The retention economics of churn

# The retention economics of churn

Take 10 million lines and 2% monthly churn. Multiply by twelve and you get 24%, or 2.4 million departures a year. Compound it properly and the answer changes: 0.98^12 = 0.785, so a cohort left alone decays to 78.5% of itself and loses 2.15 million lines. The 250,000 line gap between the two methods is bigger than the annual net adds of most European carriers, and every budget built on the wrong one is wrong by the same amount.

Now move monthly churn from 2.0% to 1.9%. Annual base decay falls from 21.5% to 20.6%: about 97,000 lines kept, on the same base, for whatever the retention programme costs. This lesson is the arithmetic underneath that: what a save costs, what a win-back costs, and the point at which a retention offer burns more margin than the departure would have.

Why churn is the master metric

Churn is the rate at which subscribers leave over a period, usually a month. Two kinds, with different economics:

• Voluntary: the customer chooses to go. Price, coverage, a service failure, a competitor's port-in credit.
• Involuntary: the customer leaves by accident. Expired card, failed direct debit, suspension for non-payment. Often 20% to 40% of total churn, and the cheapest thing in the building to fix, because a card updater and a smarter retry schedule cost a fraction of what a save desk costs.

Definitions bite here. Prepaid churn depends on where you set the inactivity window: move the cut-off from 90 days to 60 and reported churn jumps without a single customer changing behaviour. Compare your number to a competitor's only after checking they count the same way.

The acquisition trap

Winning a postpaid subscriber can cost several hundred dollars once you add the handset subsidy, the dealer commission, the port-in credit and the media. That is subscriber acquisition costacquisition costCustomer Acquisition Cost (CAC) is the total sales and marketing spend divided by the number of new customers gained in a period. It measures how efficiently you grow.View full definition → (SAC), paid up front, while the margin arrives in monthly slices over years. High churn means paying SAC repeatedly for the same customer slot. A team reporting gross adds without net adds is reporting how fast it refills a leaking bucket.

What a point of churn is worth

The value at stake per line, the expected months of margin, belongs to the lifetime valuelifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → lesson; assume it here as a figure finance already publishes, split by postpaid, prepaid and IoT. Retention economics needs three things from it: margin per line per month, expected remaining months, and how shaky both are.

Work with $12 of monthly margin ($40 ARPU at 30%).

Value preserved by a save = margin x months retained
  $12 x 24 months                              =   $288

Cost of a save = offer + handling
  $5/month x 12 + $6 of call time              =    $66

Cost of a win-back, at list level
  10,000 lapsed customers contacted at $2      = $20,000
  4% reconnect                                 =    400 lines
  $60 reconnection incentive each              = $24,000
  $44,000 / 400                                =   $110 per line
  plus 3 months of margin lost while away      =    +$36

(Illustrative. Substitute your own numbers.)

Two things fall out. A save is cheap because the customer is still on the network and still on the bill run. A win-back is usually cheaper than acquiring a stranger, since you know the old ARPU, the tenure and the exit reason. What breaks it is the response rate: at 1% instead of 4%, the same campaign costs $260 per reconnected line and you would have been better off buying a new customer.

When the offer costs more than the departure

A retention offer goes to whoever accepts it, not to whoever was leaving. That is where the money burns.

Cost per incremental save = offer value divided by incrementality, incrementality being the share of takers who would genuinely have gone. Hand an $8 monthly discount to 200,000 people when 40,000 of them were really leaving and the effective cost is $40 a month per saved line against $12 a month of margin. The programme destroys three dollars for every one it protects. The base did not churn; the margin did.

This is how haggling cultures form. UK consumer press publishes step by step guides to calling Sky and threatening to cancel, and the offers are consistent enough to make the call worth it. Sky sells pay TV and broadband, so the retention discount is a product with a known price, and customers buy it whether or not they intended to leave. Match the offer to the value tier and the risk score, never to the volume of the complaint.

Sometimes the right answer is to let them go. Bharti Airtel and the rest of the Indian industry raised prepaid tariffs sharply in July 2024, then reported subscriber losses for months while ARPU climbed. Shedding the lowest-value lines to hold price on the rest was the plan, not a retention failure.

