Most lifetime value numbers are a forecast, and the forecast is generous

Customer lifetime value card showing the number one, the single window every calculation has to declare
Contents 9 sections

Customer lifetime value is the gross profit one buyer produces before they stop buying from you. Almost every figure I get sent is something else. It is last year’s revenue over last year’s buyer count. Then it is multiplied by a repeat rate somebody hoped for. That difference decides what you are allowed to pay for a first order, so it is not a bookkeeping detail. Pay too much for too long and the loss appears two quarters later, when the cash has gone.

This is for ecommerce operators who want to bid beyond the first purchase. The number has to survive contact with your own accounts.

What customer lifetime value actually measures

It measures profit per buyer across a stated window, not revenue and not forever.

Two decisions do most of the work. The first is the window: six months, twelve months, twenty-four months. Second comes the choice between revenue and gross profit. Change either and the number moves by a multiple. Two people inside one business will quote different figures. Both will believe they are right.

The word lifetime is the problem. Nobody waits for a lifetime, so everybody picks a window and then drops the window from the sentence. A twenty-four month figure quoted to a founder who assumed twelve will justify roughly double the acquisition cost it should.

So write the window into the number every time. Twelve month gross profit per buyer is a sentence somebody can check. LTV on its own is not.

The three inputs, and the two that get guessed

You need average order value, a repeat purchase rate, and gross margin. Only the first of those is usually reliable.

Average order value comes straight out of the store. There is also a public reference point to sanity check it against. Shopify’s BFCM report of 2 December 2025 put average order value at $114.70. That is first-party data across its full merchant base. Peak weekend behaviour runs differently from the rest of the year, so treat it as a reference, not a target.

The repeat purchase rate is where guessing starts. Most stores can produce it, few have looked at it, and almost nobody splits it by acquisition channel.

Gross margin is the input clients withhold. Sometimes deliberately, more often because nobody has ever asked finance for a landed cost including shipping and payment fees.

Why customer lifetime value has to run on gross margin

Revenue LTV will talk you into spending more than the business earns.

Across my ecommerce clients gross margin runs between 20% and 50%. Take three orders at that Shopify weekend figure of $114.70 and you have $344.10 of revenue. At 50% margin that is $172.05 of gross profit. At 20% it is $68.82, and the second business cannot pay anything close to what the first can for the same buyer.

Same revenue, same order count, and an acquisition ceiling that differs by a factor of two and a half. Anybody quoting an LTV without a margin attached has told you very little.

One more deduction gets skipped. Shipping, payment processing and pick and pack are variable costs. They belong inside the margin figure, not below it in a monthly overhead line.

Orders are not customers, and my own reporting shows it

Every platform figure I have ever reported is a count of purchases, not a count of people.

Two of my own numbers make the point. On an EU fashion ecommerce account I recorded 1,715 orders at €14.18 each, across the full year of 2022. That was Meta and Google combined, with most sales coming from Google Search and Shopping. On a US electric mobility account it was more than 63,000 purchases at $5.05, also full year 2022, across Google Search, Performance Max and Display.

Both are order-level figures. Neither tells you how many distinct people sat behind them. Ad platforms count conversions, and a buyer who orders three times is three conversions.

Build an LTV from platform data and you divide by a denominator that is too high. A customer lifetime value resting on that denominator is wrong before you start. It flatters your acquisition cost and understates repeat behaviour in the same move.

Returns come off the top before anything else

A returned order is revenue you booked, refunded, and paid to acquire.

Salesforce’s 2025 holiday shopping data, published on 8 January 2026, put holiday returns at $181 billion. That is 14% of purchases, up 10% year on year. The sample covers 1.5 billion shoppers across 89 countries, and it skews towards enterprise retail. Read it as an order of magnitude, not as your rate.

Now the part that matters for the calculation. Returns are handled by my clients’ own in-house teams. They never reach my reporting, and they never flow back into the ad platform. No version of the ROAS I optimise to has ever had them deducted.

So the media buyer’s revenue and the finance team’s revenue are two different numbers, separated by roughly the return rate. Nobody reconciles them unless somebody insists. Returns belong inside customer lifetime value, deducted before anything else is counted.

Worth naming the gap in the evidence too. No primary source exists for return rate by category, apparel included, and the Salesforce figure is an overall rate rather than a category one.

What customer lifetime value does to your bidding

It raises the cost per acquisition you can defend, and nothing else changes automatically.

Say twelve month gross profit per buyer is $90 and your break-even on the first order is $30. You can defend a higher first-order cost, provided the business can fund the gap. That proviso is where most of these plans die. The arithmetic works and the bank account does not.

Predicted LTV models inside the ad platforms are built from your historic order data, so they inherit every flaw above, including the denominator problem.

If you switch to value-based bidding, give it room. Google’s own guidance is that target ROAS needs 15 conversions in the past 30 days on Search and Shopping. It also asks you to allow four weeks or three conversion cycles after a reporting change. Feeding a new value into a bidding strategy counts as a reporting change.

The channel that quietly inflates the number

Repeat orders often arrive through email, and email costs almost nothing at the margin.

