Advanced marketing · Module 2

Customer economics: segmentation, acquisition and lifetime value

Which customers are worth acquiring, what is a customer worth over the whole relationship, and how should segments be drawn so that they predict what people will do?

Beyond the IB course UndergraduateMBAProfessional About 18 minutes · 2,804 words · checked 16 November 2026

From the IB course to this module

The IB course covers market share (4.1), segmentation and targeting (4.2) and market research and sampling (4.4), and the Market Studio's market metrics tool introduces customer lifetime value as enrichment. This module develops lifetime value properly, with retention, margin and discounting worked through, sets out the economics of acquiring customers and the traps in common ratios, and asks what makes a segmentation useful.

Section 01: Conceptual foundation

A customer relationship is an asset. Its value is the stream of margins the customer will generate, discounted to the present, less the cost of acquiring and serving them. Summed across current and future customers this is customer equity, and Gupta, Lehmann and Stuart (2004) showed that it can be used to value whole companies. Thinking this way changes marketing's questions: not how many customers were won, but whether the customers won cost less than they are worth.

Two settings need different methods. In contractual settings, such as subscriptions, customers must cancel, so defection is observed and retention can be measured directly. In non-contractual settings, such as most retailing, customers simply stop buying, so whether a quiet customer has left or is between purchases must be inferred from past behaviour (Schmittlein, Morrison and Colombo, 1987).

Segmentation divides a market into groups that respond differently to marketing. Smith (1956) presented it as an alternative to product differentiation: rather than bending demand to fit the product, adjust the offer to differences in demand. A useful segmentation is measurable, large enough to serve profitably, reachable and above all actionable: the segments must call for different marketing and must predict behaviour, not merely describe people.

Section 02: Advanced explanation

Lifetime value, worked through

The simplest useful model assumes a constant annual margin m from each active customer, a constant retention rate r (the share of customers still active a year later) and a discount rate i. If a new customer yields m at the end of the first year and survives into each later year with probability r, lifetime value is:

CLV = m ÷ (1 + i − r)

A worked example: a margin of £120 a year, retention of 80% and a discount rate of 10% give CLV = 120 ÷ (1.10 − 0.80) = £400. Year by year the present values are £109, £79, £58 and £42 in the first four years; the rest of the £400 comes later.

Retention is the most powerful lever. At 85% retention CLV is 120 ÷ 0.25 = £480, 20% more; at 90% it is £600. Expected lifetime, 1 ÷ (1 − r), rises from five years to ten. The reverse matters as much: a price rise that lifts margin by 5.7% is cancelled if retention slips from 80% to about 78.3%, since 126.84 ÷ (1.10 − 0.783) is again about £400.

If acquiring the customer costs £250, each new customer creates £150 of value, but the discounted £250 is recovered only in the fourth year (cumulative present values of £109, £188, £246 and £288). That is why fast-growing subscription businesses can be valuable and cash-hungry at the same time.

The constant-retention model has a known flaw. Customers differ, and the least committed leave first, so a cohort's retention rate usually rises with tenure. A single average rate understates the value of surviving customers (Fader and Hardie, 2010). Retention curves by cohort and tenure are the honest basis for lifetime value.

Acquisition economics and the ratio trap

Customer acquisition cost (CAC) is acquisition spending divided by the new customers it wins; payback is the time a customer's margin takes to repay it. Many firms track the ratio of lifetime value to acquisition cost and aim for a figure such as three to one. The ratio is a handy summary with no theoretical basis, and it is easy to flatter:

  • Revenue instead of margin: lifetime value must be built on contribution, not sales.
  • No discounting, which overstates long relationships.
  • Averages instead of the margin: the next customer from a channel usually costs more than the average, because the cheapest customers are reached first.
  • Credit for customers who would have come anyway: if advertising takes credit for customers who would have arrived through search or word of mouth, its true incremental cost is hidden.
  • Blended channels: a healthy average can hide a channel that destroys value.

How customers are won also affects what they are worth. Lewis (2006) found that customers acquired with deep promotional discounts were less likely to stay than those acquired at regular prices, so cheap acquisition can prove expensive.

How much to spend on acquisition rather than retention depends on the marginal return to each. Blattberg and Deighton (1996) proposed a customer equity test: spend on each until the last pound returns a pound of customer equity. Reinartz, Thomas and Kumar (2005) found that getting retention spending wrong cost more long-term profit than getting acquisition spending wrong.

