ROAS Cannot Tell You If You Are Growing: New to Brand Amazon Can

New to brand Amazon is the only standard ad metric on the platform that distinguishes a customer you won from a customer you already had. ROAS cannot make that distinction and was never built to. On a brand with real awareness, a large share of ad-attributed sales would have happened without the ad, which is how spend keeps climbing while net revenue flattens and nothing in the report explains why.

• ROAS measures how efficiently you harvest existing demand, not whether you are creating any.
• New-to-brand counts orders from shoppers who have not bought your brand on Amazon in the prior 365 days.
• The metric is not reported on Sponsored Products, which is usually where most of the budget sits.
• Shifting spend from branded to non-branded targeting moves the rate, and the effect is visible within weeks.

ROAS Measures How Efficiently You Harvest, Not Whether You Are Growing

Amazon ROAS divides ad-attributed sales by ad spend, and that is the whole of its definition. A branded keyword converts because the shopper already decided. The ad appears, the click happens, the order lands, and the platform attributes the sale. Nothing about that sequence establishes that the ad caused anything.

That is the whole of the problem, and it does not require any bad faith from Amazon to produce it. Every advertising platform reports the value it created as measured by itself, using an attribution window it also sets. A strong ROAS on a brand with established awareness is a description of harvesting efficiency. It is a perfectly good number for the question it answers. It is simply not the question a growth target is set against.

New to Brand Amazon Counts First-Time Buyers, and the Definition Matters More Than the Number

Amazon classifies an ad-attributed order as new-to-brand when the shopper has not purchased any product from your brand on Amazon in the prior 365 days. Three details in that sentence carry most of the weight.

• The lookback is brand-level, not ASIN-level. Buying any product from your catalogue in the past year disqualifies that shopper from new-to-brand status on a later order.
• The window rolls. It is a moving 365 days, never lifetime history, so a customer who last bought fourteen months ago counts as new again.
• It is a family of metrics, not one figure. Amazon NTB orders and sales are raw counts. Percentage of orders and percentage of sales are derived. New-to-brand order rate, specific to Sponsored Brands, divides new-to-brand orders by clicks.

Amazon extended Sponsored Brands attribution to report new-to-brand at SKU level rather than campaign level only, which is the more useful change of the past year. It makes it possible to see which products actually recruit and which ones simply re-sell to people who already knew you.

One thing this article will not give you is a benchmark. You will find plenty elsewhere: above 50% for top-of-funnel, 60 to 80% for category keywords, below 15% for brand defence. None of the sources publishing those figures says where they come from, and a category median is close to meaningless on a metric whose lookback window does not match the purchase cycle it is measuring. Read your own trend line. Direction is the signal. Level is noise borrowed from someone else’s catalogue.

The Metric Is Blind Exactly Where Your Spend Is Heaviest

This is the structural flaw nobody mentions, and it changes how the number should be read.

New-to-brand is reported natively on Sponsored Brands, Sponsored Display, DSP and Sponsored TV. It is not reported on Sponsored Products. For most accounts, Sponsored Products carries the largest share of budget, which produces an uncomfortable consequence: account-level new-to-brand understates true acquisition, because the format doing most of the spending is invisible to the metric.

Two implications follow, and they point in opposite directions.

Do not compare your account-level rate to anyone else’s. Two brands with identical acquisition performance will report very different figures depending on their format mix. A DSP-heavy account looks like an acquisition machine. A Sponsored Products-heavy account looks like it recruits nobody. Neither reading is true.

Amazon Marketing Cloud closes the gap. New-to-brand is available across all formats there, including Sponsored Products, and it also reveals which combination of touchpoints produced the acquisition rather than attributing it to the last format in the path. That usually shifts credit toward upper-funnel placements that standard reporting under-values.

Until you look at it that way, the number you are reading is a partial view presented as an account total. This is also why NTB Amazon reporting rewards accounts that already run upper-funnel formats and penalises those that do not, regardless of how well either one actually acquires.

