- The same 4x can be a profit or a loss. In our illustrative month it is $88,000 ahead if the ads caused every sale, and $52,000 behind if returning buyers would have bought anyway.
- We found no study with a general share of ad-attributed revenue that comes from existing customers. The platforms hold it and publish nothing.
- The much-quoted rise in Advantage+ new-customer cost from $257 to $528 comes from one vendor, whose post gives two date ranges for it. [5]
- The test that needs no benchmark: required new share equals 1 / (blended ROAS × first-order margin).
- Google’s Conversion value column includes the new-customer uplift you typed in. [2]
- Meta’s API documentation says the existing-customer budget percentage is gone from new Advantage+ campaigns. [3]
- New against returning isn’t caused against not caused. Published experiments cut both ways. [15][16]
Take one month in a made-up account. Spend is $100,000 and attributed revenue is $400,000, so the dashboard shows a 4.00x ROAS, and everybody goes home happy. Every input below is ours, built to show the mechanics, and none of it describes a real brand.
| Illustrative month | New customers | Returning customers |
|---|---|---|
| Attributed orders | 1,500 | 2,500 |
| Average order | $80 | $112 |
| Attributed revenue | $120,000 | $280,000 |
| Contribution margin | 40% | 50% |
| Contribution before ads | $48,000 | $140,000 |
Now read that month twice.
+$88,000 after spend. All $188,000 of contribution counts, and the 4.00x looks modest.
-$52,000 after spend. Only the $48,000 of contribution from new customers counts, on $120,000 of revenue, or 1.20x.
Neither reading is the truth. It usually sits between them, and it can fall below the second if some of those new customers would have bought anyway. The dashboard can’t tell you where, because 4.00x is the same number in all of them. That $140,000 gap is what this piece is about.
The benchmark you’d reach for isn’t there
The natural next step is to ask what share is normal. We looked, and we found no study that measures it across brands. The platforms hold the data and publish nothing, and the figures that circulate are thinner than they appear.
The most repeated one says the new-customer acquisition cost of Advantage+ campaigns went from $257 to $528, in an analysis of 55,661 Meta campaigns. It comes from Wicked Reports, which sells attribution built on that metric. Its post attributes the same two numbers to May 2024 against May 2025 in one place, and to March through May 2025 in another. [5] The same company’s benchmark for a week in July 2026 has Meta at $91 and “89% new,” for all of Meta, across a different population. [6] Those two can’t be compared, which is the point. Neither should be carried into your planning.
The story attached to that number, that automation drifts toward people who already buy from you, isn’t settled either. Haus looked at 640 Meta incrementality tests and found Advantage+ put 70% of its measured impact on new customers against 65% for manual campaigns, with retargeting spend flat at about 15%. [7] Haus sells those tests and its customers chose to run them, but they are experiments, and they point the other way.
The ten-second test
You don’t need a benchmark, because the share you need is arithmetic. If the first order has to cover the ads, new-customer revenue times its margin has to reach spend. Rearranged, that gives one over blended ROAS times first-order margin.
The new-customer share your blended ROAS requires
Use the contribution margin on a first order: revenue minus cost of goods, shipping, payment fees, and expected returns, with marketing left out.
Model: 1 / (blended ROAS × margin). A result above 100% means no mix can pay on the first order, so the plan depends on repeat purchases. That can be a fine plan, as long as you know it’s the plan. The test assumes every attributed sale was caused by the ads, which is generous.
Our month needs 62.5% and has 30%. That one comparison ends most arguments about whether the account is healthy.
Be careful with the raw share, though. Repeat customers placed 80.2% of Wayfair’s orders in the second quarter of 2026 [8], and existing customers were about 70% of Revolve’s net sales in 2018. [9] A campaign that merely tagged sales already on their way would show a similar mix. So we compare the attributed new share to the new share of the whole store, a ratio we call the mirror index. It’s our construct, not an industry metric. If new customers are 24.9% of store revenue and 30% of attributed revenue, the index is about 1.2, and the ads look only slightly different from a random sample of your sales.
Splitting it with what you already have
Write down what new means
Google’s auto-detection means no purchase within 540 days. [1] Meta’s automatic exclusion reportedly defaults to 730 days. [14] Shopify counts any earlier order. [11] A buyer last seen 600 days ago is new at Google, excluded at Meta, and returning in Shopify. Pick one lookback and send it everywhere.
