What Is Customer Lifetime Value: Why It Is a Margin Forecast Rather Than a Revenue Total, How Cohorts Change It, and What the Number Cannot Tell You

The Number Everyone Quotes and Almost Nobody Calculates the Same Way

Published 2026-09-29 · EuroQuest International

Customer lifetime value is the most quoted number in commercial planning and one of the least standardized. Two teams in the same company will produce figures that differ by a factor of five, both correct according to their own definitions, and neither able to explain the gap in a meeting. The reason is that the phrase describes an idea rather than a formula: the total profit a business expects from a relationship over its whole course. Every word in that sentence is a modeling decision, and the decisions are usually made by whoever built the spreadsheet first.

This explainer covers what the measure is actually for, how it is calculated and where each version breaks, why cohort analysis produces a different and more useful answer than a blended average, how the figure should and should not be used against acquisition cost, and the questions it cannot answer no matter how carefully it is built. It is written for marketing, commercial, finance and customer experience teams who have to agree on one number, and for managers being shown a lifetime value figure who want to know which questions to ask about it.

What Customer Lifetime Value Actually Measures

The definition worth using is narrow: the total profit contribution a business expects to receive from a customer relationship over its remaining course, expressed in today's money. Four things are doing work in that sentence. Profit, not revenue. Expects, which makes it a forecast. Remaining, because value already collected is history rather than value. And today's money, because a contribution arriving in four years is worth less than the same contribution arriving now.

Almost every disagreement about the number traces back to one of those four. A marketing team quoting revenue and a finance team quoting contribution margin are not disagreeing about customers; they are using the same words for different quantities. Settling the definition in writing, once, before anyone builds a model, prevents most of the argument that follows.

It Is a Margin Number, Not a Revenue Number

A lifetime value built on revenue overstates every customer by whatever the gross margin is not, which in most businesses is the majority of the figure. A customer who spends 4,000 US dollars a year at a 30 percent gross margin contributes 1,200, and a model that reports 4,000 is not conservative or optimistic, it is measuring something else entirely and calling it value.

The deductions that belong inside the margin are the ones teams argue about, and the honest answer is that anything varying with the customer belongs in. Cost of goods, payment processing, shipping, and the cost to serve: support contacts, account management time, and the disproportionate handling that certain segments require. Fixed overhead does not belong, because it does not change when one customer leaves. Deciding where that line sits is a finance decision, not a marketing one, and writing it down is what makes the number comparable between teams.

Returns Are Part of the Calculation, Not an Exception to It

Returns are the deduction most consistently left out, and in consumer retail they are large enough to invert a segment's ranking. In the 2025 returns survey published by the National Retail Federation with Happy Returns, a UPS company, retailers estimated that 15.8 percent of their annual sales would be returned that year, totaling 849.9 billion US dollars, against 16.9 percent and 890 billion the year before. The same report put online returns at an estimated 19.3 percent of online sales and found that 9 percent of all returns are fraudulent.

And It Is a Forecast, Not a Record

The most consequential misunderstanding is treating lifetime value as something measured rather than projected. Historic value, what a customer has already contributed, is a fact and can be calculated exactly. Lifetime value is a statement about the future that depends entirely on an assumed retention rate, and that assumption is doing more work than any other input in the model.

This matters because retention assumptions are usually inherited rather than derived. A rate calculated from a period of market growth carries into a period of price competition, and the model keeps producing confident figures while the underlying behavior changes. Switching is real and measurable: Ofcom's 2026 report on pricing and consumer engagement found that 26 percent of United Kingdom households had changed provider for at least one communications service in the previous year, and that fixed broadband switching had risen from 14 percent in 2023 to 18 percent in 2025 after three stable years. A retention assumption set before that shift would have been wrong by a margin that compounds every year it is projected forward. Stating the assumed rate next to the output, every time, is the discipline that keeps the number honest, and it is a habit reinforced by marketing ROI measurement and budgeting practice.

How It Is Calculated, and Where Each Version Breaks

There are three versions in common use, and they are not competing formulas so much as three levels of honesty about uncertainty. Choosing between them is a question of what decision the number has to support.

The Simple Version, and Why It Is Usually Too High

The version on most slides divides average annual contribution margin by the churn rate. A customer contributing 1,200 a year in a base losing 20 percent of customers annually gives 1,200 divided by 0.20, or 6,000. It is arithmetic anyone can check, which is its real advantage, and it embeds three assumptions that are rarely true: that churn is constant over a relationship's life, that contribution stays flat, and that a payment in year seven is worth as much as one today.

The first assumption is the damaging one. Churn is almost never constant. It is highest in the first months and falls as a relationship matures, which is why a flat rate applied to a whole base simultaneously understates established customers and badly overstates new ones. In a business with heavy early attrition, the simple formula produces a figure no cohort has ever achieved.

Discounting, the Step Almost Everyone Skips

Future money is worth less than present money, and lifetime value is mostly future money. Applying a discount rate is what turns a sum of projected contributions into a present value, and skipping it inflates long-dated relationships the most. At a 10 percent discount rate a contribution arriving in year five is worth about 62 percent of its face value, and one arriving in year ten about 39 percent. For a subscription business projecting a decade of revenue, that is not a rounding difference.

Two practical rules make this manageable. Use the same discount rate finance uses for other investment decisions, because a marketing model with its own private cost of capital will not survive scrutiny. And cap the horizon: three years for most consumer businesses, five where contracts genuinely run that long. A model projecting fifteen years is not being ambitious, it is quietly assuming the business and the customer both persist through conditions nobody can forecast. Subscription models make this discipline unavoidable, which is why it is central to subscription models and customer retention work.

