New-to-Brand: What the Metric Actually Counts

The definition, precisely

Amazon's own guidance defines a new-to-brand purchase as one made by "customers who purchase your brand or product for the first time on Amazon during the last year." That's a rolling 12-month lookback against Amazon's own purchase history for your brand — not your lifetime customer list, not your Shopify orders, and not a calendar year. A shopper who bought from you in month 1 and buys again in month 15 counts as new-to-brand again, because the system only looks back twelve months from the second order.

Four metrics sit under this umbrella: new-to-brand orders, new-to-brand sales, % of orders new-to-brand, and % of sales new-to-brand. The percentage metrics are the ones worth watching day to day — raw counts move with spend and tell you nothing about mix on their own.

It's also worth being precise about which ad products even report it. New-to-brand metrics are available for Sponsored Brands, Sponsored Display and Amazon DSP — Sponsored Products historically has narrower coverage, so an account leaning heavily on Sponsored Products alone may have less NTB visibility than one running a full mix of ad types. Check which of your campaign types actually populate the metric before assuming a gap in the report means a gap in performance.

A worked example

Take a campaign that generates 400 attributed orders in a month, of which the report shows 22.0% new-to-brand. That's 88 new-to-brand orders and 312 orders from buyers who purchased the brand at some point in the trailing year. If average order value is $45, new-to-brand sales are roughly 88 × $45 = $3,960 against a total of $18,000 — so new-to-brand is 22% of orders but you'd want to check whether it's also 22% of revenue, because first-time buyers often trade down to a smaller pack size or a starter SKU, which would make the sales percentage lower than the order percentage.

On our own DSP book — across 30 advertisers in July 2026 — 20.1% of 57,137 attributed purchases were new-to-brand. That's the kind of number that looks fine sitting alone in a dashboard and only becomes useful once you can see it broken out by placement, which is where most of the real variance hides.

What the metric structurally cannot see

Three blind spots matter more than the headline number. First, the 12-month window is a default: Amazon's Amazon Marketing Cloud can extend the purchase-history lookback used for new-to-brand and lifetime-value calculations from 13 months out to 5 years, as a paid add-on, which changes the answer for durable and seasonal goods where a 12-month window is simply too short to tell repeat from new. Second, NTB only knows what happened on Amazon — a customer who has bought from your own Shopify store for years but never from your Amazon listing reads as brand-new, which is the exact seam the link-out and attribution pages on this site cover in more depth. Third, it's an order-level count, not a person-level one, so the same human can be counted new-to-brand more than once across different years without ever showing up as a repeat customer in the report.

The common mistake, including ours

The mistake we see most often — and have made ourselves when a client wants one number for a slide — is quoting the blended NTB percentage as if it describes every placement equally. It doesn't. In our own DSP book, we've seen the exact same account swing from double-digit to over 40% new-to-brand purely on the basis of which supply source the impression came from, which is a mix question, not a targeting-quality question. Reporting one NTB number for a whole account without saying which channels are driving it is the single fastest way to make a good acquisition campaign look mediocre, or a bad retargeting campaign look like it's finding new customers.

The second version of this mistake is subtler: comparing your new-to-brand rate against a number quoted in an industry deck without checking what that number was measuring. A figure describing sales driven by audience-targeted display campaigns specifically is not the same population as your account's blended new-to-brand rate across every campaign type, and treating the two as comparable sets an expectation nothing in your account was built to hit. Ask what population and what window sit behind any external benchmark before you adopt it as your own target.

When the number is bad news

If new-to-brand share is falling month over month with spend flat, check three things in order. First, audience overlap — has a prospecting audience quietly narrowed to people who already converted, through frequency capping or an exclusion list that expired. Second, budget shift — pull the placement-level breakdown before assuming targeting broke; a shift of dollars toward retargeting will lower blended NTB even if every individual audience is performing exactly as designed. Third, seasonality — a rolling 12-month window means last year's buyers roll back into the eligible pool every month, so a slow bleed downward can simply be your own repeat-purchase rate showing up in the denominator. None of these are fixed by pausing the campaign; they're fixed by reading the breakdown one level down from the summary number.

There's a fourth check worth running before any of the first three, because it catches a mistake we've made ourselves: confirm you're comparing the same lookback window period over period. A dashboard pulled on the 3rd of the month against one pulled on the 28th of the prior month isn't a clean comparison once you remember the window rolls daily, not monthly — a few days' difference in the pull date can move the denominator enough to look like a trend that isn't one. Re-pull both periods on matching day counts before reporting a decline as real.

reMKTR reports new-to-brand at the placement level on every DSP account we run, because the blended number on its own has already misled us once and we don't publish a client report built on it again.