What a Good New-to-Brand Percentage Looks Like

Why a single benchmark number is the wrong question

Ask ten agencies for "a good new-to-brand percentage" and you'll get ten different numbers, because the honest answer depends on what the placement is built to do. A campaign designed to reach shoppers who already know your brand — retargeting, branded search, Alexa device inventory — will structurally run a low NTB rate and that's not a failure. A campaign built to reach people who've never seen the brand should run high, and if it doesn't, the audience or creative is the problem, not the concept of NTB itself.

Category matters too, independent of placement. A consumable or subscription brand with a high repeat-purchase rate will naturally see its blended NTB percentage compress over time as its own customer base grows and re-enters the denominator — that's the business working, not the campaigns underperforming. A brand launching its first SKU with no purchase history at all will see an artificially high NTB rate for the opposite reason: almost everyone genuinely is new. Neither number is comparable to a category benchmark pulled from a brand at a different stage.

The worked example that shows why blending is the trap

Here's the number that makes this concrete, from our own book: across 27 advertisers over 31 days in summer 2026, Amazon-owned shopping placements — the desktop site, mobile app, mobile web and Alexa devices — took $439,875 of spend and returned 42.0% new-to-brand. Off-Amazon third-party exchange inventory, in the same window, took $259,516 and returned only 15.9% new-to-brand — while actually posting a higher return multiple, 6.05x against 4.90x on the Amazon-owned placements.

Read that pair of numbers as one blended account average and you'd conclude exchange inventory is simply "better." Read them separately and the real story is that the two supply sources are doing opposite jobs: the exchange inventory is converting people already close to a decision, and Amazon's own shopping surfaces are the ones actually recruiting new buyers. A ROAS-only lens would quietly starve the placement doing the acquisition work, because it looks worse on the metric everyone checks first.

How to actually check this in your own account

Most advertiser consoles let you break new-to-brand out by campaign, and DSP reporting can go one level further, to supply source. Pull the last full month, sort campaigns by spend, and add new-to-brand percentage as a column next to ROAS rather than looking at either one alone. The campaigns worth a closer look are the ones where the two numbers move in opposite directions — high ROAS, low NTB is a retention engine wearing a prospecting label, and low ROAS, high NTB may be exactly the acquisition spend that's supposed to look inefficient on a return basis while it's actually doing its job.

Building your own benchmark instead of borrowing one

The workable approach is to set a target per placement type, not per account. Take your last 60-90 days of DSP or sponsored ads reporting, break new-to-brand share out by campaign type — prospecting, retargeting, streaming, Amazon-owned vs. exchange — and use each segment's own trailing average as its starting benchmark, adjusted for the objective you actually assigned that budget. A prospecting line that runs below its own trailing average for two consecutive reporting periods is worth investigating; a retargeting line running at 18% NTB is not underperforming, it's doing exactly what retargeting does.

• Prospecting / cold audiences: expect the highest NTB share in the account — treat a falling trend as the signal, not the absolute level.
• Retargeting / cart abandoners: expect the lowest NTB share by design; judging it against a prospecting target will always look like failure.
• Streaming / connected TV: often the highest NTB rate of any placement precisely because attribution undercounts the sale — covered in more depth on our streaming measurement pages.

The common mistake, including ours

We've reported a single account-level NTB number in a client update before, without breaking out the placement mix underneath it, because it was faster to write and the number happened to look good that month. It wasn't dishonest — the number was real — but it invited exactly the wrong question back: "why isn't NTB higher everywhere," when the honest answer was that most of the budget was in retargeting on purpose. Reporting the blend without the mix is the mistake; we don't do it anymore, and it's worth checking whether any report you're currently reading does it either.

When new-to-brand is genuinely low and needs fixing

Three checks, in order, before touching bids or budget. First, confirm the placement mix — pull the campaign-level or supply-source breakdown and see whether the low blended number is actually a mix problem, not a targeting problem. Second, check audience freshness — a prospecting audience left unrefreshed for months will decay toward people who've already been reached and converted, which quietly turns a prospecting line into a retargeting line without anyone changing a setting. Third, check for overlap with sponsored ads — if the same shopper is being hit by both DSP and branded search, reconcile in Amazon Marketing Cloud before concluding the DSP audience itself is the problem.

If all three come back clean and the number is still genuinely low for what the placement is meant to do, the honest next step is a smaller, cheaper one than most brands reach for first: don't rebuild the whole plan, isolate the single weakest supply source or creative and test a change against it in isolation, holding the rest of the account constant. An account-wide overhaul makes it impossible to know afterward which change actually moved the number — and on a metric this sensitive to mix, that's the mistake that costs the most time to undo.