New-to-Brand Orders vs New-to-Brand Customers

What Amazon is actually counting

Amazon's own definition covers "customers who purchase your brand or product for the first time on Amazon during the last year" — but the metric it reports is new-to-brand orders and new-to-brand sales, evaluated order by order against a rolling 12-month lookback. There is no separate "new-to-brand unique customers" figure in the standard reporting. Every qualifying order adds one to the count, and the same person can qualify again after their prior purchase ages out of the trailing-year window.

A worked example that shows the gap

Take a supplement brand running steady DSP spend for two years. In year one, a customer buys once in March — that's one new-to-brand order. They don't buy again until 14 months later, in May of year three. Because their prior purchase is now outside the trailing 12-month window, that May order also counts as new-to-brand. Two new-to-brand orders, one actual customer.

Scale that pattern across a subscription-adjacent or seasonal category — sunscreen, holiday gifting, allergy relief — where genuine repeat buyers naturally return on a cycle longer than 12 months, and a meaningful share of what the report calls "new-to-brand orders" is really "customers coming back after a gap." If your reported customer-acquisition cost divides ad spend by new-to-brand orders, and 10% of those orders are actually repeat buyers on a long cycle, your true acquisition cost is understated by roughly that same 10%.

Why this matters more than it sounds like it should

Customer-acquisition cost (CAC) is one of the few Amazon-adjacent numbers finance teams actually put in a board deck, and CAC math built on new-to-brand orders rather than distinct people will always look slightly better than reality in any category with a purchase cycle longer than a year — supplements, home goods, seasonal apparel, gifting. It looks worse than reality in the opposite case: a fast-consumable or subscription category where the same person can legitimately re-qualify within the window through gifting or a household's second account, inflating the apparent customer count without any real new person involved.

The direction of the error matters for how you use the number. If you're understating CAC in a slow-repeat category, a growth plan built on that number will assume acquisition is cheaper than it is, and the gap only shows up months later when lifetime-value assumptions stop reconciling with actual repeat revenue. That's a more expensive mistake to unwind than simply reporting a conservative number up front, which is the reason to flag the caveat before finance builds on top of it rather than after.

Why the metric is still worth trusting, with the caveat attached

None of this means new-to-brand orders is a bad metric — it's the only standardised, comparable acquisition signal Amazon reports natively, and it's directionally reliable for spotting whether an audience or campaign is recruiting at all. The problem is specifically what happens when someone downstream — finance, a growth model, a board deck — takes an order-level number and silently treats it as a person-level one, because nothing in the metric's name warns them not to. The fix is naming the gap once, clearly, wherever the number gets used for anything beyond a weekly trend line.

How to get closer to a real customer count

Amazon doesn't publish a native unique-customer metric to close this gap directly, but two things reduce the error. First, Amazon Marketing Cloud's Amazon Retail Purchases dataset extends the purchase-history lookback used for new-to-brand and lifetime-value calculations from 13 months out to 5 years — available as a paid AMC feature — which shrinks (though doesn't eliminate) the false-new-again pattern described above, because a much longer purchase history has to be clear before someone re-qualifies. Second, if you sell direct-to-consumer as well as on Amazon, cross-reference your own CRM's repeat-purchase distribution by category cycle length; if your typical repeat window off-Amazon runs 14-18 months, assume Amazon's 12-month on-platform metric is quietly recycling a similar share of "new" orders back through your CAC math.

The common mistake, including ours

We've reported new-to-brand order counts as "new customers acquired" in a client summary before, because that's the phrasing clients actually ask for and the distinction feels pedantic in the moment. It isn't pedantic once a finance team starts dividing total marketing spend by that number to get a CAC figure that goes into a growth model — at that point the gap between orders and people becomes a real forecasting error, not a rounding one. We now say "new-to-brand orders" explicitly in reporting and flag the category-cycle caveat rather than letting the shorthand imply a distinct-person count it can't actually support.

When the gap is distorting a real decision

If a category has a natural repeat cycle longer than 12 months and new-to-brand orders are feeding a CAC number used for paid-media budget decisions, don't try to hand-correct the Amazon figure with a guessed discount factor — that just substitutes one unverified number for another. Instead, pull the actual repeat-purchase distribution from your own order history (Amazon Seller/Vendor Central order reports, or your DTC CRM if you sell there too) and use the real median repeat interval to judge how much of the new-to-brand order count is plausibly a returning customer, rather than assuming the 12-month window is a good proxy for your specific category.

If that analysis shows a material gap — say your category's real median repeat interval is 16-18 months against Amazon's 12-month window — the honest move is to state the correction as a range rather than a single adjusted number, and to keep using new-to-brand orders as the operating metric day to day. Rebuilding your entire reporting stack around a custom-adjusted "true new customer" figure usually costs more in reporting complexity and internal disagreement than the correction is worth, unless the category gap is large enough to be materially changing a budget decision.