Start from the store
Use order-level data, where Shopify has a new or returning label on each order. [10] Take first-time buyers and new revenue net of refunds. Avoid customer-level reports for this, because they read a customer’s whole order history, including orders placed after your period. [11]
Your ground truthGoogle: read the right column
The native split needs a lifecycle goal and doesn’t backfill. [1] An option called Report on new customers acquired, spotted in the interface in August with no announcement, appears to give the split without touching bidding. [12] If you’ve set a new-customer value, read Original conversion value, because Conversion value includes your uplift. [2]
Meta: define the audiences first
Set your existing customers in ad account settings, then use the audience segments breakdown. The API field is user_segment_key. [4] Anyone Meta can’t match lands in New Audience, and Engaged Audience means no purchase yet. [13] So a store-new figure is the New row plus the Engaged row. That sum is our reading, not Meta’s definition.
Reconcile
Add the new customers each platform claims and divide by first-time buyers in the store. If Meta claims 1,900 and Google 1,300 against 2,800 real ones, that’s 114%, and at least one of them is double counting.
Where the split flatters you
Splitting the number doesn’t make it clean. Most of the errors below push in the same direction, toward a better new-customer figure than you’ve earned.
| Trap | What happens | In our month |
|---|---|---|
| Unmatched buyers read as new | Google can count anyone missing from your list as new [1], and Meta files the unmatched under New [13] | If only 70% of returning buyers are recognized, a true 1.20x reports as 2.04x (our model) |
| The uplift sits in the revenue column | Conversion value adds what you specified [2], and Google suggests starting near twice average order value [17] | A flat $200 uplift turns 4.00x into 7.00x |
| Repeat orders inside the window count as new | Google says repeat purchases by a new customer are classified as new [1] | New value is no longer first-order value |
| Share of orders used as share of revenue | It assumes both groups spend the same | 1.50x against a true 1.20x |
| Different email, or a store purchase | Shopify holds one email per profile [19], and Google doesn’t split offline store sales [18] | Even your ground truth has an error rate, so test it with a phone or address match |
One piece of older advice no longer fits. Plenty of guides, including our own June piece on Advantage+, tell you to set an existing-customer budget cap. Meta’s Marketing API documentation now says the field isn’t available for new Advantage+ campaigns, and points to two ad sets, one including the audience and one excluding it. [3]
Who pays for the spend, and when it comes back
Dividing new revenue by all spend assumes that money shown to returning customers did nothing. Dividing by only the new-customer portion assumes it fully earned its keep. Neither is known, and in our illustrative month the assumption moves the answer a long way.
| Share of spend charged to new customers | New-customer ROAS | Cost per new customer | Flat payback |
|---|---|---|---|
| 100% | 1.20x | $66.67 | 12.4 months |
| 80% | 1.50x | $53.33 | 7.6 months |
| 60% | 2.00x | $40.00 | 2.9 months |
| 40% | 3.00x | $26.67 | Paid on the first order |
Report the first and last rows together. A single number here is an opinion about causality presented as a fact.
Months until a new customer pays back
Take new-customer revenue from your store rather than from a platform, and take monthly repeat contribution from a real cohort.
Model: (spend - first-order contribution) / new customers / monthly repeat contribution. A negative result means the first order already paid. The flat rate is optimistic: with repeat buying that decays, our month pays back in month 15 rather than 12.4. Treat this as a lower bound and check it against a cohort.
What the split can’t tell you
New against returning is a sort by who bought. It says nothing about what caused the purchase, and this is where a clean spreadsheet starts to mislead.
The eBay paid search experiments are the warning in one direction. Frequent shoppers accounted for most of the sales attributed to paid search and didn’t respond to it, while new and infrequent users did. Measured with ordinary methods, the channel’s return exceeded 4,100%. Measured with the experiment, it was -63%. [15] That was non-brand search in 2012, for a very strong brand, so don’t carry the number over. Carry the pattern over: attribution gathers around people who were likely to buy.
Lewis and Reiley are the warning in the other. Display ads shown to a retailer’s existing customers returned more than 11 times their cost, and 93% of the effect showed up offline, where a pixel never looks. [16] A team that sees “returning” and reads “waste” could switch off the part that works.
Google’s documentation says plainly that incrementality testing of lifecycle goal settings isn’t supported inside Google Ads. [1] So the split is a question, and the answer comes from a holdout test, run on new and returning customers separately. Until then, bring four numbers to the budget meeting: blended ROAS, the share you need, the share you have, and new-customer ROAS on both full and allocated spend. It’s less tidy than one number. It’s also the first version a finance lead can’t take apart in a minute, and once you know the split, the value of the next dollar is the question that decides the budget.
