Loyalty & retention: foundations & core concepts

Tesco knew what it sold in 1994. It did not know who bought it. In February 1995 it launched Clubcard: a point for every pound, a voucher posted back each quarter, and, quietly, a name attached to every basket. Chairman Ian MacLaurin's reaction to the first analysis, that these outsiders knew more about his customers after three months than he had learned in thirty years, has been repeated at conferences ever since. Within about a year Tesco had passed Sainsbury's to become Britain's largest grocer. Two things happened at once, and they are still the two things any loyalty programme does: a slice of margin was traded for repeat visits, and anonymous transactions became identified histories.

Bain's retention research put the economics bluntly: a 5% increase in customer retention raises profits by 25% to 95%, depending on the industry. Acquisition gets the budget. Retention gets the compounding.

What retention, churn and loyalty actually mean

Retention is behavioural, and it is always measured against a window: the share of customers active at the start of a period who are still active at the end of it. Define "active" first (a purchase, a login, a paid month), because the same customer base can show 40% or 80% retention depending on which definition you pick.

Churn is that measurement read from the other end: the share who stopped. Voluntary churn is a decision, someone cancels. Involuntary churn is a failure, a card expires and the renewal declines. At subscription scale the second category is large, unglamorous and fixed by billing plumbing rather than by campaigns.

Loyalty is attitudinal: preference that survives a cheaper or more convenient alternative. Retention without loyalty is ordinary (a 24-month contract retains you whether you like the brand or not). Loyal customers are retained, resistant to competitive offers, and more likely to bring someone else with them.

A loyalty programme is a standing exchange offered to identified customers: recognisable value (points, member prices, tiers, access) in return for repeat behaviour and the data that identifies it. It has two costs and produces one asset. The costs are the margin given away and the operation needed to run it. The asset is the identified purchase history.

Sub-concept 1: customer lifetime valuecustomer lifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition →, the number everything else feeds

CLV is the total net profit expected from a customer across the relationship: average purchase value x purchase frequency x expected lifespan, minus acquisition and servicing cost. In a subscription, lifespan falls straight out of churn, as 1 divided by the periodic 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 →. Third-party subscription trackers put Netflix's monthly churn at roughly 2%, the lowest of the large US streamers, which implies an average paying life of around 50 months. A rival churning at 5% a month holds the same customer for 20 months. Same price, same content bill each month, less than half the lifetime revenue. That ratio explains why a streaming service will spend heavily to remove a reason to cancel and still refuse to buy a subscriber with a discount that expires at renewal.

Sub-concept 2: churn rate is the symptom, churn cause is the diagnosis

Most boards see one churn number a quarter. Few marketing teams can name the reasons behind it with exit data attached. The distinction is operational, because the fixes sit in different departments: price objections go to commercial, declined cards go to billing, "I finished what I came for" goes to product.

Netflix's 2023 paid-sharing enforcement is the case worth holding on to. Charging for viewers outside the household was widely predicted to trigger a cancellation wave. Cancellations came, and then Netflix added close to 9 million paid memberships in the third quarter of 2023 and closed the year at roughly 260 million. For most borrowed-password users, the obstacle had never been willingness to pay. It was that nobody had asked.

Sub-concept 3: the loyalty loop against the purchase funnelpurchase funnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition →

McKinsey described the consumer decision journey in 2009, and the loop it identified still holds up. A funnel ends at purchase. A loop routes the post-purchase experience back into consideration and skips awareness entirely. Tesco's shift from collecting points to Clubcard Prices, member-only prices printed on the shelf edge, is loop machinery: the reason to come back is visible at the moment of choosing, not three months later in an envelope. A member who opens the app in the aisle did not enter your funnel. They re-entered a loop you built, at a marketing cost per visit that no acquisition channel can match.

Sub-concept 4: bought, built-in and earned loyalty

Bought loyalty runs on points and member prices. It works, it is measurable, and it can be repriced by any competitor with a spreadsheet. Built-in loyalty runs on switching costs: over three billion people use a Meta app every day, and leaving takes your contacts, your photo history and a decade of conversations with you. Meta pays no points to anyone. Earned loyalty is preference: the customer would still choose you at parity. Most real programmes mix all three, and the useful question is which one is actually holding the customer, because only the third one survives a rival matching your discount.

Real-world case 1: Tesco Clubcard, the reference design

The mechanics were deliberately simple: one point per pound, 150 points converted into a £1.50 voucher, statements mailed quarterly, and reward partners who would take the voucher at a multiple of its face value so Tesco's cost stayed lower than the perceived gift. The analysis behind it came from dunnhumby, a firm that sells exactly this kind of customer data work; Tesco bought control of dunnhumby in 2001 and the remainder a decade later, which tells you where the value sat. Clubcard now covers north of 20 million UK households and the data drives range decisions, store formats and promotions, not just the vouchers. The reason it stayed relevant for thirty years is that Tesco kept changing what the card pays out (points, then coupons, then member prices) while never giving up the identification.

Real-world case 2: Netflix, retention without a programme

Netflix has no tiers, no points and no members' club. Its retention instruments are the price ladder (the ad-supported plan launched in November 2022 gave price-sensitive viewers somewhere to go that was not the exit), the release schedule, and the payment machinery that quietly recovers failed cards. This is what retention looks like when the product is bought again every 30 days: the cancel button is one click away, and every decision about content spend is an argument about lifespan.

Real-world case 3: Meta, where retention is the business model

Meta sells no subscription and runs no rewards scheme, yet it reports daily and monthly active people because the ratio between them is its retention number. Roughly $165 billion of 2024 revenue rests on attention that comes back tomorrow, and every product decision (feed ranking, notifications, Reels) is a retention decision priced in advertising. Useful counterweight to the assumption that loyalty work means launching a card: in some businesses the programme is the product.

CMO action items

• Write down your definition of an active customer and the window you measure it over, then re-run retention on that definition. If marketing, finance and product are using three different ones, fix that before anything else.
• Split last quarter's churn into voluntary and involuntary before approving a single win-back campaign. The involuntary half rarely needs a marketing budget.
• Calculate CLV per acquisition channel using a lifespan derived from that channel's own churn rate rather than a blended average. Channels almost always differ by a factor of two or more, and blended CLV hides it.

Common mistakes that kill results

• Reporting a retention rate with no stated definition of activity. The number moves when someone changes the window, and nobody notices.
• Counting enrolled members instead of identified sales. Cards issued is a vanity figure; the share of revenue that arrives attached to a known customer is the one that tells you whether the programme is working.
• Calling a discount a programme. If you reduce the reward and behaviour falls with it in the same period, you were renting purchases, not building preference.
• Treating loyalty as a marketing-only asset. Churn caused by a declined card is a billing problem, churn caused by a dead-end onboarding is a product problem, and a CMO who is not in those rooms is reporting on retention rather than managing it.