The save desk

The save desk intervenes when a customer calls to cancel or downgrade, working from a tiered menu: bill credits, a plan upgrade, a loyalty reward, a streaming add-on.

Three numbers matter: save rate, cost per save (offer plus handling), and value preserved. The only honest way to read them is against a holdout, a small slice of at-risk customers randomly given nothing. Carriers that run holdouts routinely find the naive save rate is well above the incremental one. Without a control group you cannot tell a retention programme from a discount programme.

Contract renewals: the quiet churn battle

Churn clusters. It spikes at contract end, and again at device payoff, when the handset instalment drops off the bill and the customer notices they are free and still paying the same. Both dates are known months ahead, which makes them scheduled events.

Reaching someone six weeks before the cliff costs a message and a modest offer. Reaching them after they hold a competitor's quote costs the save desk, and the offer has to be bigger because the alternative now has a price on it.

Structure changes the shape of the cliff. Vodafone's EVO plans in the UK split the airtime agreement from the handset agreement, so the two end dates stop coinciding and no single moment unlocks everything at once. The cost is a more complicated proposition to sell in store.

🎬 [VIDEO: "Customer ChurnCustomer ChurnChurn rate is the percentage of customers or revenue lost over a period. It measures how fast a business loses its existing customer base.View full definition →: A Study of Definition and Measurement" - youtube.com - a clear walkthrough of how subscription businesses define, measure, and reduce churn]

Predicting churn before it happens

Scoring subscribers for churn propensity runs on the behavioural signals the engagement metrics lesson catalogues. The question here is what the score costs to act on.

Ranking matters more than raw accuracy. If your top decile carries five times the base churn ratechurn rateChurn rate is the percentage of customers or revenue lost over a period. It measures how fast a business loses its existing customer base.View full definition →, contacting 10% of subscribers reaches half the leavers. Extend to three deciles and coverage rises while incrementality falls, because most of the extra contacts were staying anyway. Set the threshold where the marginal offer cost equals the marginal margin protected, and recompute it whenever the offer changes.

A false positive is not free: it is a discount handed to a loyal customer, the same leak as before with a model attached. For a primer on how these models are built, see Google Cloud's overview of customer churn prediction, keeping in mind that Google sells the platform the article recommends.

Win-back: reacquiring the ones who left

Segment by exit reason before spending anything.

• Price, or a service failure since resolved: cheap to reverse, the objection has an answer.
• Coverage at home or at work: no offer fixes it until the network does.
• Structural: Vodafone Germany lost a large share of its cable TV households when the German rule letting landlords bill TV through building service charges ended in mid-2024. No save desk reverses a change in the law. The only route was converting those households to individual contracts, an acquisition campaign wearing a retention badge.

Cap the offer at the value you expect to recover, respect cooling-off periods and consent rules (GDPR-style opt-in in many markets), and watch the clock. Six months out, the leaver has a new handset, a new contract and a new habit, and your win-back is now competing with a competitor's own save desk.

Putting it together

1. Measure churn precisely, voluntary and involuntary apart, and pin the definition of an inactive prepaid line before comparing with anyone.

2. Fix involuntary churn first. Failed payments and expired cards are the cheapest saves available.

3. Score for risk and value together, and act before the contract-end cliff rather than after the cancellation call.

4. Price every offer against the margin it protects, not against the threat that produced it.

5. Run holdouts, so cost per save means cost per incremental save.

6. Win back only where the exit reason has an answer.

One question sits under all six: does the margin preserved exceed the margin given away, counting everyone who takes the offer and not only those who were leaving?

Key Takeaways

• Compound, do not multiply: 2% monthly churn is a 21.5% annual base decay, and a tenth of a point on 10 million lines is roughly 97,000 subscribers a year.
• Involuntary churn runs at 20% to 40% of the total in many carriers and is the cheapest slice to recover.
• Cost per incremental save = offer value / incrementality. An $8 discount taken by five people to keep one is $40 a month against $12 of margin.
• Win-back economics turn on response rate rather than offer size: the same list at 4% and at 1% differs by more than double per reconnected line.
• Some churn is bought on purpose. Airtel's 2024 tariff rise traded low-ARPU lines for price on the rest.