Loyalty programme consolidations are sold as retention investments. In practice most of them are platform migrations with a marketing wrapper, approved on a business case that nobody has tested against the reasons customers actually give for leaving.
PLDT and Smart have just unified their loyalty programmes on Capillary, the latest in a run of operator loyalty consolidations justified on retention grounds. The logic is familiar: one programme, one data layer, better personalisation, lower churn. The first three are engineering outcomes and they are achievable. The fourth is an assumption.
Does a loyalty programme reduce churn?
Sometimes. Rarely as much as the business case claims, and almost never for the reasons the business case states.
When you run a proper churn diagnosis in a subscription or retail business — CRM and billing data first, then interviews with frontline staff, then short targeted research with customers who have actually left — the drivers cluster in three places:
- Price perception. Not price. Perception. The customer concluded they were paying more than the thing was worth, or more than the new customer down the road.
- A service failure that was never closed. An outage, a billing error, a support loop. The cancellation often comes weeks later, which is why the data alone makes it look unrelated.
- A competitor's bundle. Someone else packaged the same utility with something the customer was already buying, and the maths changed.
Rewards appear in these conversations. They almost never appear first. They show up as a tiebreaker — "I had points, so I stayed another year" — which is a real effect, but a marginal one, and it is a different claim from "the programme reduced churn".
A loyalty programme can tip a customer who is already ambivalent; it cannot argue with a bill the customer has decided is too high.
Why the loyalty business case survives anyway
Three reasons, and none of them are bad faith.
First, the measurement is self-confirming. Members churn less than non-members in nearly every base. That is selection, not causation: heavy users, long-tenure customers and people who have already decided to stay are the ones who enrol. Comparing members to non-members tells you who joined, not what joining did.
Second, the alternatives are harder to own. Fixing price perception means a pricing conversation with finance. Fixing service failures means operations. Fixing a competitor's bundle means product and partnerships. A loyalty platform sits inside marketing's remit and has a vendor who will help write the deck.
Third, consolidation genuinely is sensible when you have three programmes and four databases. The cost and data arguments usually hold. It is the retention number bolted onto the end that does the heavy lifting in the approval meeting, and it is the number nobody is accountable for afterwards.
The affordability backdrop
There is a reason this matters more in 2026 than it did in 2021. In several markets operators are openly repositioning around affordability rather than loyalty mechanics — Business Day has reported Cell C shifting focus to affordability as cash-strapped consumers reshape the South African telecom market (https://news.google.com/rss/articles/CBMi2wFBVV95cUxOb3VPeUxYQkRQZFZPRDNsZUNnNExLQ3psWmlhWWloQ21jSnlSOWJJTS14bXd3ZEF3UWswbjJKekZDV0FQQ2FmM1NnSTBsb21raDN1cE93V2RVSmZBRUtLMnhVQ0RmZnI2WC11VXYySC1NM25UU1R3RURodWJKWlRCSGRJMHRWd19rWlRWSGs5dTVjQzFFVXRrNXpNSjFaOU9kenBKNDR2NkM1U1Z6T1ZUdTlJX3pPd1JaT0tCOFB4TmI3ODRWOHJJb25PNXdDY19sTEtRWHhyblIxa1E?oc=5).
That is one operator in one market, and Nordic conditions are not South African conditions. But the direction is worth noting: when household budgets tighten, the competitive battleground moves towards the headline price and the bundle, which are precisely the two things a points programme does not address.
Run the diagnosis first
The sequence is not complicated. It is just unpopular, because it can produce an answer that reduces the loyalty budget.
Step one: quantify who leaves, from your own data. Churn by tenure, by product depth, by discount history, by whether a service ticket was opened in the ninety days before cancellation, by whether the customer had a price change. You are not looking for a model. You are looking for two or three cohorts that behave differently enough to deserve different treatment.
Step two: ask the frontline. Retention agents, store staff and account managers can usually name the top three reasons customers leave before you finish the question. Their account is biased — they hear the stated reason, not always the real one — but it is fast, free and it sharpens the research design.
Step three: talk to people who actually left. Short, targeted, properly incentivised. Not an exit survey in the cancellation flow, where the customer is irritated and clicking the fastest option. A conversation, two to four weeks later, that reconstructs the decision: when they first considered leaving, what triggered it, what they compared, what would have changed their mind.
Step four: rank the drivers by addressable value. Not by frequency. A driver that affects twelve per cent of churners but can be fixed with a packaging change is worth more than one affecting thirty per cent that requires rebuilding your network.
Only then do you ask the loyalty question: of the churn we have just mapped, how much of it is plausibly influenced by rewards, and is that the best available use of this retention euro?
What a good outcome looks like
Sometimes the answer is yes, proceed. If your diagnosis shows churn concentrated among low-engagement customers with no strong price objection and no service history, a well-designed programme has something to work with.
More often the answer is split. Consolidate the platforms on the cost and data case, which stands on its own, and move the retention claim — and part of the budget — to whatever the diagnosis actually surfaced. That might be a proactive outreach to customers whose price increased above a threshold, a service recovery protocol with teeth, or a counter-bundle.
The point is not that loyalty programmes do not work. It is that you should not spend eighteen months and a seven-figure budget finding out whether they work on a churn problem you never defined.