How to use LTV Measurement
A practical methodology guide for measuring LTV on Shopify, WooCommerce, and Magento stores — covering historical vs predictive models, cohort curves, and the margin choices that decide whether your number is decision-useful.
LTV Measurement
The methodology layer that decides how customer lifetime value is calculated — historical vs predictive, cohort vs average, revenue vs contribution margin.
LTV Measurement is the set of methodology choices that sit underneath your customer lifetime value number: whether you compute it from realised order history or a predictive model, whether you slice it by acquisition cohort or take a sitewide average, and whether the value you sum is gross revenue, gross profit, or contribution margin after variable costs.
The choices matter because the same store can publish four LTV figures that differ by 3-5x depending on methodology. A decision-useful LTV is the one that maps cleanly to a CAC payback window, a paid-media bidding ceiling, or a retention investment case — not the biggest number you can defend.
Most stores in the €1M-€15M revenue band measure LTV once, in a spreadsheet, using whatever method the previous analyst picked. The number then drives bid caps, retention budgets, and board slides for two years without anyone re-examining whether the methodology fits the decision.
This guide walks through the four methodology decisions that actually move the number — historical vs predictive horizon, cohort vs sitewide aggregation, revenue vs contribution margin, and reporting cadence. Each section ends with the question you should be able to answer before publishing a single LTV figure internally.
Historical LTV vs predictive LTV
Historical LTV sums the actual order value a cohort has generated to date. It is auditable, defensible, and always too small — by definition it excludes the future purchases the cohort hasn't made yet. For a 6-month-old cohort, historical LTV captures maybe 40-60% of what they'll eventually spend.
Predictive LTV extrapolates forward using a model — usually BG/NBD plus Gamma-Gamma for the spend distribution, or a simpler retention-curve fit for stores without a data team. It answers the question paid media actually asks: what is this customer worth over the next 24 months, so what can I afford to pay to acquire them today?
The right default for an apparel or beauty store is to publish both. Use historical LTV for finance and board reporting where you need an auditable number, and use predictive LTV for paid-media bid caps and retention ROI cases. The detailed comparison lives in Predictive LTV vs Historical LTV — including when each model breaks.
The 12-month trap
Many Shopify stores publish 'LTV' as the average revenue per customer in their first 12 months. That is neither historical nor predictive — it is a censored historical number that systematically under-counts repeat buyers and over-counts one-and-done cohorts that happen to have a full year of data. If your LTV moves every time you refresh the dashboard, this is why.
Cohort curves vs sitewide averages
A sitewide average LTV is one number for the whole store. A cohort LTV curve is a separate value for every acquisition month, tracked at 30, 60, 90, 180, and 365 days post-first-order. The curves are the only honest way to see whether retention is improving or degrading.
Sitewide averages hide cohort mix shifts. If you spent heavily on a discounted Black Friday campaign, the resulting cohort drags the sitewide LTV down for the next 18 months — and you can't tell whether it's because the campaign was bad or because the cohort is simply newer. Cohort LTV Curves separate the two by holding tenure constant.
Cumulative LTV by acquisition cohort (apparel store, contribution-margin €)
2023 Q1 cohort (full-price)
2023 Q4 cohort (BFCM-acquired)
2024 Q2 cohort (subscription push)
Three cohorts from the same store, three completely different LTV trajectories. The sitewide average across all three at M12 would be around €83 — a number that describes none of them and would mislead any bid-cap decision based on it.
Revenue LTV vs contribution-margin LTV
Revenue LTV sums what the customer paid you. Contribution-margin LTV sums what's left after COGS, payment processing, fulfilment, and returns. The gap between them is usually 55-75% of the revenue figure — meaning a €200 revenue LTV is closer to a €60-€90 contribution-margin LTV.
Paid-media bid caps should never reference revenue LTV. If your CAC is €40 and revenue LTV is €200, the 5:1 ratio looks healthy — but if contribution-margin LTV is €70, you're actually at 1.75:1 and burning cash on every acquisition. The choice of margin matters more than the choice of model.
