New vs returning customers is the split that decides whether your prospecting budget is working, and most ad accounts cannot show it cleanly. A returning buyer is cheap to reach, already trusts you, and converts at a rate that flatters whichever campaign takes the credit. A first-time buyer costs the auction price of a stranger’s attention. Blend the two and the resulting figure looks healthier than the business underneath it. On more than one account the fix turned out to have nothing to do with targeting. It was an exclusion list and a different weekly report.
This page is for stores whose retargeting looks brilliant while the new customer count refuses to move.
New vs returning customers are two different purchases
One is a sale. The other is an introduction that happens to end in a sale.
A returning buyer already knows the brand, the sizing, the delivery time, and whether the last parcel arrived intact. None of that has to be bought again. The ad works as a reminder, and reminders are cheap.
A first-time buyer needs all of it built from nothing, usually across several visits and a good many impressions. That is what acquisition money actually buys.
So the two belong in separate campaigns, on separate budgets, judged against separate ceilings. Running them together starts as a bidding problem rather than a reporting one, because the algorithm will always chase the cheaper conversion, and the cheaper conversion is the repeat purchase that was arriving anyway.
Every ecommerce account I have worked on rewards that shortcut by default. Nothing in the platform stops it, and the reported numbers improve while it happens.
Orders are not customers, and my own figures show the gap
Every headline figure I publish counts orders, and I label it that way deliberately.
There is €14.18 across 1,715 orders from an EU fashion ecommerce client, full year 2022, on Meta and Google together with the majority of sales arriving through Google Search and Shopping. There is also $5.05 across more than 63,000 purchases from a US electric mobility client, again full year 2022.
Both are a cost per order. Neither is a customer acquisition cost, because the ad platforms could not tell me how many distinct people sat behind those orders, and guessing would be worse than saying so.
In a store with real repeat buying that gap is not a rounding error. Cost per order drops every time somebody comes back, while the cost of finding a stranger does not move at all.
Your commerce platform knows the answer. It can tell you how many first-time buyers you had last month. The ad account cannot, and it will cheerfully let you believe otherwise. That is the whole of the new vs returning customers problem in one line: the platform counts events, while the store counts people.
What a returning customer costs when email does the work
Close to nothing at the margin, which makes the comparison unfair from the start.
I scaled one client from roughly $10,000 to $50,000 a month, and not by raising the Meta budget. I added campaign objectives, added awareness campaigns, built custom audiences and lookalikes, and brought in channels the client was not using, including TikTok and Klaviyo.
Klaviyo is email rather than an ad platform. That distinction matters more here than anywhere else, because the second order it produces costs almost nothing to generate.
So a store with a working email programme should not be buying repeat orders in an auction. Paid retargeting there competes with a channel already reaching the same people at no marginal cost.
Spend auction money on strangers. Let owned media look after the people who have already handed over an email address.
The prospecting problem nobody has measured properly
A claim circulates widely that most prospecting campaigns are quietly reaching existing buyers.
Littledata is the usual source cited for it. Its write-up says the company audited dozens of Shopify stores, and that is the whole disclosure: no sample size, no date range, no methodology, and no figures. Its only quantified example is explicitly hypothetical.
I believe the underlying problem is real, because I have found it inside accounts myself. What I cannot do is point at evidence for it, because none has been published.
Holding those two things apart matters. The gap between a real problem and a documented one is where most marketing statistics live, and you can settle this one in your own account faster than you can argue about it.
Exclude your customer list from one prospecting campaign for a fortnight. If its cost per purchase jumps, the campaign was leaning on people you already had. New vs returning customers is precisely the sort of question two weeks of your own data answers better than any citation.
How I separate new vs returning customers in an account
Three moves, in this order, and none of them needs new software.
Exclude the purchaser list from every prospecting campaign. Upload it, keep it fresh, and accept that the excluded campaign will report a worse number afterwards. That worse number is the honest one.
Build the same customer list into its own retention audience, then decide deliberately whether it deserves paid budget at all. In stores with email running properly, it usually does not.
Report first-time buyers from the commerce platform every week, beside the ad account’s purchase count. Those two will disagree, and watching the size of the disagreement move tells you more than either figure alone.
New vs returning customers stops being a philosophical argument the moment both numbers sit in the same row of the same spreadsheet.
Frequency tells you more than the audience name does
A campaign is not prospecting because somebody typed prospecting into the name field.
In my US flag and banner ecommerce account, Meta sales campaigns reached 165,298 people between June and August 2026 at a frequency of 2.48. The average person in that pool saw the ads roughly two and a half times.
Frequency is the cheapest early warning signal available. Where it climbs while reach stays flat, the campaign has stopped finding anybody unfamiliar, whatever the naming convention claims. It will not label new vs returning customers for you, though it does tell you when a campaign has run out of strangers.
Performance inside that account varied enormously. My best ad set bought purchases at $29.37 on a 4.84 return, while the worst active one paid $93.96 on 0.98.
What I cannot tell you is how much of the best ad set’s revenue came from people who had bought before. Meta does not report it that way, and inventing the split would be worse than admitting the hole.
Blended reporting flatters new vs returning customers
That same account produced four defensible return figures in one period, and every difference was definitional.
Mixed attribution windows, as Meta blends them by default, gave 1.11. Every campaign including awareness came to 1.25. The six sales ad sets together read 1.58. Those same ad sets excluding one failed test read 1.99.
Not one of the four separates first-time buyers from repeat ones. A blended figure that also blends the audience is being asked two questions at once, and it answers neither properly.
Retention revenue sitting inside an acquisition judgement is the specific danger. It makes a weak engine look adequate, right up to the week growth stalls and nobody can explain the cause.
Judge acquisition on first-time buyers alone. Judge the business on everything together.
Two ceilings instead of one for new vs returning customers
Set them separately, because the two do not have the same shape.
Your acquisition ceiling comes from first-order contribution. Shopify reported an average order value of $114.70 across its full merchant base for Black Friday Cyber Monday 2025, published on 2 December 2025. Across my ecommerce clients gross margin runs between 20% and 50%, so an order that size contributes somewhere between $22.94 and $57.35.
Your retention ceiling works differently, since the person has already been paid for once. The only question worth asking is whether the paid touch caused the order or claimed one that was coming regardless.
Returns cut both. Salesforce put holiday returns at 14% of purchases, up 10% year on year, in its 2025 holiday shopping data published on 8 January 2026, drawn from 1.5 billion shoppers across 89 countries.
A first-time buyer who sends the order back cost the full acquisition price and left nothing behind.
What to change this week
Three things, and the first one takes an afternoon.
Upload your customer list and exclude it from every prospecting campaign. Leave it alone for a fortnight afterwards, and let the reporting get worse before you pass judgement on it.
Then pull first-time buyer counts from your commerce platform and set them beside the ad account’s purchase count every week. The distance between those two lines is the figure you have been missing.
Then decide, deliberately, how much paid budget your existing buyers deserve. Where email is working properly, the honest answer is usually less than you are spending now. Splitting new vs returning customers on the budget line is where this decision actually gets made.
Why a cheap enquiry is often the expensive one makes a similar argument on the lead generation side. What an online order really costs me at volume gives two dated figures with their sample sizes attached. How the acquisition calculation should be built covers what belongs in the numerator and the denominator, and why the famous rising-cost statistic is four years old covers what is genuinely happening to media prices. Audience structure and exclusions are the part of my Meta Ads work that shifts this fastest. To have the split checked in your own account, book a teardown.