Customer Churn Warning Signs and What to Do About Them

The most reliable customer churn warning signs are changes in behavior: a customer buys less often, spends less, uses your product less, stops opening your messages, raises complaints that stay unresolved, gives a low satisfaction score, or their main contact person changes. Any one signal can be noise. Two or three together usually mean a customer is drifting away, and that is the moment to reach out personally, before they have made up their mind.

This guide lists the signals by business type, shows how to combine them into a simple health score, and gives you a response playbook for each one.

Why churn is rarely sudden

From the outside, a customer's departure often looks sudden: one day they cancel, or they simply never come back. From the inside, it is usually gradual. Something disappoints them, they look at alternatives, they buy a little less, they stop reading your emails. By the time they cancel, the decision was made weeks or months earlier.

That gap between the first signal and the final decision is your window. Small businesses have an advantage here: you can often notice changes personally, and a call from the owner carries weight that an automated email does not.

Warning signs by business type

Signals differ depending on how customers interact with you.

Repeat retail and e-commerce

Signal	How to spot it
Longer purchase interval	Days since last order exceed the customer's own usual interval, for example by 50% or more
Smaller baskets	Average order value falls over the last few orders
Narrower range	They used to buy several product categories, now only one
Returns or complaints	A recent return, damaged delivery or complaint
Email disengagement	No opens or clicks for several campaigns
Discount-only buying	They only purchase when there is an offer

Subscriptions and software

Signal	How to spot it
Falling usage	Fewer logins, fewer active users, key features unused
Downgrade	Moved to a cheaper plan or reduced seats
Failed payments	Card declined and not updated
Support pattern	A spike in tickets, or complete silence after early problems
Champion left	The person who bought the service no longer works there
Export activity	They download all their data

Services and B2B accounts

Signal	How to spot it
Fewer enquiries	No new requests after a project ends
Slower responses	Your emails take longer to be answered
Cancelled meetings	Check-ins postponed or declined
New stakeholders	A new manager or procurement asking for quotes
Invoice friction	Payments later than usual, more questions about invoices
Scope reduction	Smaller follow-up projects, fewer services

The two signals people overlook

Silence. Many businesses only react to loud signals like complaints. But a customer who complains still cares enough to tell you. A customer who quietly stops engaging may already be gone. Track "days since last contact" for every important account.

A change of contact. When your main contact leaves, the relationship often goes with them. The new person has no history with you and may have their own preferred suppliers. Treat a contact change as a reason for a proper introduction meeting, not just an email update.

A simple customer health score

A health score combines several signals into one number so you can sort customers by risk. It does not need special software. Here is a template for a service or subscription business:

Signal	Healthy (2 points)	Watch (1 point)	At risk (0 points)	Weight
Usage or purchase frequency vs. their normal	At or above	Up to 30% below	More than 30% below	3
Last satisfaction score	9–10 NPS or 5 CSAT	7–8 or 4	0–6 or 1–3	2
Open issues	None	One, in progress	Unresolved for over a week	2
Days since last two-way contact	Under 30	30–90	Over 90	2
Relationship	Known champion, engaged	Champion less active	Champion left	1

Score each signal, multiply by the weight and add up. The maximum is 20 (2 points × total weight of 10).

Worked example. A customer's usage has fallen 40% (0 × 3 = 0), their last NPS answer was 8 (1 × 2 = 2), they have no open issues (2 × 2 = 4), you last spoke 45 days ago (1 × 2 = 2) and the champion is still there but less active (1 × 1 = 1). Total: 9 out of 20.

You might set thresholds like this:

• 15–20: healthy, normal contact rhythm
• 9–14: watch, check in within two weeks
• 0–8: at risk, personal contact this week

Keep the score simple at first. Adjust weights after a few months by looking at which customers actually left and which signals they showed beforehand. If you survey regularly, the NPS calculator helps you track the satisfaction input over time; see how to calculate NPS for the method.

The response playbook

A signal is only useful if it triggers an action. Agree in advance what happens for each one.

Signal	First response	Who
Purchase interval stretched	Personal note: "We noticed it has been a while, is everything OK with…?" plus something useful (new product, reminder)	Account owner or automated with personal sender
Falling usage	Offer a short session to help them get more out of the product	Customer success or owner
Low survey score	Call within two working days to understand and fix	Manager or owner
Unresolved complaint	Escalate, set a resolution date, inform the customer	Support lead
Champion left	Request an introduction meeting with the new contact	Account owner
Downgrade	Ask what changed; check whether a different setup fits better	Account owner
Failed payment	Friendly reminder with a direct link to update details	Automated, then personal
Discount-only buying	Test value-based messages rather than more discounts	Marketing

A check-in script that does not sound like a sales call

"Hi Alex, it's Jordan from [business]. I'm not calling to sell anything. I noticed we haven't [seen you / worked together] since [month], and I wanted to check how things are going. Was there anything about [last order / project] that didn't work as well as you hoped?"

Then listen. The goal is to learn what changed. If there is a problem, fix it or say honestly what you can do. If their needs simply changed, thank them and leave the door open.

Common mistakes

• Waiting for the cancellation. By then, the customer has usually found an alternative.
• Treating every signal the same. A slightly late reorder from a small customer does not need the owner's personal attention; a drop in usage at your largest account does.
• Leading with a discount. It signals that price was the issue, even if it was not, and it trains customers to wait for offers.
• Building an elaborate score nobody checks. A five-signal spreadsheet reviewed weekly beats a complex model reviewed never.
• Not learning from losses. When a customer does leave, look back at which signals appeared and when. Use that to tune the score.

Measuring whether it works

Track two things:

1. Churn rate over time, using the same definition each period. The churn rate calculator gives you churn and retention for any period and converts monthly to annual rates.
2. Save rate for flagged customers: of the customers who reached "at risk", how many are still customers three months later?

If the save rate is low, look at the response, not just the score. Are the right people making contact? Is the contact early enough? If churn does not move at all, your signals may not be the ones that predict leaving in your business, which is a reason to review recent losses in detail. For a fuller picture of when customers drop off, see cohort retention analysis.

Your next step

List your 20 most valuable customers in a spreadsheet and score them with the five signals above. It takes about an hour. Anyone below 9 gets a personal call this week. Then review the list monthly, and fold the signals into your wider customer retention strategies.