Unit Economics Wiki

Customer Lifetime Value

Customer lifetime value is the present value of a customer's expected future gross profit — a forecast built on retention, margin, and discounting assumptions, not an observed fact.

Unit EconomicsUpdated Aug 13, 202611 min read

Snapshot

What it is

The present value of the future gross profit a customer is expected to generate, weighted by the probability that the customer is still there to generate it.

Why it matters

LTV sets the economic ceiling on acquisition spend, prices the value of a retention program, and — via LTV:CAC — is the number most often quoted and least often defined in a seed deck.

What it is not

It is not revenue times expected life. It is not an observed quantity — it is a forecast, and its error bars are wide. And it is not a single defensible number: the same cohort honestly produces LTVs from $24,941 to $80,000 depending on which of five standard methods you pick.

Key takeaways

  • Use gross profit, never revenue. At a 75% gross margin, a revenue-based LTV overstates value by exactly 33%.

  • ARPA × margin ÷ churn is an undiscounted perpetuity. It assumes the customer lives forever at a constant hazard and that a dollar in year eight is worth a dollar today. At 85% annual retention and a 12% discount rate it produces roughly 1.8x the discounted answer.

  • Retention is the dominant input, by a wide margin. Retention elasticity of customer value is 3–7; discount-rate elasticity is about a fifth of that.

  • Quote a range. An LTV estimated from 100 customers with one year of history carries a 95% confidence interval spanning roughly $26,000 to $45,000 on a $33,000 point estimate. The decimal places are decoration.

On this page11 sections

What is customer lifetime value?#

LTV = Σ  [ P(active in period t) × gross profit in period t ] / (1 + d)^t
      t=1..T

Three things are doing the work: a survival curve, a margin, and a discount rate. Everything else is bookkeeping.

Gupta, Lehmann and Stuart define the value of a customer as "the expected sum of discounted future earnings," and their infinite-horizon closed form is the version most worth knowing:

LTV = m × r / (1 + i − r)

where m is periodic margin, r is the retention rate, and i is the discount rate. Note that this convention survival-weights the first period too, which is conservative; the common founder convention treats the newly acquired customer as certain in period 1 and applies survival from period 2. Both are defensible. State which one you used — the two differ by a factor of r, which at 85% retention is a 15% swing.

The margin has to be gross profit#

This is not a detail; it is the definition.

BasisWhat it measuresVerdict
Revenue
Money in
Wrong. Ignores the cost of delivering the service you sold
Gross profit (revenue − cost of revenue)
Money left after delivery
Standard. Use this and label it gross-profit LTV
Contribution (also deducting customer-specific success, support, payment, and renewal cost)
Money left after keeping the customer
Most complete. Use for operating decisions; label the boundary

A revenue-based LTV is not "conservative in a different way." It is arithmetically the gross-profit LTV divided by the gross margin — at 75% margin, exactly 1.33x too big, before any other error. Because CAC is a real cash cost, comparing a revenue-based LTV to CAC compares a pre-cost number to a post-cost number, which is the single most common error in the metric.

Simple LTV versus discounted LTV#

The shortcut every founder learns first:

simple LTV = ARPA × gross margin / churn rate

It is a perpetuity with three buried assumptions: a constant hazard (a customer in month 60 is as likely to leave as one in month 2), an infinite horizon, and a zero discount rate. None of those hold. It is a fine back-of-envelope sanity check and a bad basis for a spending decision.

Gupta, Lehmann and Stuart make the related point precisely: converting a retention rate into an "expected lifetime" and then valuing that many years of margin overestimates lifetime value. Their example — $100 annual margin, 80% retention, 12% discount — gives a true LTV of $250, against $360 for the naive five-year annuity, an overstatement of about 44%.

Why does LTV matter to founders?#

It sets a ceiling on acquisition, not a target. LTV tells you the most a customer could ever be worth. What you can afford to spend is that number minus a margin of safety for forecast error, minus the fixed costs LTV does not fund, minus the financing cost of the gap before payback.

It prices retention work. A customer success program is worth building when its incremental cost is below the discounted contribution it preserves. Without an LTV model that is an argument about vibes.