Segmentation that predicts behaviour

Segments can be built on geographic, demographic, psychographic, behavioural or needs-based variables. Yankelovich and Meer (2006) argued that many segmentations fail because they describe people in ways that do not predict what they buy: demographic and attitudinal segments are easy to name and hard to act on. Segmentations built on behaviour and on the job a customer is trying to get done (Christensen, Hall, Dillon and Duncan, 2016) tend to serve product and channel decisions better.

Analytically, segments are often found with cluster analysis or latent class models applied to survey or transaction data. Within an existing customer base the workhorse is RFM, the recency, frequency and monetary value of past purchases, which predicts future value well because past behaviour is the best guide to future behaviour.

For retention, segmenting by risk is not the same as segmenting by response. Ascarza (2018) showed in field experiments that the customers most likely to leave were not necessarily those whose behaviour a retention offer changed most, so targeting by risk alone could waste the budget. Uplift modelling, which uses experimental data to estimate each customer's response to an offer, targets those whose behaviour will actually change.

Segmentation now meets legal and ethical limits. Offers and prices personalized from detailed data can be efficient, but they can also exploit inattentive or vulnerable customers, which regulators increasingly examine.

Customer equity, loyalty and the value of the firm

Treating customers as assets connects marketing to finance. Rust, Lemon and Zeithaml (2004) proposed judging marketing by its effect on customer equity, so that a campaign, a service improvement or a loyalty scheme is assessed by the change in the value of the customer base. McCarthy, Fader and Hardie (2017) showed how to value subscription businesses from the customer data companies disclose, such as subscriber numbers, churn and revenue per user.

Two cautions temper enthusiasm for loyalty. Long-lived customers are not always profitable: Reinartz and Kumar (2002) found that many loyal customers were costly to serve or bought only on discount, while some short-lived ones were highly profitable, so loyalty should be managed for profit, not for its own sake. And loyalty schemes rarely create much loyalty: Sharp and Sharp (1997) found only small effects of a large programme on repeat-purchase patterns. Their value lies more in the data they generate, the targeted offers they allow and, increasingly, the advertising they make possible.

Disclosed metrics need care. Each company defines an active customer or a subscriber in its own way, and a change of definition, or a choice between counting at a point in time and over a period, can move the numbers more than any change in behaviour.

Section 03: Models and frameworks

Where each comes from, how to use it, and where it stops being useful.

Gupta and Lehmann (2003)

Customer lifetime value

With constant margin m, retention r and discount rate i, lifetime value is m ÷ (1 + i − r), a multiple of annual margin that rises steeply with retention. Customer equity is the sum across current and future customers.

How to use it

Compute lifetime value by acquisition channel and cohort, on contribution rather than revenue, and compare it with the marginal cost of acquisition. Use the formula to test how much retention a price rise or a service cut can afford to lose.

Where it fails

Constant margin and retention are rarely true: heterogeneity makes cohort retention rise with tenure, and margins change as customers mature. Small errors in retention produce large errors in value.

Sunil Gupta and Donald R. Lehmann (2003). Customers as assets. Journal of Interactive Marketing. Sunil Gupta, Donald R. Lehmann and Jennifer Ames Stuart (2004). Valuing customers. Journal of Marketing Research.

Schmittlein, Morrison and Colombo (1987); Fader, Hardie and Lee (2005)

Counting your customers

In non-contractual settings, models such as the Pareto/NBD and the simpler BG/NBD use each customer's recency, frequency and length of history to estimate the probability that the customer is still active and the purchases to expect.

How to use it

Fit the model to transaction histories to rank customers by expected future value, to decide whom to contact, and to forecast a cohort's purchases. It needs only data most firms already hold.

Where it fails

The models assume stable buying rates and treat customers' departures as unobserved and permanent; promotions, seasonality and returning customers break the assumptions. They predict purchases, not the effect of marketing on them.

David C. Schmittlein, Donald G. Morrison and Richard Colombo (1987). Counting your customers: who are they and what will they do next?. Management Science, 33(1). Peter S. Fader, Bruce G. S. Hardie and Ka Lok Lee (2005). 'Counting your customers' the easy way: an alternative to the Pareto/NBD model. Marketing Science, 24(2).

Smith (1956)

Market segmentation

Segmentation adapts the offer to differences in demand, as an alternative to differentiating the product and persuading the whole market. Segments must be measurable, substantial, reachable and actionable.

How to use it

Build candidate segments on behaviour and needs, then test whether they respond differently to price, product or channel. Keep a segmentation only if it changes decisions; describe each chosen segment with a persona so that the whole team can design for it.

