Lifetime value is the number that justifies a CAC. It is quoted in board meetings, used to defend an acquisition budget, and it is usually a forecast wearing the clothes of a measurement.
There is a real, useful number underneath it, and it is smaller and duller than the one most stores quote. This guide separates the two: what you can actually measure from your own order history, and what you are assuming the moment you project past it.
Measured lifetime revenue vs predicted LTV
Two different things travel under the same three letters.
- Historic lifetime revenue — what a customer has already spent with you, summed across their delivered orders. It is a fact. It is also backward-looking and systematically understates a cohort that has not finished buying yet.
- Predicted LTV — a model projecting what a customer will be worth over some horizon. It is more useful and it is an estimate, and it inherits every assumption in the model: retention curve, time horizon, discount rate, whether margin or revenue is being projected.
Neither is wrong. What is wrong is quoting the second as though it were the first — because by the time the number reaches a budget decision, the assumptions are invisible and the figure looks like something that was observed.
Start with revenue, then make it profit
Most LTV figures in circulation are revenue figures, which makes them unusable for the decision they are meant to support. You cannot compare a revenue LTV to an acquisition cost; one is a top line and the other is real money out.
Lifetime revenue all delivered orders, at the price paid
× Gross margin % after COGS
= Lifetime gross profit
− Lifetime fulfilment shipping, handling, returns processing
= Lifetime contribution
Lifetime contribution is the number that answers "what can I afford to pay to acquire one of these?". A customer with $420 of lifetime revenue at a 38% gross margin and $46 of fulfilment cost has contributed about $114 — which is a very different budget from $420.
The counting problem that comes first
Before any of this arithmetic means anything, the thing being counted has to be a person. This is where most customer analytics quietly break.
A shopper who checked out once as a guest, once with a phone number typed with a country code, and once without, is three customers in a naive count. Your repeat rate falls, your lifetime revenue per customer falls with it, and the business concludes it has a retention problem it does not have.
Resolving identity properly means preferring the platform's own customer id where there is one — the storefront has already decided which orders belong to one person — and falling back to a normalised phone and email rather than the raw string. The difference between a naive count and a resolved one is frequently large enough to change a strategy.
Repeat rate: two orders, not two line items
The other definition that quietly breaks is "repeat customer". A customer who buys three products in one checkout has placed one order. Counting that as repeat purchasing inflates retention across the whole business, and it inflates it most for stores that sell bundles — exactly the stores most likely to be making decisions about bundling.
A repeat customer has at least two distinct delivered orders. And the window matters as much as the rule: "repeated within August" and "has ever repeated" are different questions, and a retention figure that does not say which one it answers cannot be compared to anything.
Cohorts, because averages hide the trend
A single blended LTV across all customers mixes people who bought last week with people who have been buying for three years, and it will drift upward simply because your oldest customers have had more time.
Grouping customers by the month they first bought fixes that, and it answers the question that actually matters: is a customer acquired this year worth more or less than one acquired last year, at the same age? A rising blended LTV with a falling cohort curve is a business whose acquisition is getting worse and whose average is being propped up by its history.
What to do with the number
- Set an acquisition ceiling from contribution, not revenue. And decide explicitly how long you are willing to wait to recover it — a payback period you can fund is worth more than an LTV:CAC ratio you cannot.
- Segment before you average. The customers who arrived through a discount campaign and the ones who arrived organically rarely have the same curve, and merging them produces a number that describes neither.
- Watch concentration. If a fifth of revenue rests on a handful of buyers, that is a risk figure, not a success metric.
What ORVX reports, and what it does not
ORVX resolves a canonical customer per store and reports measured figures on that identity: lifetime revenue per customer, repeat rate on distinct delivered orders, revenue per customer, orders per customer, revenue concentration, and estimated CAC from ad spend against delivered customers acquired.
It does not publish a predicted LTV. That is a deliberate limit. A projection is only as good as its assumptions, and a number presented without them becomes a fact in the next meeting. What ORVX gives you is the measured base a projection would have to start from — and the honest statement that the projection is yours to make.
Payback period beats the LTV:CAC ratio
The ratio everyone quotes — LTV to CAC, with 3:1 held up as healthy — hides the question that actually decides whether you survive the quarter: when does the money come back.
A 3:1 ratio recovered over four years and a 3:1 ratio recovered in two months are the same ratio and completely different businesses. The first one needs funding for every customer it acquires; the second funds itself. If you are choosing one number to manage acquisition by, choose the payback period, because it is the one your bank balance responds to.
Calculate it from contribution, not revenue: how many months of a customer's contribution it takes to recover what you paid to acquire them. Then compare that to how long you can actually fund the gap.
Frequently asked questions
What is a good customer LTV for ecommerce?
There is no useful benchmark, because LTV is meaningless without the margin and the acquisition cost beside it. A $200 LTV at a 60% margin and a $40 CAC is a strong business; a $600 LTV at a 15% margin and a $250 CAC is not. Compare your own lifetime contribution to your own CAC rather than to a figure from another company's catalogue.
Should LTV use revenue or profit?
Profit — specifically contribution, after COGS and the variable costs of fulfilling the orders. A revenue LTV cannot be compared to an acquisition cost, because one is a top line and the other is money that has actually left the business. Most LTV figures in circulation are revenue figures, which is why they look so encouraging.
Why is my repeat rate lower than I expected?
Most often because the thing being counted is not a person. A guest checkout, a phone number typed with a country code and the same number typed without it can become three customers. Resolve identity on the platform's own customer id first and normalise phone and email before comparing. The second most common cause is counting line items instead of orders, which inflates it in the other direction.
Does ORVX forecast what a customer will be worth?
No, and that is deliberate. ORVX reports lifetime revenue that has already happened, alongside repeat rate, revenue per customer, orders per customer and concentration. A predicted LTV inherits every assumption in its model, and those assumptions become invisible the moment the number reaches a budget meeting. What ORVX gives you is the measured base a projection would have to start from.