A churn signal is an observable behaviour or data change that predicts a client is likely to leave. Churn signals matter because retention work is only affordable when it is targeted — an agency cannot give every client in a 3,000-policy book personal attention, but it can give it to the 90 showing warning signs.
The strongest signals
Roughly in order of predictive strength:
- A request for loss runs or a copy of the dec page — this is usually the client shopping. Treat it as urgent, not administrative.
- A large renewal rate increase — typically anything above 10–15%, and the effect compounds if the prior year also increased.
- Coverage reduction requests — raising a deductible or dropping optional coverages often signals price pressure preceding a move.
- A payment failure or late payment, especially a first-time one.
- Monoline status combined with short tenure — a first-year single-policy client is structurally the most fragile client in the book.
- Silence — no contact in twelve months, which is less a signal than an absence of the thing that prevents churn.
- A service complaint, particularly one that took multiple contacts to resolve.
The value of ranking
Signals do not carry equal weight, and combining them beats any single one. A monoline first-year client with a 14% increase and a recent complaint is not "three flags" — it is a near-certain loss unless someone calls this week.
Why this is analytics, not intuition
Every one of these signals already exists in the AMS and the carrier feed. What is missing in most agencies is not data but a standing process that evaluates the book against these conditions every week and puts a ranked, reasoned list in front of the person who can act.
That is a modest technical problem and an enormous commercial one.