask for a documentation handover while things are still friendly. What was done, what is scheduled, what is half finished, and any credentials for tools that will transfer.

Port call tracking numbers into an account you control. A number printed on a truck or a menu that stops working is the most expensive mistake in a bad transition.

Check what is running unattended. Ad campaigns still spending, automated posts, scheduled emails, and any tracking or messaging script the agency installed in your name. Know what will keep going after they stop watching it, because it is your business name on those accounts.

What does the transition gap actually cost?

Here is the arithmetic, and it is the section most owners have never run.

Reviews are the clearest case because they are the thing that quietly stops the day nobody owns the account.

Set the benchmark. Darren Shaw of Whitespark gives the method plainly: find how often your top competitors are getting new reviews, and aim for that rate plus one. Say your three closest competitors are each adding two a month.

Map the gap. You send notice on 1 June. The final month is contracted but the asking quietly stops around 15 June. You sign a new vendor on 1 August and their system goes live on 20 August.

That is a 66 day gap. Your competitors each added four to five reviews. You added none, and your newest review is now more than two months old.

Price the recovery. Catching up at three a month against their two closes a nine review deficit in about nine months. Trying to close it in three weeks by blasting your whole customer list runs into Google's policy, which names content exhibiting unusual volumes or patterns of review contributions as a violation. A posting restriction or a public consumer alert banner stops your velocity dead while competitors keep accruing.

Why this matters more than the count suggests. Shaw's position is that the moment you stop getting new reviews, local rankings start to slip. Joy Hawkins of Sterling Sky documented a client whose rankings dropped with no technical cause until an audit found the review flow had flat-lined, because the owner had stopped rewarding staff for asking. Rankings recovered when reviews resumed. In Whitespark's 2026 factor survey, Recency of Reviews and sustained influx over time are scored as two separate factors, both inside the top fifteen.

Run the same arithmetic on anything else with a cadence: profile posts, content publishing, competitor monitoring. The gap compounds at your competitors' rate, not at yours.

What is the right order for the transition?

Notice, then overlap, then handover, then a documented restart.

Send notice in the required form, plainly and without a list of grievances. You are not writing a review, you are ending a contract. A short professional message keeps the final month cooperative, and you want that month worked.

Ask for the final report and a status list of anything in progress. Anything half done is your biggest risk: a partial migration, a citation cleanup abandoned midway, a redirect map half built.

Try not to leave a long gap. Ongoing tasks stop when nobody owns them. Review responses, profile monitoring, holiday hours, and competitor issues all sit unattended, including the ones that need active reporting, such as a competitor stuffing keywords into their business name, which only gets fixed by somebody filing it.

And do not take anything down in frustration. Deleting pages, cancelling hosting, or letting a domain lapse during a handover causes damage that outlives the grudge.

If the reason you are leaving is that you paid for something that does nothing, gather the evidence calmly rather than arguing about it. Some of these are settled. Ahrefs examined 137,210 domains and found that 97% of published llms.txt files received zero requests in a month from any bot or human, that 28% of sites published one, and that the AI retrieval bots which actually produce citations accounted for 1.1% of the small amount of traffic that did occur. Google states Search ignores the file. That makes the case against paying for one straightforward without anyone needing to be called dishonest.

What should I not let the next agency tell me?

That your numbers are below the industry average. There is no current industry average, and the confident version of that sentence is a sales move.

The only large-sample study of Google Business Profile performance is BrightLocal's, published in July 2019 on data from September 2017 through December 2018, covering 45,264 listings. It reported a median view-to-action conversion rate of 4.68%. It has never been refreshed. Worse, several of the metrics it was built on no longer exist: Google retired photo insights and the direct-versus-discovery search breakdown when it moved from the old insights to the current performance reports, and views are now a deduplicated count of unique visitors, capped at one per person per day, which is a different quantity from what 2018 called a view.

So that number cannot be reproduced on a modern account, and it should not be used to grade yours. Other published benchmarks in this category state no sample size and no method at all.

The honest replacement is the baseline you recorded before you sent notice, plus your named local competitors. Compare your trailing ninety days against your own prior ninety days. Any vendor who leads with a published average instead is either using a number they have not checked or hoping you will not.

What to do this week

Make the access list and work through it. Six logins, confirmed by password reset to your own email, not by being told you have access.

Record the baseline. Ninety days of profile performance, current calls and forms, one rank grid, one screenshot of your review count and the date of your newest review.

Write the thirty-day corrective ask if you have not already. Specific deliverables, specific date, sent in writing. If it lands, you may not need any of the rest.

Then decide what happens next before you send notice. A vendor, an interim plan, or handling it yourself for a quarter. Any of those is fine, and an honest comparison of doing it yourself against hiring is worth running before you commit, because the answer depends on hours you may not have.

Questions about an agency transition?

Email me at eric@seod.com with your offboarding list, the accounts and assets you plan to recover, and I will tell you what is missing from it before you send notice. It is a checklist reply, nothing more. I am not applying for the work and will say so in the first line.

There is more on hiring and working with agencies if you want to prepare properly.