Churn prediction: how to notice a client is leaving while they're still yours
Clients3 min read
What client churn is, how to calculate it and how AI in a CRM warns about clients at risk of leaving. Formula, examples and retention steps in Mestra.

Churn prediction estimates which clients are about to stop buying. In Mestra AI tracks purchase and communication history and warns about clients at risk, giving the manager a reason to reach out while the relationship is still alive.
The one-minute math
Churn formula: churn = clients lost in the period ÷ clients at the start × 100%.
- At the start of the quarter there were 400 clients; 48 left during it.
- Churn: 48 ÷ 400 × 100% = 12% per quarter.
- Cutting churn to 8% saves 16 clients per quarter.
- At an average 30,000 ₽ per client per quarter, that's 480,000 ₽.
Signs a client is leaving
| Sign | What it means |
|---|---|
| Gap between purchases grows | Found an alternative or lost interest |
| Average order shrinks | Some purchases moved to competitors |
| Slow to reply to messages | The connection is weakening |
| Didn't renew a contract or membership | A direct risk signal |
| Complained and got no solution | High churn risk |
How it works in Mestra

- Client history on one card. Deals, invoices, chats from every messenger, tasks.
- AI churn prediction. AI analyzes activity and warns about clients at risk.
- Return reminders. Birthdays, contract end, renewal, repeat visit — automatically.
- Segment mailings. A message to a group of clients who haven't bought in a while.
- Upsell suggestions. AI suggests to the manager what else might interest the client.
- Dissatisfaction alerts. If there's negativity in a chat, the owner knows immediately.
On retention strategies see «Client retention»; on repeat sales — «Repeat sales».
What to do with at-risk clients
- Call or write personally. Ask if everything is OK, without selling in the first message.
- Find the reason. Price, quality, timing, or they simply forgot you.
- Offer a solution. A personal offer, a new product, a convenient format.
- Create a task. So the next contact isn't lost.
Why reminders aren't everything
Repeat-visit reminders exist in many CRMs. The difference is that Mestra sees the whole client: chats, payments, complaints and activity. So the risk signal is based on behavior, not just a single date.
Client churn in plain words
Client churn is the share of customers who stopped buying in a chosen period. Internationally it's called churn rate.
The reasons matter as much as the percentage. Some leave over price, others over quality, others simply forget about you.
So reduce churn two ways at once: fix the causes and reach out in time to those who are drifting away.
FAQ
How do I calculate client churn?
Divide the number of clients lost in a period by the number at the start, and multiply by 100%.
Can a CRM predict a client leaving?
Yes. Mestra's AI analyzes client activity and warns about churn risk.
What churn rate is normal?
It depends on the industry and model. It matters more to track your own trend from period to period.
How do I win back a client who stopped buying?
Reach out personally, find the reason and offer a solution. Mestra suggests whom to contact and helps with a segment mailing.
Which plan includes churn prediction?
Mestra's AI features are on the Standard and Pro plans.
Summary
A client rarely leaves suddenly — they leave gradually. Mestra (Местра) notices the signs and tells you whom to go back to right now. Try it free for 15 days.


