Seven of the eight clients one agency lost in August had been clients for two months or less.
In the last two weeks, six different companies asked me the same question about their agency's client churn: what monthly churn is acceptable?
Google has answers. A lot of them. And they contradict each other — one page says 5% is fine, another calls it a crisis, and none of them agree on what "agency" even means. An SEO shop and a lead-gen agency don't churn the same way, and benchmarking one against the other is nonsense.
So here's something the SERP doesn't have: real churn numbers from four seven-figure agencies we ran diagnostics on in the last 45 days. All four came to us because churn was the problem.
Honest caveat up front: most of these numbers are the agencies' own accounts, not pulled reports. That tells you something about the state of churn data in this industry — the people who need it most barely track it.
Why agency churn benchmarks contradict each other
Three reasons.
First, everyone defines churn differently. Some sources quote monthly, some annual. Some mean clients lost, some mean revenue lost. A 10% annual churn agency and a 10% monthly churn agency are different species — the first keeps clients ten years on average, the second keeps them ten months.
Second, they pool every model together. Retainer agencies, project shops, lead-gen operators and creative boutiques churn for completely different reasons. One blended benchmark is useless to all of them.
Third, two of the four agencies we looked at couldn't state a churn rate at all. Not "it's bad" — they genuinely didn't track one. When half the industry is guessing, the benchmark pages are just ranking guesses.
The only honest benchmark is math, and it's three lines long. I'll show you that math below.
The real numbers: four agencies, four models
Here's what churn actually looked like inside the four shops. I ran a diagnostic on each — different services, different teams, and the reasons their clients left were almost identical.
| Agency type | Monthly churn we found | Contract setup | Healthy looks like |
|---|---|---|---|
| SEO + AI search, B2B SaaS and local | ~10% of clients a month (about 1–2 clients of 25–30) | 3–4 month terms | Clients clearing the 60-day cliff; SaaS and local tracked separately |
| Lead-gen, UK home services | 5–12% across three straight months (owner's recollection) | Month-to-month | Range narrowing month to month; fewer exits after month two |
| Paid creative, e-commerce brands | No rate tracked — but 7 of 8 August exits were clients of 2 months or less | Month-to-month | Fewer month-two exits; early wins visible in the first 30 days |
| Full-service DTC growth | No rate tracked | Not disclosed in the diagnostic | Churn tracked at all; clients clearing the first 90 days |
A few things jump out.
The SEO agency had the same lifetime value for SaaS and local clients — about $25–35K — but SaaS clients churned at nearly double the rate of local clients after month six. Same service, same team, different model. One benchmark for "SEO agencies" would hide that completely.
The lead-gen agency's numbers swung from 5% to 12% month to month. On month-to-month terms, a bad month is a free month or a cancellation, so the range is the story.
And the DTC growth agency didn't track churn at all — while raising its average client bill from about $5K to $20–25K a month over three years, stretching its strategists past 7–8 clients each. Bigger retainers, bigger expectations, no instrument to see clients drifting.
In none of the four was the core work the cause. The failures were in onboarding, communication, reporting, and visibility.
The only benchmark math that never lies
Here's the thing.
Monthly churn converts directly to how long a client stays. Roughly: divide 1 by your churn rate.
| Monthly churn | Average client lasts about |
|---|---|
| 12% | 8 months |
| 10% | 10 months |
| 5% | 20 months |
| 3% | ~33 months |
So when someone says "10% monthly churn is normal," they're saying an average client lasts ten months. Read it as months, not percentages, and suddenly every benchmark argument on page 1 gets very easy to judge.
My working rule from doing this work: clients who churn badly show it inside the first 60 days. Clients who clear that bar tend to stick around about two years. A two-year average life works out to roughly 4% a month, blended across the book.
Is that aggressive? Maybe. But "what churn is acceptable" is the wrong question anyway. The right one is how many months does your average client stay, and what does each lost month cost you?
How to calculate your real churn (logo vs. revenue)
Most agencies that do track churn track it wrong. There are two numbers and they answer different questions.
Logo churn is clients lost divided by clients at the start of the month. Start with 30, lose 3, that's 10%. It tells you how fast your book is leaking.
Revenue churn is the MRR you lost divided by your starting MRR. It tells you how much money the leak costs. If the 3 clients who left were your smallest, revenue churn reads better than logo churn. If they were your biggest, it reads a lot worse.
Then there's net revenue churn, which subtracts expansion and upgrades from the clients who stayed. You can lose 10% of clients and still grow MRR if the survivors expand. It happens — and it's why a "bad" logo number sometimes coexists with a growing business.
The trap that catches everyone: a big signing month hides churn. Five clients in, three clients out — the book is bigger, so it feels fine. The churned ones were quietly getting worse for months.
