Agency client churn rate benchmark: what is normal, what is a warning, and which churn you should be causing on purpose

You lost two accounts this quarter and want to know whether that is normal. The honest answer depends on what you sell, how you count, and whether you would have kept those two accounts if they had offered to stay. Most benchmark pages skip all three questions and give you a single percentage, which is why the figures they quote run anywhere from 2% to 50%.

The published numbers, and where each one comes from

The agency-specific figures that trace to a named source are narrower than the spread suggests.

Karl Sakas, who consults to agencies, writes that at a retainer-oriented agency he would be concerned if annual client turnover ran higher than 20%, and that a project-based agency could reasonably sit closer to 30% to 50% a year. That is the most-quoted operator figure in this market, and it is the one to measure against if you run retainers.

Predictable Profits' 2025 Agency Growth Benchmark, as reported by Swydo, puts the benchmark annual retention for retainer-based agencies at 80% or better, with 70% as the red flag. The same study splits by size: seven-figure agencies benchmark at 78% retention and eight-figure agencies at 92%, which reads as annual churn of roughly 22% and 8%. Project-based agencies benchmark at 50% to 70% retention.

The ANA and 4As 2025 Client-Agency Relationship Tenure study reports average agency-of-record tenure at about seven years, 7.3 years for independent agencies against 5.8 for holding company shops. That sample is 67 agencies reporting on 426 relationships, mostly large brand accounts, so it describes a different market from a twelve-person agency with £8k retainers. Treat it as the ceiling on what a relationship can look like, and the Sakas and Predictable Profits figures as the ranges to run your own agency against.

The figures you will see elsewhere, like a 70% or 82% average retention rate across all agencies, repeat from blog to blog without a primary source we could find, and we have left them out for that reason.

Count revenue, not logos

Two agencies can both lose three clients from thirty and have had completely different years. One lost three £2k accounts it had been meaning to release. The other lost its largest client, which was 25% of revenue. Both report 10% logo churn.

The number that describes what happened is gross revenue retention: the recurring revenue you held from the clients you started the year with, as a share of what they were paying at the start, ignoring any upsell. Add the upsell back and you get net revenue retention, which can run above 100% in an agency that expands its accounts faster than it loses them.

The calculation for a year:

GRR = (starting recurring revenue - revenue lost to churn and downgrades)
      / starting recurring revenue

Downgrades count. A client who cuts the retainer from £10k to £4k has churned 60% of their account, and a logo count will tell you nothing has changed. The cuts that happen quietly, through scope creep in reverse, are usually the earliest signal that a client is halfway out.

Separate the churn you chose from the churn that chose you

This is the split that benchmark pages never make, and it changes what the number means. Hydra works a 40% gross margin floor per account, checked quarterly, with the options for any account under the floor being re-price, re-scope, or release. An agency running that process will release accounts every year, and those releases show up as churn.

So keep two lines. Chosen churn is revenue from accounts you released, or re-priced and lost on purpose. Unchosen churn is everything else. A retainer agency with 18% total churn where 10 points were chosen is healthier than one at 12% where every loss was a surprise, because the first agency is pruning and the second has a delivery or sales problem it has not found yet.

Unchosen churn above the Sakas 20% line on a retainer book points at one of three places: the account was sold to a client who was never a fit, the delivery stopped showing value in a form the client could repeat to their boss, or the account manager carries too many accounts to notice the warning signs. The when to fire a client piece covers the first, and the question of how many clients per account manager covers the third.

Why churn hides a growth ceiling

ZeroTo1 was 100% referral dependent when we started working together. New referrals roughly matched churn each year, so the agency sat at $78k a month and felt plateaued, when the real problem was that nothing but referrals was feeding the top of the funnel. That is the referral equilibrium: churn at a normal rate, acquisition at a rate that exactly cancels it, and a founder convinced the problem is retention.

Check for it by writing the two numbers next to each other. If annual revenue churn and annual new revenue from referrals are within a few points, the constraint is acquisition, and no retention program will move the top line. The binding constraint is one thing at a time, and churn is the wrong one to attack in that situation.

What it looks like when it works

ZeroTo1 went from $78k to $414k a month over 36 months once a go-to-market system existed alongside the referrals, with contract value rising from $8k to $20k a month over the same period. Fraggell Productions more than doubled its average retainer from £5.5k to £13.5k while growing from £23k to £110k a month over 24 months, which is net revenue retention doing the work that new logos usually get credit for.

If you want to know whether your churn is a retention problem or an acquisition problem wearing its clothes, the diagnostic runs the split for you.

Read next

• When to fire a client, and how to know before it costs you a year
• Agency financial benchmarks: the numbers a healthy agency actually hits