Where it fails

Segments found statistically can be unstable and hard to reach in media. Many segmentations end as attractive descriptions that change nothing; finer targeting also raises fairness and privacy concerns.

Wendell R. Smith (1956). Product differentiation and market segmentation as alternative marketing strategies. Journal of Marketing, 21(1). Daniel Yankelovich and David Meer (2006). Rediscovering market segmentation. Harvard Business Review, 84(2).

Blattberg and Deighton (1996)

The customer equity test

Spend on acquisition and on retention until the last pound of each returns a pound of customer equity. Both face diminishing returns, so the best split depends on the marginal return to each, not the average.

How to use it

Estimate response curves for acquisition and retention spending, ideally from experiments, and shift money from the one with the lower marginal return to the other until the returns are equal.

Where it fails

Response curves are hard to estimate reliably, and the test treats acquisition and retention as independent, though acquired customers' quality depends on how they were won.

Robert C. Blattberg and John Deighton (1996). Manage marketing by the customer equity test. Harvard Business Review, 74(4).

Section 04: Real business applications

Real organizations appear only with dated, cited facts; illustrative examples are labelled.

Real organization, 2025-03

MultiChoice: price rises and a shrinking subscriber base (2025)

In results for the year to 31 March 2025, MultiChoice, the African pay-television group, reported that linear subscribers had fallen by 1.2 million, or 8%, to 14.5 million, with losses split evenly between South Africa and the rest of Africa. In South Africa prices rose by about 5.7%, yet subscription revenue fell 3% as the active base shrank to 7.0 million; average revenue per subscriber rose 4%, from ZAR281 to ZAR292 a month. Management blamed the cost-of-living crisis, saying many households had no choice but to give up their subscription for the time being, and the shift to cheaper streaming, social media video and free or pirated video. Its own streaming service, DStv Stream, grew subscribers by 38%. The group also counted 18.6 million subscribers active at some point in the 90 days to the year end, against 14.5 million at the year end itself.

Interpretation. Rising revenue per subscriber can hide an eroding base: in the lifetime value formula a higher margin is easily cancelled by lower retention. The gap between the two counts reveals customers who leave and return, where the definition of active shapes the story.

Real organization, 2026-08

Tesco: a loyalty scheme as an engine of data and pricing (2026)

Tesco's interim results for the 26 weeks to 29 August 2026 describe its Clubcard scheme as "the glue that holds the whole Tesco ecosystem together". The company reported extending Your Clubcard Prices, prices personalized to the individual, to around 2.5 million customers, alongside more than 10,000 Clubcard Prices open to all members and about 700 products price-matched to Aldi. Subscribers to Delivery Saver, its delivery subscription, rose 6.1% to 836,000, and average weekly orders in its grocery home shopping business grew 4.9%.

Interpretation. The scheme works less as a creator of loyalty than as an engine of data and targeted pricing: member prices reward customers for identifying themselves, personalized prices segment them one by one, and a delivery subscription turns a fee per order into a commitment that makes each extra order cheaper.

Illustrative

A cohort curve changes a software firm's numbers

A software company tracks customers by the quarter they joined. Of 1,000 who joined in one quarter, 700 remain after a year, 600 after two and 550 after three: retention of 70%, then 86%, then 92%. Its board papers had valued customers as if three in ten left every year. Valued with its own curve, each cohort is worth much more, but the curves also show that recent cohorts start lower, a decline the average had hidden.

Section 05: Case studies

Fictional case, simulated figures

Pebble & Pine: which customers are worth buying?

Pebble & Pine, a fictional subscription box for houseplants, wins customers through three channels: paid social media advertising, a referral scheme that rewards existing customers, and a discount partner that offers a heavily reduced first box. The marketing team reports a healthy blended position and wants to double spending on paid social. The finance director asks for the economics by channel.

Customer acquisition by channel, last quarter
Channel New customers Acquisition spend (£) Monthly margin per active customer (£) Monthly retention
Paid social4,000180,0009.0085%
Referral scheme1,50030,0009.0092%
Discount partner3,00060,0006.0075%

Simulated figures. Margin is contribution after product, packing and delivery. Retention is the share of customers active in one month who are still active the next, measured over customers' first three months.

Questions

  1. Calculate the acquisition cost per customer and the lifetime value for each channel, using a discount rate of 1% a month. Which channels create value?
  2. Calculate the blended acquisition cost and the customer-weighted average lifetime value. Why do these figures mislead?
  3. Retention was measured over customers' first three months only. In which direction is that likely to bias the lifetime values, and what data would you ask for?
  4. Doubling paid social spending is expected to raise the acquisition cost of the extra customers to £60. Should the company do it?

