Telecom churn: retention calls before they leave

Churn is not decided in a day

A customer does not wake up one morning and change operator. The decision matures over weeks: dissatisfaction appears, an alternative is looked at, it is discussed with someone, and only then is the paperwork filed. The company usually sees only that last step.

That explains why retention calls arrive late. If the customer has already decided, the call's job is to reverse a decision — which is always expensive and usually ends in a discount. A call made while the decision is still forming is a different conversation: there the problem can still be solved.

This article is about signal-based retention: which signals justify a call, how signals are prioritised, how the script is built, and what to do to stay out of the discount trap.

Which signals justify a call

In telecom most of the signals are already in the system — nobody turns them into a calling queue. Five of them work in practice.

1. 1A sharp drop in usageIf calls, traffic or service usage fall for several weeks running, the customer has either moved or started shifting to an alternative. This is the earliest signal there is.
2. 2Repeat contact with supportA customer who has come back two or three times with the same problem is already in the risk group. Even when the problem is solved, the experience remains.
3. 3Late payment, or a question about cheaper tariffsA change in payment behaviour, or a request for a cheaper plan, shows the decision is maturing.
4. 4Approaching the end of a contractWhen a contract ends the customer looks at the market anyway. This is the one signal whose date is known in advance and can be planned for.
5. 5A question about number portabilityThis is no longer a signal — it is a decision. The call has little chance here, but learning the reason is valuable for later.

Reacting to all five the same way is a mistake. The first three are early signals and the conversation there is about a problem; the last two are late signals and the conversation is about terms.

From signal to call: prioritising

Calling on every signal is neither possible nor useful. The queue is built from two dimensions: the customer's value and the level of risk.

• High value + early signal: call immediately, with the most experienced team
• High value + late signal: call, but always ask the reason before making an offer
• Medium value + early signal: a call or a message, in campaign form
• Low value + late signal: an automated channel is enough; calling capacity is not spent here
• In every case: a customer who has had a retention call in the last 30 days does not get another

That last line is the system protecting itself. Two retention calls to the same customer in a month do not retain them — they prompt them to act, because the call itself confirms there is a problem.

The script: question first, offer second

The most repeated mistake on a retention call is opening with a discount. Said in the first sentence, it turns the call into a price negotiation, and the customer learns to repeat it every time.

1. 1State the reason for the callShort and honest: we noticed a change in your usage, or we wanted to ask how things went after your last contact. This separates the call from a sales call.
2. 2Ask the reasonOne open question: what changed. The answer usually falls into one of four groups — price, quality, service, a change in need. Without it any offer is a guess.
3. 3Offer the step that matches the answerA quality problem needs a technical fix; a tariff problem needs the right plan; a service problem needs a handover to the person responsible. A discount is one option among these and always the last.
4. 4Outcome and follow-throughWrite the outcome code, create a task if there is a technical issue, and tell the customer when they will be called back. A promise that is not tracked makes the situation worse rather than better.

How a sales opportunity is recognised inside a support call is covered in cross-sell on support calls — there the goal is an extra sale, here it is retention, and the two scripts should not be mixed.

The discount trap

A discount is the fastest retention instrument and the most expensive one. The problem is not the immediate cost — it is the behaviour it teaches: a customer retained once with a discount opens the same way next time.

• A discount is offered only when the reason is price, and not otherwise
• Its duration and conditions are stated plainly — an open-ended concession dismantles the tariff policy
• Value is offered as an alternative: extra volume, a service, a device, contract terms
• Customers retained with a discount are tracked separately — they are in the risk group again next period
• The level of discount is set by policy rather than by the script, and an operator cannot raise it

That last line is a management question. When a retention discount is a free parameter of the script, average revenue drifts down quietly and it takes months for that to appear in a report.

Limits and care

Retention calls happen in a regulated sector, and that constrains several parts of the script. The points below are not legal advice — they are practical items to settle alongside the company's own rules and its regulatory obligations.

• Information about contract terms, penalties and portability comes only from approved text
• If the customer has clearly said they are leaving, there is no pressure — one offer, one answer
• Call recording and data handling follow the company's rules
• Retention calls are not scheduled for nights or early mornings
• Nothing is promised that cannot be delivered: phrases like «the network will improve» do not belong in a script

How Vexvon supports this flow

Vexvon's AI voice agent calls from each company's own scenario, approved information and escalation rules. The parts signal-based retention uses:

• The target list is built from a CRM filter: separate campaigns by signal type, value segment and date of last contact
• The scenario defines the fields to extract: reason category, description of the technical problem, the agreed step, the callback date
• When a condition is met the call is handed to an operator or to the technical team
• Reminders are checked every minute — a promised callback is not lost
• Every call, note and status change appears on one timeline

Date-driven retention — where the moment is known in advance, as with a contract ending — follows the same logic as the model in insurance renewal reminder calls. Recovery after an incident is in the negative review recovery call.

Measurement: the save rate is not the only number

The result of a retention programme is read in four numbers, and one of them is usually forgotten.

• Save rate — what share of the called risk group stayed
• A control group — the result for a comparable segment that was not called; without it the save rate says nothing
• Signal precision — what share of the customers a signal flagged were genuinely preparing to leave
• Share retained by discount, and how those customers behave in the next period

The second line is the most skipped and the most important: without an uncalled control group there is no way to claim the programme works. How the numbers are assembled into a report is shown on the analytics page.

To start, pick one signal — a drop in usage, say — and run it on one value segment for two weeks. To build this flow, get in touch.