It reorders your segments. Expensive customers who retain and expand routinely beat cheap customers who churn before onboarding cost is recovered. The CAC page's worked example shows exactly this inversion.

It makes a valuation narrative inspectable. Customer-based valuation links active customers, acquisition rate, retention, and margin to enterprise value. It does not remove uncertainty — it moves the uncertainty into named assumptions someone can argue with.

How do you calculate LTV?#

1. Define the customer and freeze the cohort#

Use the contracting unit — account, location, household, merchant, workspace — and group by acquisition period. The same unit must be used in CAC or the ratio is meaningless.

2. Observe survival by tenure; do not assume it#

Pull the actual cohort curve. Real churn is front-loaded: onboarding failure, then a renewal cliff, then relative stability. A constant-hazard model understates early loss and overstates late loss, and gets the shape wrong even when it gets the average right.

3. Build gross profit per period, not revenue#

Deduct cost of revenue on the same basis your accountant uses. Then decide, explicitly, whether recurring customer-specific success, support, payment, and renewal-commission costs come out too — and label the result accordingly.

4. Discount#

Use a rate consistent with the cash-flow frequency and the risk. For an early-stage company, 10–15% annually is a common range; the Gupta paper uses 12%. Do not spend an afternoon choosing between 11% and 13% — retention elasticity is roughly five times discount-rate elasticity, so that afternoon is better spent on the survival curve.

5. Cap the horizon#

A perpetuity in a business with two years of history is a claim you cannot support. Use an explicit 3–5 year window plus, if you must, a conservative residual — and show the horizon on the face of the number.

6. Back-test against matured cohorts#

Compare predicted cumulative gross profit with realised cumulative gross profit by cohort age. Attribute the error to retention, price, or delivery cost. An LTV model that has never been back-tested is a spreadsheet, not a forecast.

Key Facts

01

Retention dominates every other lever

Across five firms, Gupta, Lehmann and Stuart estimate a retention elasticity of customer value of 3 to 7 — a 1% improvement in retention raises customer and firm value by 3–7% — against a margin elasticity of about 1 and an acquisition-cost elasticity of 0.02 to 0.32.

Gupta, Lehmann & Stuart, *Valuing Customers*, JMR 2004
02

The discount rate matters about a fifth as much as retention

In the same study, a 1% improvement in retention raised customer value by 2.45–6.75%, while a 1% reduction in the discount rate raised it by only 0.5–1.2%.

Gupta, Lehmann & Stuart, *Valuing Customers*
03

The "expected lifetime × margin" shortcut overstates by double digits

With $100 annual margin, 80% retention and a 12% discount rate, the discounted value is $250; valuing five years of $100 gives $360.48, an overstatement of 44%.

Gupta, Lehmann & Stuart, *Valuing Customers*
04

The margin you multiply by is roughly three-quarters, not one

Confluent's fiscal 2025 subscription gross margin was 78.1% ($1,119.7M subscription revenue against $245.4M subscription cost of revenue). Using revenue instead would overstate every LTV in that business by 28%.

Confluent FY2025 Form 10-K
05

Even best-in-class retention leaves real annual loss

Workiva reported a 97.2% gross retention rate as of December 31, 2025 — about 2.8% annual gross revenue churn. At 2.8%, a perpetuity model implies a 36-year customer life, which is roughly twice the age of the company.

Workiva FY2025 Form 10-K

Worked example: five honest LTVs for one cohort#

Take the cohort from the CAC page: 30 accounts, blended ARPA of $1,000 MRR ($12,000 ARR), 75% gross margin, 85% annual logo retention, fully loaded CAC of $12,000, discount rate 12%.

annual revenue      = $12,000
annual gross profit = $12,000 × 75% = $9,000
annual churn        = 15%   (monthly equivalent: 1 − 0.85^(1/12) = 1.345%)

Method 1 — revenue-based simple LTV (wrong, and common):

$12,000 / 0.15 = $80,000

Method 2 — gross-margin simple LTV, annual basis:

$9,000 / 0.15 = $60,000

Method 3 — gross-margin simple LTV, monthly basis:

$750 / 0.013452 = $55,754

Methods 2 and 3 use the same inputs and differ by $4,246, purely because the annual version credits a full year of gross profit to customers who churned in month two. The monthly version is the more careful of the two.