The fix is cheap: group your clients by the month they started, then check how many are still here at month two, three, six. That one view shows you the month-two cliff I'm about to describe. If you only build one report, build that one.
Why month two is the danger zone
Every agency we looked at lost clients in the first 30–90 days. The first 7 to 30 days decide whether a client stays — one owner told us exactly that. Here's what's actually killing clients that early.
Onboarding is the weakest system in the building.
Rushed launches: leads routed to the wrong CRM, wrong phone numbers on landing pages, one client who got zero leads because an automation failed. Manual onboarding means tasks slip where the client can see them. In one agency, the sales handoff brief — the thing that carries what was promised into delivery — was used about 80% of the time. That missing 20% is where clients go to die.
The honeymoon ends and there's no proof yet.
The work is happening, but the client can't see it in terms they care about. Agencies report clicks and impressions; clients want leads and revenue. One agency lost a client while 137 finished pieces sat unpublished — the work existed, the proof didn't. In another, the team talked to clients "very little," and early wins were never surfaced. Clients left while results were starting to show.
What sales promised isn't what delivery heard.
Expectation mismatch is set at the sale and carried into delivery. One churned client thought the scope was wider than SEO. Another expected a £50 cost-per-lead on a £20K job and was seeing about £500. These aren't delivery failures. They're handoff failures.
Nobody's watching.
Clients judge daily. Agencies review monthly. Problems surface at the month-end invoice, after the client has spent four weeks seeing no return. Nobody was running an early-warning system — sentiment, cancelled calls, quiet clients. Every agency we looked at is building one now, because of the churn.
The contract amplifies all of it.
Three-to-four-month terms for a service that takes longer than that to show ROI. Month-to-month where one bad month means a free month or an exit. The contract structure doesn't cause churn, but it decides how fast a wobble becomes a cancellation.
Above the benchmark? Fix the first 90 days, in this order
When an agency is above benchmark, I don't start with retention tactics. I start with the first 90 days, because that's where clients are actually leaving.
- Set expectations at the sale, in writing, and carry them into delivery. The handoff brief isn't optional. The promise has to survive the jump from sales to delivery intact.
- Build an early-win plan for month one. Not a wish list — specific proof the client will see in the first 30 days, reported in terms they care about (leads, revenue, costs — not impressions).
- Run a month-three ROI review before the client runs their own. If the client does the math first, you're already defending.
- Put approval gates on early deliverables — and a silent-approval rule. Work dies waiting for client sign-off. A rule that says work moves forward after a set deadline keeps months of deliverables from sitting unpublished.
- Watch for quiet clients. Cancelled calls, slipping sentiment, invoices drifting — that's your churn forecast. One score per client, checked weekly, alerts when it drops. This is the system every agency builds too late.
- Match contract length to the service's time-to-value. If the work needs six months to prove out, three-month terms are a churn machine. Ninety-day initial terms give both sides a fair shot.
- Check day two. Leads arriving, tracking live, access confirmed — before the client has to ask. The first 48 hours set the tone for the first 90 days.
Retention isn't a program you bolt on. It's what happens when the first 90 days work. The client communication layer and your capacity to actually respond are the two things that decide it.
FAQ
What is a good monthly churn rate for a marketing agency?
It depends on the model — which is why blended benchmarks are useless. My working rule: a healthy book keeps clients around two years, which works out to roughly 4% a month. If you're losing clients inside the first 60 days, that's bad churn no matter what the blended number says.
What's the difference between logo churn and revenue churn?
Logo churn is the share of clients you lose. Revenue churn is the share of MRR you lose. Track both — losing 10% of clients hurts differently depending on whether they were your smallest or your biggest accounts.
Why do agency clients leave in month two?
Because month two is when the gap between what was promised and what has been delivered is widest. Onboarding wobbles, no visible ROI yet, and the client re-evaluates with almost no relationship built. Seven of eight exits at one agency hit exactly this window.
How do I calculate my churn rate?
Clients lost in the month divided by clients at the start of the month. Then group clients by start month and see how many are still there at month two, three, and six. That grouped view matters more than the monthly number.
Is 10% monthly churn bad?
It means the average client stays about ten months. For the SEO agency we diagnosed at that rate, it also meant replacing a tenth of the book every month just to stand still — while contracts ran only 3–4 months. Sustainable? No.
Should account managers be scored on churn?
Only on the parts they control. One agency measured success reps on churn while the ad quality sat with a separate team the reps couldn't influence. Score the outcome someone actually owns.
What this looks like in your business
If your churn looks like the numbers above, the problem usually isn't your service. In the last 45 days we've run this diagnostic for four seven-figure agencies, and every time the diagnosis was the same: the work was fine, the first 90 days weren't.
We look at your last six months of churn, find where clients actually leave, and rebuild the first 90 days — onboarding, proof of value, early warnings — so they stop.