Sources
- Google Ads Help · Measure your lifecycle goals campaignsthe 540-day window, list gaps counted as new, repeat purchases labeled new, no retroactive data, and no in-platform incrementality testing
- Google Ads Help · Step 2 of 5: Configure your lifecycle goalsthe Conversion value column includes the additional value you specify
- Meta for Developers · Advantage+ campaign experiencethe existing customer budget percentage is unavailable for new campaigns
- Meta for Developers · Insights API breakdownsthe user_segment_key breakdown and how existing users are specified
- Wicked Reports · Meta Advantage+ vs Manual: 55,661 campaigns analyzedthe $257 to $528 figure with both date ranges, from a vendor selling attribution on this metric
- Wicked Reports · The Paid Traffic Truth: Why your Meta cost rose but your click price didn’tMeta at $91 and 89% new, all Meta, client accounts only
- Haus · The Meta report: lessons from 640 incrementality experimentsvendor data from self-selected customers, published July 2025
- Wayfair · Q2 2026 results, SEC exhibitrepeat customers placed 80.2% of orders delivered
- Revolve Group · Form S-1/Aexisting customers about 70% of 2018 net sales
- Shopify Help Center · Analytics fieldsthe order-level new or returning customer label
- Shopify Help Center · Customers reportsfirst-time and returning definitions, and reports that read the entire order history
- Search Engine Roundtable · Google Ads report on new customers acquiredin-product text spotted by practitioners, with no Google announcement
- Jon Loomer · A guide to audience segmentspractitioner account of how unmatched people and engaged audiences are filed
- Jon Loomer · Meta adds automatic customer exclusionsone practitioner’s account of the 730-day default, October 2026
- Blake, Nosko, and Tadelis · Consumer heterogeneity and paid search effectivenessEconometrica 2015, a large-scale field experiment at eBay
- Lewis and Reiley · Does retail advertising work?a Yahoo display experiment matched to a retailer’s customer database
- Accelerate with Google · Best practices for new customer acquisition goalsthe suggested starting point of twice average order value
- Google Ads Help · Using new customer acquisition goal with store goalsnew customers are not reported for offline store sales conversions
- Shopify Help Center · Managing customerseach customer profile holds one email address
Frequently asked questions
What is new-customer ROAS and how do I calculate it?
It is revenue from first-time buyers divided by ad spend. Take the revenue from your store’s order export, net of refunds and discounts, rather than from the platform’s conversion value column. Divide by all the spend for a full-load figure, and by the spend on new-customer segments for an allocated figure. Report both, because the gap between them is an assumption about whether spend on returning customers does anything.
What share of my ROAS should come from new customers?
We found no study that gives a general figure, and the platforms that hold the data do not publish one. Calculate the share you need instead: one divided by blended ROAS times first-order contribution margin. At 4x and a 40% margin, 62.5% of attributed revenue has to be new for the ads to pay on the first order.
Why did my Google ROAS jump after I turned on new customer acquisition?
Check whether you set a value for new customers. Google says the Conversion value column includes both the original purchase value and the additional value you specify. In our illustrative month, a flat $200 uplift turns a 4.00x into a reported 7.00x with no change in sales. The Original conversion value column shows the figure before the adjustment.
Can I still cap existing customers in Advantage+?
Not with the old field in new campaigns. Meta’s Marketing API documentation says the existing customer budget percentage is not available for new Advantage+ campaigns. Its replacement is two ad sets, one including and one excluding the audience. Meta also opened a Customer Lifecycle Strategy to all advertisers in October 2026.
How do Google and Meta decide someone is a new customer?
Google’s auto-detection counts someone as new if they have not purchased within 540 days, and a person missing from your customer list can count as new as well. On Meta you define existing customers with your own audiences, and practitioner Jon Loomer reports that anyone Meta cannot match falls under New Audience. Shopify counts any earlier order on the account. One buyer can be new on one system and returning on another.
How do I calculate CAC payback for an e-commerce brand?
Subtract first-order contribution from the cost of acquiring the customer, then divide by the contribution a customer brings per month after that. That flat version is optimistic, because repeat buying usually slows. The proper version follows a cohort and finds the month when its cumulative contribution per customer reaches the acquisition cost.
Is new-customer ROAS the same as incremental ROAS?
No. It sorts sales by who bought, not by what caused them. A new customer can be a sale the ad did not cause, and a returning customer can be a sale it did. Google states that incrementality testing of lifecycle goal settings is not supported within Google Ads. Only a holdout answers that question.
Mikołaj Salecki
Writes about media, tech, and AI business for people who actually run digital. Former agency lead. Skeptic of frameworks that read better than they perform.