Section 06: Critical evaluation

What the evidence supports, where the debate is, and what the models leave out.

What is established. Retention is the dominant driver of lifetime value in subscription businesses, and customer heterogeneity makes simple constant-retention models misleading; probability models built on recency and frequency predict future purchases well in many settings. That customers acquired through deep discounts tend to be worth less is supported by field evidence.

What is overstated. Reichheld and Sasser (1990) argued that small gains in retention can raise profits sharply, and the idea became a slogan. Later work found that the relationship between loyalty and profit is weaker and more varied than the slogan implies: long-lived customers are not always profitable, and loyalty programmes produce modest behavioural effects. Rules of thumb such as a three-to-one ratio of lifetime value to acquisition cost have no theoretical basis and depend entirely on how both are measured.

Measurement problems. Acquisition costs are usually attributed rather than measured experimentally, so they include customers who would have come anyway. Disclosed metrics such as subscribers and revenue per user are defined differently by each company and sometimes redefined, which limits comparisons.

Segmentation. Statistical segments can be unstable from one study to the next, and many segmentations never change a decision. The evidence that targeting by predicted risk can waste retention budgets is a warning against confusing prediction with influence.

Ethics and law. The same data that make customer economics precise allow firms to charge different customers different amounts for the same service. Where that rests on inattention or vulnerability rather than cost or value, regulators have intervened, as the pricing module shows.

Section 07: Decision exercise

A fictional situation. Decide before you read the guidance.

Where should the next £2m go?

Pebble & Pine has £2m to spend next year beyond its current budget. Its analysts have built a model that predicts which customers are likely to cancel, and the referral scheme has the best economics of any channel but has never been scaled. The board must choose where the money goes.

  1. A. Scale acquisition

    Spend the £2m on paid social and other paid channels to grow the customer base as fast as possible.

  2. B. Retention by risk

    Spend it on offers and service calls to the customers the churn model ranks as most likely to cancel.

  3. C. Test, then target

    Spend a fifth on randomized tests of retention offers and referral rewards, then put the rest where the tests show the largest incremental gain in customer equity.

Judge the options on

  • Incremental customer equity per pound
  • Confidence in the measurement
  • Risk of paying customers who would have stayed or come anyway
  • Scalability of the channel
Read the guidance

The case exhibit already shows the danger in A: paid social barely creates value at its current cost and destroys it at the higher marginal cost of scaling. Growth bought this way is growth in customers, not in customer equity.

B sounds rigorous but targets by risk rather than response. Many of the highest-risk customers will leave whatever they are offered, and some offers will go to customers who would have stayed anyway; experiments in other businesses have found exactly this.

C is usually the strongest choice. It spends a small share to learn where the money changes behaviour, then applies the customer equity test to the rest. The referral scheme is the obvious candidate to scale, but referral economics usually worsen with scale as the most enthusiastic customers are tapped first, so it should be tested at higher reward levels before the budget is committed. The cost of C is a delay of a few months, which matters little compared with the cost of spending £2m on the wrong customers.

Section 08: Further reading

  • Sunil Gupta, Donald R. Lehmann and Jennifer Ames Stuart (2004). Valuing customers. Journal of Marketing Research.

    Shows how customer lifetime value can value whole companies; the clearest bridge between marketing and finance.

  • Peter S. Fader, Bruce G. S. Hardie and Ka Lok Lee (2005). 'Counting your customers' the easy way: an alternative to the Pareto/NBD model. Marketing Science, 24(2).

    A practical model for estimating which customers are still active, with data most firms already hold.

  • Werner Reinartz and V. Kumar (2002). The mismanagement of customer loyalty. Harvard Business Review, 80(7).

    Why loyal customers are not always profitable ones.

  • Robert C. Blattberg and John Deighton (1996). Manage marketing by the customer equity test. Harvard Business Review, 74(4).

    The customer equity test for balancing acquisition and retention.

  • Daniel Yankelovich and David Meer (2006). Rediscovering market segmentation. Harvard Business Review, 84(2).

    Why so many segmentations fail, and how to build ones that guide decisions.

  • Eva Ascarza (2018). Retention futility: targeting high-risk customers might be ineffective. Journal of Marketing Research, 55(1).

    Field experiments showing why targeting the riskiest customers can waste a retention budget.

Every work cited here is listed in the platform's sources register. Real organizations appear only with facts from the sources shown; everything else is labelled illustrative or fictional.