Method 4 — discounted perpetuity at 12%:

$9,000 / (1 + 0.12 − 0.85) = $9,000 / 0.27 = $33,333

(Under the Gupta convention, which also survival-weights year 1: $9,000 × 0.85 / 0.27 = $28,333.)

Method 5 — explicit five-year discounted cohort:

YearSurvivalGross profitDiscount factorPresent valueCumulative
1
100.0%
$9,000
0.8929
$8,035.71
$8,035.71
2
85.0%
$7,650
0.7972
$6,098.53
$14,134.25
3
72.3%
$6,503
0.7118
$4,628.35
$18,762.60
4
61.4%
$5,527
0.6355
$3,512.59
$22,275.19
5
52.2%
$4,699
0.5674
$2,665.80
$24,940.99

The full ladder, one cohort:

MethodLTVvs. CAC of $12,000
Revenue-based simple
$80,000
6.67x
Gross-margin simple, annual
$60,000
5.00x
Gross-margin simple, monthly
$55,754
4.65x
Discounted perpetuity, 12%
$33,333
2.78x
Discounted perpetuity, Gupta convention
$28,333
2.36x
Explicit 5-year discounted
$24,941
2.08x

Same customers, same retention, same margin, no arithmetic error — and a 3.2x spread in the headline number. This is why "our LTV:CAC is 3:1" carries almost no information until the method is stated. See LTV:CAC Ratio.

How precise is an LTV number actually?#

Far less precise than the way it is usually written. The dominant source of error is not the discount rate — it is that retention itself is estimated from a short history on a small sample.

Suppose you observe 100 accounts for twelve months and 15 of them churn. Your point estimate is r = 0.85, but the standard error of that proportion is sqrt(0.85 × 0.15 / 100) = 0.0357, giving a 95% interval of roughly 78.0% to 92.0%. Push that through the discounted perpetuity:

Cohort size observed95% CI on annual retentionImplied LTV rangeImplied LTV:CAC range
50
75.1% – 94.9%
$24,392 – $52,624
2.03x – 4.39x
100
78.0% – 92.0%
$26,472 – $44,997
2.21x – 3.75x
400
81.5% – 88.5%
$29,509 – $38,297
2.46x – 3.19x

Three things follow. First, sampling noise alone spans the 3:1 benchmark at every realistic early-stage cohort size — a 100-account cohort cannot distinguish 2.2x from 3.7x. Second, this table assumes the constant-hazard model is correct; it prices sampling error only, not model error, and real survival curves are not constant-hazard. Third, twelve months of history tells you nothing about year four, which is where a perpetuity gets most of its value.

The practical rule: report LTV as a range with the cohort size and observation window attached, and treat any LTV quoted to the nearest dollar from under two years of data as a presentational choice rather than a measurement.

What are the common mistakes?#

  • Using revenue instead of gross profit. Mechanically inflates LTV by 1 / gross margin. At 75% margin that is 33%; at 45% it is 122%.
  • Reporting the simple formula as if it were discounted. ARPA × margin ÷ churn is an undiscounted infinite-horizon perpetuity. Say so, or discount it.
  • Assuming a constant hazard. Real churn is front-loaded. Use the observed cohort curve where you have one.
  • Booking expansion without its cost. Upsell that requires sales, success, or implementation effort is not free contribution.
  • Averaging across a concentrated base. When five accounts produce most of the gross profit, a mean LTV describes no customer in the business. Report the distribution.

When does LTV break?#

Young cohorts. With eighteen months of history you are extrapolating year four from nothing. Cap the horizon and widen the range rather than pretending otherwise.

Non-contractual purchasing. For transactional and marketplace businesses, inactivity is not churn. Use repeat-purchase and probabilistic-inactivity models; a subscription retention rate applied to episodic buying produces a confident, meaningless number.

Concentration. If the top 5% of accounts drive most of the gross profit, the mean is an artefact. Value the large accounts individually and model the tail separately.

Active repricing or repackaging. LTV assumes the offer that produced the historical curve still exists. During a pricing or packaging change, the survival curve you measured belongs to a product you no longer sell.

