How to Calculate LTV for SaaS

LTV is three different metrics wearing the same name, and nobody tells you which one they used. That’s why two people can look at the same company and quote numbers that differ by a factor of three, both sincerely.

This page runs one real-shaped 24 month cohort through all three methods so you can see exactly where each one bends.

The naive formula and its two failure modes

ARPA divided by churn rate. At $500 monthly ARPA and 2 percent monthly logo churn, that’s $25,000, times gross margin.

It’s popular because it fits in a tweet. It fails in two specific ways, and both get worse as your company gets better.

Failure one: it breaks at low churn. The formula treats churn as a constant, so lifetime is 1 divided by churn. At 2 percent monthly, that’s 50 months. At 1 percent, 100 months. At 0.5 percent, 200 months, which is sixteen and a half years. You are now planning against revenue from 2042. A twenty basis point measurement error at the low end swings LTV by tens of thousands of dollars.

Failure two: it ignores expansion entirely. If your customers grow, ARPA at month 24 is not ARPA at month 1. Any product with seat-based or usage-based pricing violates the formula’s core assumption from day one. Snowflake and Datadog both built businesses where the month-36 customer is worth multiples of the month-1 customer. The naive formula cannot represent that at all.

Cohort-based LTV: what a real retention curve looks like

Real retention curves are not exponential decay at a constant rate. They drop steeply in the first three to six months and then flatten, sometimes almost completely. The customers who survive early are structurally different from the ones who did not.

Cohort LTV works from the observed curve. Take one signup cohort, sum the revenue it produced each month, and extrapolate the remaining months using the shape the curve has settled into rather than a single average rate.

Here’s the worked cohort. One hundred customers signed up in month zero at $500 a month, gross margin 78 percent.

Notice what’s happening. Logo retention flattens near 59 percent, but revenue per surviving customer climbs from $500 to $670. Cohort MRR at month 24 is nearly 80 percent of month one despite losing 41 percent of the customers.

Extrapolating the flattened curve to a 36 month cap gives roughly $10,400 gross-margin LTV per original customer. Extend the cap to 60 months and it reaches about $16,200.

The same cohort under all three methods

Now run that identical data through each formula.

Method	Calculation	Result	Where it bends
Naive ARPA over churn	$500 ARPA, 2.1% monthly churn, x 0.78 margin	$18,571	Assumes 48 month flat curve and zero expansion
Cohort-based, 36 month cap	Observed curve, extrapolated flat tail, x 0.78	$10,400	Conservative, defensible, needs 18+ months of data
Cohort-based, 60 month cap	Same curve, longer horizon	$16,200	Reasonable only if you have a 5 year old cohort
NRR-adjusted	108% NRR compounded, 36 month cap, x 0.78	$13,900	Overstates if NRR is carried by a few large accounts

The naive formula lands 79 percent above the conservative cohort figure. That gap is the difference between a growth plan that works and one that runs out of cash in month fourteen.

Worth noting the naive result here is not wildly wrong because churn is relatively high at 2.1 percent. Repeat the exercise at 0.7 percent monthly churn and the naive method returns roughly $55,700 against a cohort figure near $19,000. The lower your churn, the more dangerous the shortcut becomes.

NRR-adjusted LTV for expansion-heavy products

If your net revenue retention sits above 100 percent, the naive formula doesn’t just mislead, it fails arithmetically. Net churn is negative, and dividing by a negative number produces an infinite or nonsensical lifetime.

The NRR approach projects cohort revenue compounding at the observed net retention rate, then caps it. At 108 percent annual NRR, a cohort starting at $50,000 MRR is worth $54,000 in year two and $58,320 in year three, before margin.

Two cautions. First, NRR is often carried by a small number of large accounts, so the cohort average flatters the median customer badly. Check whether your expansion is broad or concentrated before you model it as a cohort property. Second, NRR is mean-reverting in practice. A company running 130 percent NRR during a usage boom should not extrapolate that across five years.

For a fuller treatment of how retention metrics differ, NRR vs GRR covers why gross retention is often the more honest planning input.

The gross margin adjustment, and why 2026 makes it mandatory

LTV without a margin adjustment is a revenue number pretending to be a profit number. Since you compare it against CAC, which is spent in real dollars, the comparison is invalid without it.

For years this felt like a rounding detail because classic SaaS ran at 75 to 85 percent gross margin. That assumption is breaking. AI-native products carrying inference costs commonly run 50 to 70 percent, and some usage-heavy products run lower during their growth phase when they’re subsidising consumption.

Which costs belong in COGS for this calculation: hosting and infrastructure, model inference, customer support, customer success salaries where CS is a delivery function, third party data or API fees passed through to the product. Not sales, not marketing, not R&D.

The lifetime cap, and picking the right horizon

Every honest LTV has a cap. The question is where to put it.

• 36 months is the default for most B2B SaaS. It’s long enough to capture expansion and short enough that you’re not planning against speculation.
• 60 months is defensible for enterprise products with multi-year contracts, but only if you have a cohort that old to validate the curve shape.
• 24 months is right for early stage companies and any product in a fast-moving category, particularly AI tooling where the competitive set may not exist in three years.

Never use an uncapped horizon. The compounding of a flat retention tail across an infinite series produces numbers that are mathematically valid and commercially fictional.

Which method to use, and where I’d draw the line

Use cohort-based LTV with a 36 month cap and a gross margin adjustment for anything that informs a decision. Planning, hiring, CAC targets, board reporting. It requires more work and at least eighteen months of cohort data, and it’s the only version that survives scrutiny.

Consider using NRR-adjusted LTV as a secondary view if you’re expansion-heavy, and present it alongside the cohort number rather than instead of it.

Use ARPA over churn for exactly one thing: a fast directional check on a company you know nothing about. Never in a board deck. The formula assumes a retention curve that no SaaS company has ever observed, and presenting it to people making capital allocation decisions is a quiet form of misreporting.

Who should reasonably disagree: a pre-product-market-fit company with six months of data has no cohort to work from. Use the naive formula, label it clearly as a placeholder, and state the implied lifetime in months so everyone can see the assumption.

Instrumenting this so you can rerun it quarterly

What to do next

Pull one cohort that’s at least eighteen months old and run all three methods on it. The spread you get is the most useful number on this page, because it tells you how much your current reporting is off by.

The Customer lifetime value calculator will do the cohort arithmetic, and the LTV to CAC Ratio Calculator pairs it with the acquisition side. Make sure the CAC input is fully loaded before you compare them: How to Calculate CAC for SaaS covers which costs belong in the numerator, and the LTV to CAC ratio calculator gives you the ratio with both sides stated.

Since retention drives everything above, How to Calculate SaaS Churn Rate is the natural companion, and Average Contract Value (ACV) clarifies the terminology people mix up when quoting ARPA. The SaaS Metrics and Analytics hub covers how these numbers connect to payback, which is where B2B SaaS CAC Payback Calculator picks up.