Almost no client leaves over a single missed deadline. They leave over a pattern, and the pattern almost always starts in the same place: the people who do the best work ran out of hours, and everything after that was triage.
Clients Leave Patterns, Not Incidents
One late deliverable is forgivable and usually forgiven. What is not forgivable is the third one in a quarter, because by then the client has updated their model of you. They are no longer dealing with a slip, they are dealing with what they now believe you are like.
That shift happens quietly and usually before anyone says anything. By the time a client raises it formally, they have often already started looking.
The Sequence That Precedes Churn
It runs the same way most times.
Capacity tightens, usually because you won something. Proactive work stops first, because it is the only thing with no deadline attached. Nobody notices, including you, because all the deadlines are still being hit.
Then response times stretch. Not dramatically, from two hours to a day. Then reviews get lighter, because reviewing is the second thing with no hard deadline. Then something ships that should have been caught.
Then the client asks a question they never used to ask, which is some version of what are we actually paying for. That question is the tell, and it comes about 60 days before the cancellation.
Four Leading Indicators
All four are visible in your own systems, and all four move before revenue does.
Proactive recommendations per account per month. The clearest signal. When this hits zero, you have become an order taker on that account, and order takers are replaceable on price.
Time to first response. Not resolution time, first response. Clients read silence as being deprioritised, and they are usually right.
Percentage of deliverables that skipped review. If you are not tracking this, assume it is rising whenever your team is busy, because review is what gets cut first.
Ratio of client-initiated to agency-initiated contact. When clients are chasing you more than you are updating them, the relationship has already inverted.
What Capacity Actually Buys
Not the ability to take more clients. That is what it looks like on a spreadsheet and it is the least valuable thing it does.
Capacity buys back the work with no deadline. The proactive recommendation, the second review pass, the analysis nobody asked for that finds the thing. That work is invisible when it happens and expensive when it stops, because it is the entire difference between a partner and a vendor.
It also buys absorption. Every agency has weeks where two accounts need more than planned. With no slack, that week costs you review quality on both. With slack, it costs you the slack.
The honest framing: added capacity does not make you money directly. It stops you losing accounts you already have, which is worth more and is much harder to see on a P&L.
The Wrong Time to Add It
Adding delivery capacity during a churn spiral rarely works, because onboarding new people consumes the senior time that was already the bottleneck. You make the next six weeks worse in exchange for making month three better, and the accounts at risk are leaving inside six weeks.
Add capacity when the leading indicators start moving, not when revenue does. Proactive recommendations falling to zero is the earliest reliable signal, and it typically shows up a full quarter before anything appears in your churn numbers.
If accounts are already leaving, triage first. Stabilise the specific at-risk accounts with the people who know them, and add capacity underneath once the immediate fires are out.
Part of: The GCC Model for Marketing Agencies, covering how agencies get offshore economics without building a center.