LTV methodology choices by store maturity (typical setup)
| Stage | Aggregation | Horizon | Margin basis | Refresh cadence |
|---|---|---|---|---|
| €1M-€3M revenue, no data hire | Sitewide average | Historical, 12-month | Gross revenue | Quarterly |
| €3M-€7M, first analyst | Quarterly cohorts | Historical 24-month + simple curve fit | Gross profit (after COGS) | Monthly |
| €7M-€15M, data team forming | Monthly cohorts by channel | Predictive 36-month (BG/NBD) | Contribution margin | Weekly |
| €15M+, full data function | Cohort × channel × product | Predictive 36m + scenario bands | Contribution margin + returns reserve | Continuous |
Most stores skip the middle row and jump from 'sitewide average revenue LTV' to 'we should build a predictive model'. The middle row is where the real measurement work lives — quarterly cohorts on gross profit, refreshed monthly, will resolve 80% of the decisions a predictive model would. For the underlying math, see LTV Calculator and the published LTV Benchmarks by Industry.
Reporting cadence and decision fit
LTV is not a daily metric. The signal-to-noise ratio at weekly cadence is poor for stores under €10M — you're looking at small cohorts with high variance, and you'll chase noise. Monthly cohort refresh with a 90-day rolling LTV curve is the cadence that matches how acquisition decisions actually get made.
The decision fit test: name the decision your LTV number is going to drive this quarter. If it's a paid-media bid cap, you need predictive contribution-margin LTV by acquisition channel. If it's a board update, historical revenue LTV by quarterly cohort is fine. If you can't name the decision, you don't need a new number — you need a clearer question.
The contribution-margin shortcut
If you only do one thing differently this quarter, switch the LTV number on your dashboard from revenue to contribution margin. It will drop by 60-70%, every CAC:LTV ratio in the business will look worse, and the conversations that follow will be the right ones. Most stores discover their 'profitable' acquisition channels are not.
Frequently asked questions
Use a simple retention-curve fit rather than a predictive model — fit an exponential decay to your 30/60/90-day repeat-purchase rates and project forward 12 months. Avoid BG/NBD models until you have at least 18 months of data; the parameter estimates are too unstable below that. The historical 6-month number is also useful as a floor.
Nothing material — Customer Lifetime Value (CLV) and LTV are used interchangeably in e-commerce. Some textbooks reserve CLV for the discounted-cash-flow version that applies a discount rate to future cash flows, but in practice the two terms describe the same metric. Pick one and use it consistently across reporting.
Contribution-margin LTV for any decision involving spend — paid media bid caps, CAC ceilings, retention ROI. Revenue LTV is fine for top-line board reporting where the audience understands the difference. Never compare CAC (a cost) to revenue LTV (pre-cost) — the resulting ratio is structurally misleading.
A sitewide LTV averages across customers of every tenure, which means it mixes 30-day-old customers with 3-year-old ones. A cohort LTV holds acquisition month constant, so you compare like-for-like — the M12 LTV of the January cohort vs the M12 LTV of the June cohort. This is the only way to see whether retention is genuinely improving.
Usually no. The marginal accuracy from a BG/NBD model over a well-built cohort retention curve is small at that scale, and the maintenance overhead is real. Spend the analyst-week on cohort hygiene and channel-level historical LTV instead. Revisit predictive models when monthly cohort sizes exceed 500 customers.
Returns should be netted out of both revenue and contribution margin before you compute LTV — gross order value massively over-states LTV in categories with 20-40% return rates like apparel and footwear. Pull net revenue (after refunds) from your Shopify or commerce platform export, not gross order value.
3:1 on contribution-margin LTV is the common benchmark, with payback inside 12 months. Below 2:1 you're not generating enough margin per customer to fund growth; above 5:1 you're likely under-investing in acquisition. The ratio is more meaningful by channel than at the sitewide level.
Monthly cohort refresh is the right cadence for most €1M-€15M stores. Weekly creates noise without signal at that cohort size; quarterly is too slow to catch retention deterioration. The exception is around major campaigns (Black Friday, big launches) where a 30-day post-campaign cohort check is worth running.
Yes — a quarterly-cohort historical LTV by acquisition channel can be built in a spreadsheet from a Shopify customer export in a few hours. The blocker isn't tooling; it's discipline around cohort definitions and consistent margin assumptions. Tools like the LTV Calculator handle the math once you've decided the methodology.
Subscriptions compress the variance — repeat behaviour is contractual rather than probabilistic — so a simple churn-rate-based LTV (ARPU / monthly churn) works well. Measure subscription and one-time customers as separate cohorts; blending them produces a sitewide LTV that describes neither group accurately.
Track CAC, channels, and funnel conversion in one place
Metricuno connects ad spend, funnel events, and revenue so you can see CAC by channel, cohort, and campaign — without stitching together five tools.