How to Analyze Amazon Ads New-to-Brand Metrics

Amazon defines sponsored ads new-to-brand purchases as purchases from shoppers who have not purchased from the brand during the prior 12 months. It applies a comparable 12-month lookback to determine whether a detail page view is new to the brand. That is a platform-specific classification, not a lifetime customer identity and not necessarily the same as a retailer's CRM definition. Put the lookback and brand scope in every review so stakeholders do not read the label as never purchased anywhere before.

Amazon's help documentation states that sponsored ads new-to-brand metrics are available for Sponsored Brands and Display, while the same data is not available for Sponsored Products. Availability can also differ by account, campaign, marketplace, and reporting surface. Build an eligibility matrix before calculating portfolio rates. If a row is blank or absent, mark it unavailable or unsupported. Do not convert missing data to zero or blend unsupported Sponsored Products spend into a denominator that implies comparable acquisition measurement.

Choose one business question. Examples include whether a Sponsored Brands campaign reaches more first-time Amazon purchasers, whether a Display campaign has a defensible acquisition cost, or whether new-to-brand detail page views progress toward purchases. State the campaign set, promoted products, brand, reporting dates, currency, attribution view, and excluded rows. The question should name the decision it may support, such as creative review or a bounded budget test, rather than assuming the metric proves incremental growth.

Export the supported campaign report and preserve the console filters and download time. Review total spend, impressions, clicks, purchases, sales, and units beside their new-to-brand counterparts. Amazon documents percentage measures such as percent of purchases new to brand and percent of sales new to brand, plus rates and cost measures for applicable products. Recalculate any derived metric from the exported numerator and denominator when possible. This exposes filtered rows, rounded values, currency mixtures, and incomplete periods before they become a recommendation.

Keep counts, shares, rates, and costs distinct. A high percentage of purchases new to brand can occur in a small campaign with few total purchases. A campaign can drive many new-to-brand purchases while the percentage falls because repeat purchases grew faster. Cost per new-to-brand purchase requires verified spend and supported attributed purchases, while economic acquisition cost also needs margin, returns, fees, and repeat-value evidence. Present the count and denominator beside every rate so scale remains visible.

Segment only where the report supports a stable comparison. Useful cuts can include ad product, campaign, brand, marketplace, promoted product family, creative, audience, or time period. Avoid comparing a recent launch with a mature campaign without labeling inventory, price, promotion, seasonality, and learning differences. Keep currencies separate and use complete attribution windows. If a campaign changed name, targeting, creative, or product scope, create a version boundary rather than merging incompatible operation periods into one trend.

New-to-brand reports describe attributed behavior under Amazon's definitions. They do not by themselves prove that advertising caused a shopper to become new to the brand or that the purchase would not have happened otherwise. Brand search, retail availability, price, promotions, seasonality, external marketing, and organic discovery can move at the same time. Use careful language such as attributed new-to-brand purchases or observed share, and reserve incremental claims for an experiment or measurement design that can support them.

Investigate sudden changes before reallocating spend. Check whether campaign type, brand mapping, product eligibility, marketplace, attribution setting, creative, targeting, retail readiness, price, inventory, or reporting completeness changed. Compare the raw supported rows with the prior export and look for suppressed or newly available fields. A movement may reflect real customer mix, but it may also reflect a reporting boundary. Document the hypotheses and the evidence that would confirm or reject each one.

Join the report to product economics only at a governed grain. Map campaign and promoted products to verified landed cost, marketplace fees, fulfillment cost, returns, discounts, and contribution margin. If that mapping is incomplete, show new-to-brand cost as an advertising measure rather than a profitability conclusion. A lower cost per attributed new-to-brand purchase can still be unattractive for a low-margin product, and a higher cost can be reasonable for a product with verified repeat value. The metric needs economics, not a universal target.

Amazon introduced long-term sales and long-term sales ROAS measures that estimate the incremental sales value expected over the next 12 months from selected new-to-brand engagements. Amazon says the measures use historical 12-month return patterns and can appear in campaign and downloadable reporting for supported customers. Treat them as modeled forward-looking measures, not booked sales. Record the eligibility, metric definition, observation date, horizon, currency, and report source next to any use.

Do not add long-term sales to immediate attributed sales as if they were separate realized revenue. The long-term measure represents an estimate tied to an engagement and horizon, while current sales are observed within an attribution system. Use the metric to frame a hypothesis about customer value or to rank campaigns for deeper analysis. Verify that the advertiser, marketplace, ad product, and report support the measure, and show unavailable values honestly. Model changes can also break comparability across time.

Convert the analysis into a bounded action. Choose one campaign or creative, define the expected acquisition metric and economic guardrails, set a review window, and specify rollback conditions. Preserve the baseline report and recheck retail readiness before launch. After the window closes, compare complete attributed counts, rates, spend, economics, and any supported long-term measure using the same definitions. Record the result as observed evidence. Do not claim causal lift or lifetime value until the measurement design and realized customer data can support it.