The Twelve-Month Window Is a Convention, Not a Fact About Your Category

The 365-day lookback is a platform decision, applied uniformly across categories that behave nothing alike.

For a consumable bought every few weeks, twelve months is generous: a genuine lapsed customer is correctly counted as new. For a considered purchase with a multi-year replacement cycle, twelve months is close to meaningless, because almost every buyer qualifies as new-to-brand and the metric approaches a simple order count. Some retail media networks extend the window to 18 or 24 months for exactly this reason.

The practical correction is not to abandon the metric. It is to stop reading the absolute rate as an acquisition truth and start reading the delta. If your rate moves 6 points after a targeting change, the window did not change and the category did not change. Something you did changed.

Branded Spend and New to Brand Amazon Move in Opposite Directions, Which Is the Lever

Here is the part that is actionable within weeks rather than quarters.

Branded campaigns produce a low new-to-brand rate by construction. The shopper typing your name is almost never new to your brand, so branded spend converts beautifully on ROAS and contributes close to nothing on acquisition. Non-branded category terms and competitor targeting do the opposite: lower ROAS, materially higher new-to-brand.

Which means the rate is not a passive indicator you observe. It is a direct function of your targeting mix. Every point of budget moved from branded to non-branded and competitor terms mechanically moves the number, and the effect shows up in weeks rather than quarters because the metric is reported per campaign.

Two conditions make this more urgent than it was. Branded cost per click on Amazon has risen sharply over the past year as competitors bid on rival brand names, so branded defence costs more for the same harvest. And with new-to-brand now reported at SKU level, the reallocation can be surgical rather than portfolio-wide: fund the products that demonstrably recruit, defend the ones under genuine competitive attack, and stop paying full price for the rest.

That reallocation is the kind of decision Primelis Market is built to run continuously, detecting when branded advertising has stopped being incremental and moving budget toward the terms that produce first-time buyers.

A High New to Brand Rate With Weak Repeat Is a Treadmill

Now the correction, because everything above becomes dangerous without it.

New-to-brand is not a profitability metric. It counts acquisition. It says nothing about whether the customers acquired are worth acquiring. A high rate paired with a weak repeat purchase rate describes a business buying strangers who never come back, and doing so at a cost per acquisition that only the first order ever has to justify.

The two must be read as a pair:

• High new-to-brand, strong repeat. The spend is an investment. A higher first-order acquisition cost is defensible because subsequent orders carry no acquisition cost at all.
• High new-to-brand, weak repeat. The spend is a treadmill. Volume holds only as long as budget holds, and growth stops the day spending does.
• Low new-to-brand, strong repeat. A healthy customer base being harvested efficiently, and a ceiling approaching. This is the profile that produces flat net revenue on rising spend.
• Low new-to-brand, weak repeat. The account is buying single orders from existing customers at full price.

Anyone selling new-to-brand as a standalone north star is selling you a replacement vanity metric. It is a better metric than ROAS for the question of growth, and it is the wrong metric on its own for the question of whether that growth is worth having.

There is also a boundary worth naming: new-to-brand is measured inside Amazon. A customer new to your brand on Amazon may have bought from you directly last month, which is a different question entirely and one most brands have never asked. That question is treated separately.

What Changes When You Pilot on New to Brand Amazon Instead of ROAS

Among the Amazon advertising metrics available in the console, ROAS answers how efficiently the account converted the demand it found. New to brand Amazon answers whether the account found any demand that was not already yours. Both belong in the report. Only one of them moves in the same direction as the business.

The first read is quick and does not require a project. Pull your new-to-brand rate by campaign type over the last twelve months, separate branded from non-branded, and put your repeat purchase rate alongside it. Most accounts discover the rate is flat, that branded carries a disproportionate share of spend, and that nobody had ever looked at the two lines together.

That is the diagnostic our teams run before touching an Amazon budget, drawing on 15 years of work with more than 300 global brands and over $1B in annual media managed across 200+ specialists.