Who Are Your Amazon Ads Really Buying? New vs Existing Customers
How much of your ad spend re-buys existing customers, which campaign types really win new ones, and what each new customer costs.
A big share of your Amazon ad sales go to people who already buy from you. On the consumable brands I've analysed, it was often a third to a half. That doesn't make those ads worthless, but it does mean ACoS and ROAS can't tell you whether your advertising is actually growing the business.
The number that can is cost per new customer. This guide covers what new-to-brand data shows, which campaign types really bring in new customers, and how to start judging your ads on it.
What "new-to-brand" means
Amazon counts a purchase as new-to-brand when the shopper hasn't bought from your brand in the previous 12 months. Everything else is a repeat customer.
You can see new-to-brand figures in your campaign reports in the ads console. Amazon Marketing Cloud gives you a much more detailed view: new-to-brand by targeting type, placement, match type and keyword, and what those customers went on to do afterwards.
How much of your ad spend re-buys existing customers
Across several consumable products I looked at in AMC, 33% to 55% of ad-attributed purchases went to existing customers.
Some of that is unavoidable. Customers search your brand name, see your ad and click it. But it means two campaigns with the same ACoS can be doing completely different jobs: one finding new customers, the other collecting sales you'd probably have got anyway.
On one product, branded search made up half of all ad-attributed purchases and had the lowest new-to-brand rate in the account.
Which targeting types bring in new customers
The order was the same in every account I analysed:
| Targeting type | Share of ad purchases from new-to-brand customers |
|---|---|
| Category and competitor product targeting | 73–83% |
| Generic keywords | 67–72% |
| Auto campaigns | 62–67% |
| Targeting your own product pages | 41–61% |
| Branded keywords | 41–49% |
Overall, new-to-brand share ran from 45% to 84% depending on the brand. It was highest where organic pull was weak, such as a course supplement at 84%, and lowest on a well-known brand where ads were covering a lot of branded demand.
Cost per new customer by campaign structure
New-to-brand share is only half the picture. What matters is how much you pay for each new customer, which is ad spend divided by new-to-brand customers.
Here's what that looked like across one account over 12 months, shown as an index where 100 is the account average:
| Campaign structure | Cost per new customer (account average = 100) |
|---|---|
| Branded keywords | ~45 |
| Targeting your own product pages | ~63 |
| Auto | ~78 |
| Isolated single-ASIN competitor targeting | ~79 |
| Generic exact match, top of search | ~129 |
| Sponsored Brands | ~140–170 |
| Bundled competitor targeting | ~160 |
| Catch-all auto across every ASIN / Sponsored Display competitor | ~300+ |
Branded looks cheapest, but be careful with it. Plenty of those "new" customers already knew the brand and were searching for it by name. They're cheap partly because many would have found you anyway.
A few things stood out from the rest:
Isolated campaigns won new customers at about half the cost
Competitor ASINs in their own single-target campaigns brought in new customers at roughly half the cost of the same targets bundled together. Part of that is because winners are the ones that get isolated, but the gap was big and consistent.
In the same account, spend on these isolated campaigns had quietly fallen by about three-quarters year on year, despite them having the best cost per new customer in the account.
Head terms can be the most expensive way to grow
One big generic keyword cost about 1.9 times as much per new customer as a more specific term. It also had the lowest new-to-brand share of any generic keyword, at around 55%.
A deliberate push to win back rank on head terms roughly doubled the account's cost per new customer over the year, and bought fewer new customers, not more.
Top of search isn't always the bargain it looks
Top of search was the cheapest placement overall, because it converts so well. But that average was propped up by branded traffic. Within generic exact-match keywords, the top slot cost about 43% more per new customer than other placements.
Broad match was cheapest per new customer
Across match types, broad was the cheapest per new customer and exact was the most expensive, even though exact took the biggest share of spend.
Not all new customers are equal
Winning a new customer is only the start. What they do next matters just as much.
- Ad-acquired customers subscribed less. Customers who found the brand organically signed up to Subscribe & Save at 1.15 to 1.83 times the rate of ad-acquired customers. The gap was biggest where brand loyalty was strongest.
- Some "new" customers were really old ones coming back. On one brand, about 10% of ad-acquired customers had bought organically first and were then recaptured by an ad.
- Some brands depend on ads for almost all their growth. On one course supplement, about 85% of new customers had touched an ad before buying. When spend doubled, new-customer numbers rose with it.
All of these are correlations, not proof. If a decision is big enough, test it properly with a holdout.
Judge acquisition on payback, not ACoS
Here's how long it took a new customer to pay back their acquisition cost on one brand, by campaign type:
| Campaign type | Share of ad spend | Months to pay back a new customer |
|---|---|---|
| Auto | ~12% | ~6 |
| Competitor, category and generic | ~25% combined | ~8–9 |
| Branded | ~59% | 12+ |
The account's overall ACoS looked disciplined, but only because branded spend was holding it down. The non-branded campaigns were running at an ACoS about 1.5 times higher than the first-order break-even, and they still paid back within 9 months. A flat ACoS target would have cut the campaigns actually growing the business.
My recommendation to that brand was simple: judge acquisition campaigns on cost per new customer against a payback target, not on ACoS.
How to start tracking it
- Add new-to-brand columns to your campaign reports. New-to-brand orders, sales and percentage are all available in the ads console.
- Work out cost per new customer, not just cost per order. On the same product they differed by 2 to 3 times.
- Group campaigns by job, such as branded, own-product, competitor, generic and auto, and compare cost per new customer for each.
- Set a payback target. Use the Amazon LTV calculator to see what a customer is worth and when they pay back.
- Wait at least 14 days before reading new-to-brand figures for a month, so late conversions are counted.
- Use AMC for the detail if you have it: new-to-brand by placement, match type and keyword, and what those customers did next.
The bottom line
ACoS tells you how efficiently you're getting sales. It doesn't tell you whose sales they are. For a consumable brand, the customers who matter most are the new ones, because they're the ones who'll reorder for months to come.
Start tracking cost per new customer by campaign type. It usually changes where the budget should go.
Related reading: How to calculate Amazon customer lifetime value and TACoS explained: why it misleads repeat-purchase brands.
Want more like this? Subscribe free for practical, data-backed guides to growing a consumable brand on Amazon.