Cash. LTV can be comfortably positive while the business runs out of money — the value arrives over five years and the CAC leaves the bank this quarter. LTV must always be read alongside CAC payback and runway.

Frequently asked questions

01

Should LTV use gross margin or contribution margin?

Gross-profit LTV is the standard for comparison over time and against peers, because cost of revenue is a defined accounting boundary. Contribution LTV — which also deducts customer-specific success, support, payment, and renewal costs — is more economically complete and is the right input to an operating decision. Compute both, label both, and never compare one company's contribution LTV to another's gross-profit LTV.

02

Is the simple `ARPA × margin ÷ churn` formula useless?

No — it is a good sanity check and a fast way to see how sensitive the business is to churn. It is unsuitable as the basis for a spending decision, because it capitalises an infinite customer life at a zero discount rate. In the example above it returns 1.80x the discounted perpetuity ($60,000 versus $33,333) and 2.41x the explicit five-year value ($60,000 versus $24,941).

03

What discount rate should an early-stage company use?

Something in the 10–15% range, stated openly. It matters much less than you think: retention elasticity is roughly five times discount-rate elasticity in the Gupta study. Spend the effort on the survival curve instead.

04

How do I handle expansion revenue and negative churn?

Model expansion as a growth term on per-period gross profit for surviving customers, and deduct the cost of producing it. Do not fold expansion into the retention rate — a net revenue retention above 100% used as r in a perpetuity formula produces a negative or infinite denominator, which is where the genuinely absurd LTV numbers in pitch decks come from.

05

How far out should I project?

No further than roughly twice your longest observed cohort. If you have 24 months of retention data, a five-year explicit horizon is already an extrapolation, and a perpetuity is a claim about a decade you have not seen.

Sources#

  1. Gupta, S., Lehmann, D. R., & Stuart, J. A., Valuing Customers, Journal of Marketing Research 41(1), 7–18, February 2004. The m × r / (1 + i − r) formulation, retention/margin/acquisition elasticities, the retention-versus-discount-rate comparison, and the $250-versus-$360 illustration of the expected-lifetime error.
  2. Confluent, Inc., FY2025 Form 10-K, for the year ended December 31, 2025. Subscription revenue and subscription cost of revenue used for the 78.1% gross margin figure.
  3. Workiva Inc., FY2025 Form 10-K, for the year ended December 31, 2025. Gross retention rate of 97.2% as a best-in-class enterprise SaaS reference point.
  4. Similarweb Ltd., First Quarter 2026 Results (Form 6-K, Exhibit 99.1), May 13, 2026. Illustrates that customer-economics metrics are published as company-defined supplemental measures, not standardised ones.
  5. Skok, D., SaaS Metrics 2.0 — A Guide to Measuring and Improving What Matters, For Entrepreneurs. Origin of the simple SaaS LTV formulation and the author's own subsequent replacement of it with a discounted-cash-flow approach.

Source-use note: The 78.1% subscription gross margin and the 36-year implied life at 2.8% churn were calculated by us from the figures those companies reported. The worked example and the sampling-error table are illustrative constructions, not company data.

Note: This page is educational and does not constitute accounting, financial, or investment advice. LTV is an unaudited forecast built on assumptions the preparer chooses, it falls outside GAAP and IFRS, and small changes in retention produce large changes in the result. Consult a qualified accountant or financial adviser before relying on it in investor reporting, a valuation, or a financing document.

Author

Dr. Sarah Zou

Independent economist · EconNova

Commercial strategy for technical products, with a focus on pricing, unit economics, and the operating choices behind the model.

About Sarah

Topics

customer lifetime valueCLVLTVretentiondiscountingunit economics

Cite this page

Suggested citation

Zou, S. (2026). Customer Lifetime Value: From Retention Assumptions to Cash Economics. In Unit Economics. Pricing & Monetization Wiki. https://sarahzou.com/wiki/unit-economics/customer-lifetime-value

Open license

Reuse with attribution

This content is available for reuse. When referencing or republishing it, please credit Dr. Sarah Zou and link back to the original source.

Licensed under Creative Commons Attribution 4.0 International. You may share and adapt the material with appropriate credit.