The Probabilistic Version, and When It Is Worth the Cost

The third version models each customer's purchase frequency and survival probability separately, fitting the distribution to observed behavior rather than assuming a single rate. It handles the non-contractual case properly, where nobody cancels and a customer is only revealed as lost by their absence, which is the situation most retailers are actually in.

It is more accurate and more expensive, and the honest assessment is that it is worth building when the number drives money at scale: bidding on acquisition, deciding service tiers, or valuing a book of customers in a transaction. For a team trying to establish whether one segment is worth more than another, a cohort-based calculation on margin, discounted, with the retention assumption written on the page, gets most of the way there at a fraction of the effort.

Aggregate Against Cohort: The Distinction That Decides Whether the Number Is Useful

A single blended lifetime value for an entire customer base is the most common form of the measure and the least useful, because of a specific statistical problem rather than a question of precision.

The Survivorship Trap

Measure the average tenure of the customers you currently have and the answer is biased upward by construction, because the customers who left early are not in the sample. The base is composed of survivors, so the average survivor looks durable and the resulting figure describes a population that does not include the failures it should be warning you about. The effect is largest in exactly the businesses that most need the warning: those with heavy early churn.

The correction is to calculate by acquisition cohort. Group customers by the month or quarter they joined, and track each group's retention and cumulative margin over time on its own. This produces a curve per cohort rather than one number, and the curves answer questions the average cannot: whether customers acquired this year behave better or worse than last year's, which channel produces cohorts that persist, and whether a retention initiative changed the shape of the curve or merely coincided with a better intake. Cohort work is also what makes segment definitions defensible, which is the substance of data-driven marketing and customer segmentation.

One warning about cohort curves. A young cohort has not lived long enough to show its full value, so comparing a three-month-old cohort's cumulative margin against a three-year-old one measures elapsed time rather than quality. Compare cohorts at the same age: month six against month six. It sounds obvious and it is the error most often found in a cohort dashboard that has stopped being trusted.

What the Figure Is Genuinely For

Lifetime value earns its keep in a small number of decisions. Outside them it tends to become a slide that everyone nods at and nobody acts on.

Setting What You Can Afford to Pay for a Customer

This is the primary use. If a cohort's discounted contribution is 900 and acquisition costs 300, the relationship pays back three times over its modeled life. The widely repeated target of a three-to-one ratio is a convention rather than a law, and it exists mainly to leave room for the model being wrong. What matters more than the ratio is the payback period: how many months until cumulative margin covers acquisition cost. A business that recovers acquisition cost in five months can grow on its own cash; one that takes twenty-six months is financing growth from somewhere else, whatever the ratio says.

The ratio should be calculated per channel and per cohort, never once for the whole business. Blended acquisition cost against blended lifetime value can look healthy while one channel quietly loses money on every customer it brings in, and the blend is precisely what hides it.

Deciding Who Gets What Level of Service

The second legitimate use is allocating effort. Where lifetime value differs materially between segments, it is a defensible basis for deciding who gets a named contact, priority handling or proactive outreach, and who is served well through self-service. Done openly this is ordinary commercial sense. Done without stating the criterion it becomes an unexplained difference in treatment that is hard to justify when a customer asks, and harder when a regulator does.

That last row is worth dwelling on. A lifetime value model takes current pricing as an input, so using it to justify a price rise is circular unless the retention assumption is re-estimated at the new price, which is a different exercise entirely and the subject of pricing strategies and market positioning.

What It Cannot Tell You

Three limits are worth stating plainly, because each one is routinely crossed.

It cannot tell you why. Lifetime value is an output. It reports that a cohort decays faster without offering any account of what the customers experienced, which is why it belongs next to qualitative work rather than instead of it. The companion guide on mapping a customer journey covers the side that explains the curve; this measure only draws it.

It cannot tell you that service quality is fine. Operational responsiveness and customer loyalty are different variables, and one of the clearest demonstrations available is the US Consumer Financial Protection Bureau's report on 2025, which records that the bureau received more than 6.6 million complaints and sent more than 5.9 million to companies, and that companies provided a timely response to 99.6 percent of them. Speed of response was close to universal while complaint volume stayed in the millions, more than 5.8 million of them about credit or consumer reporting alone, which was 88 percent of the total. A dashboard showing response times will look excellent in that situation and will tell you nothing about whether people want to stay.

It cannot tell you the future will resemble the past. Every version of the calculation extrapolates observed behavior, and spending patterns move underneath it. The US Bureau of Economic Analysis reported that in July 2026 personal consumption expenditures rose 36.3 billion dollars, or 0.2 percent, and that the increase reflected 86.2 billion more spending on services partly offset by 49.9 billion less on goods, with the personal saving rate at 3.0 percent. That is national accounting rather than evidence about any company's customers, and it is quoted here only to make the structural point: a category can be shifting composition while individual retention curves still look stable, and the model will not notice until the cohorts do.

Where Teams Build This Capability

EuroQuest International runs customer experience, loyalty and commercial measurement programs in Dubai, Vienna, Barcelona, Paris and Budapest. Sessions work from participants' own cohort data where they can bring it, so teams leave with a calculation their finance function will accept rather than a template. The wider program sits under marketing management and customer experience, and the service side is covered in customer experience and service excellence and customer loyalty programs and retention.

Frequently Asked Questions