I run Meta, Google Ads and Klaviyo for a US flag and banner ecommerce client. Klaviyo is email, not an ad platform, which means the second and third orders from a buyer can land at close to zero marginal cost. Credit those orders to the paid channel that found the person and paid LTV looks excellent. Credit them to email and the paid channel looks ordinary.

Both readings are defensible, and they justify very different budgets. An honest version states the split.

That same client has never disclosed gross margin. I can give you their cost per purchase to the cent. Their profit per buyer I cannot give you at all, because one of the three inputs was never shared.

When customer lifetime value is safe to bid on

Three conditions, and failing any one of them means bidding to first-order profit instead.

You need twelve months of customer-level order history, taken from the store and not the ad platform. You need a stated gross margin that already includes shipping and payment fees. Third, you need a returns figure, even a rough one from the finance side.

Miss one and the number turns into an argument for spending more, which is exactly how it will get used.

New businesses fail the first condition by definition. A store trading for four months has no repeat data, only hope. Hope priced into a bid is how a young brand runs out of cash while the dashboard looks healthy.

Single-purchase categories fail it differently. Mattresses, wedding items, some home improvement work: a second order may never come, so the first one has to pay for itself.

What to change this week

Three things, in this order.

Pull twelve months of order history from the store rather than the ad platform, and count distinct buyers instead of orders. That step alone usually shrinks the denominator and changes the repeat purchase rate you thought you had.

Then ask finance for one number: gross margin after shipping, payment fees and pick and pack. Not the markup, the margin.

Finally, rewrite your target as twelve month gross profit per buyer, with the window inside the sentence. Then check whether your current cost per acquisition still clears it. A customer lifetime value written down that way is one somebody else can audit.

Working out what a lead or an order is worth covers the valuation method underneath all of this. What a purchase actually costs at volume gives two of my own figures with their date ranges attached. The order size benchmark worth trusting covers the AOV input. Raising order size before cutting acquisition cost is the cheapest way to move this arithmetic. On Search most of the work sits in feeding the right value back into bidding, which my Google Ads work sets up deliberately. If your numbers refuse to reconcile, book a teardown.

Frequently asked questions

What is customer lifetime value?

It is the gross profit a single customer produces across a stated window, usually twelve or twenty-four months. Revenue versions of the number are common and misleading, because they ignore the cost of the goods. The window matters as much as the arithmetic. A twenty-four month figure will justify roughly double the acquisition cost that a twelve month figure justifies, so quote the window every time you quote the number.

How do I calculate customer lifetime value for an ecommerce store?

Take average order value, multiply by the orders a buyer places inside your chosen window, then multiply by gross margin after shipping and payment fees. Pull the order history from the store rather than the ad platform, and count distinct buyers, not purchases. Platform conversion counts treat one repeat buyer as several people, which inflates the denominator and distorts everything downstream of it.

Should customer lifetime value include returns and discounts?

Yes, both of them. A returned order is revenue you booked, refunded and paid to acquire, so leaving it in overstates the figure by roughly your return rate. Discounts come off the margin directly, which bites hardest in categories that trade on promotion. Returns are usually handled by an in-house team and never reach the ad reporting, so the media buyer and the finance team end up holding different revenue numbers.

How long should the window be for customer lifetime value?

Twelve months suits most ecommerce businesses, because it covers a full seasonal cycle and stays inside a horizon finance can plan against. Twenty-four months is defensible for subscription-like categories with proven repeat behaviour. Anything longer is a forecast rather than a measurement. Whichever you choose, hold it constant, because changing the window changes the number and makes every comparison with last quarter meaningless.

Is predicted LTV in the ad platforms worth using?

It can help larger accounts with genuine repeat behaviour and plenty of order history behind them. Predicted models are built from your own past data, so they inherit its faults, including counting repeat buyers as new ones. Treat the output as a ranking signal rather than a profit figure. Check it against store data before you let it set budgets, and give any new bidding strategy several weeks.

Why do platform purchase counts overstate my customer count?

Ad platforms count conversion events, not people. Someone who orders three times in a year is three purchases in Meta or Google, and often three attributed conversions spread across different campaigns. Divide revenue by that count and you get revenue per order dressed up as revenue per person. The store database is the only place with a reliable count of distinct buyers, so start there.

What if the client will not share gross margin?

Then you cannot produce a profit figure, and you should say so rather than publish a revenue proxy. Ask for a range instead of an exact number, because most finance teams will confirm a band even when they will not confirm a percentage. Across my ecommerce clients that range runs from 20% to 50%. Work the arithmetic at both ends and show what changes between them.

Can I bid to a longer horizon than the first order?

Only if the business can fund the gap between paying for a buyer now and collecting the profit later. The arithmetic and the bank balance are separate constraints, and the second one closes companies. Prove the repeat behaviour with store data first, then extend the horizon in steps rather than all at once, and watch cash rather than the dashboard while you do it.

Want a second pair of eyes on your account?

Book a free teardown

Book a free teardown

Send me your worst-performing campaign.

I will tell you what is wrong on the first call. No obligation, no hard sell, and I will say honestly if I am not the right fit.

Book a free 20-minute account teardown

Replies within one business day, Lahore time.