Most Amazon ad reporting answers one question: which sales did an ad touch? The question that matters to a founder is different. Which sales would not have happened without the ad?
That gap is where budget quietly leaks. Branded search, retargeting, and high-intent placements often look efficient because they sit close to a purchase that was already going to happen.
Why ROAS overstates performance
Attributed ROAS credits an ad a shopper interacted with before buying. It does not separate demand you created from demand you simply intercepted.
- Branded keywords capture shoppers who were already looking for you.
- Retargeting reaches people who were close to buying anyway.
- Upper-funnel media gets little credit because its effect is delayed and indirect.
Efficient is not the same as incremental.
What AMC makes possible
Amazon Marketing Cloud is a clean-room environment where advertisers can query pseudonymized, event-level advertising data with SQL. Results are returned in aggregate, which protects shopper privacy while still allowing far deeper analysis than standard reports.
Instead of looking at one touchpoint at a time, AMC lets you study full paths to conversion across campaign types, placements, and time.
Three analyses worth running first
- Path to conversion. Which combinations of touchpoints precede purchases, and how often do shoppers convert after only branded search?
- Exposure versus non-exposure. How do conversion rates compare between shoppers who saw a campaign and similar shoppers who did not?
- New-to-brand by path. Which campaigns bring in customers who have not bought from you before, rather than re-buying existing customers?
Holdouts are the gold standard
Observational analysis gets you closer to the truth, but the cleanest answer comes from deliberately withholding spend. Pausing a campaign for a defined audience, region, or period and measuring the difference in sales tells you what the ads actually caused.
The design matters. Holdouts need enough volume, a stable comparison period, and controls for seasonality and promotions. Run poorly, they produce confident answers that are wrong.
Turning analysis into budget decisions
The output is not a report. It is a reallocation: less spend on placements that mostly harvest existing demand, more on the ones that create new buyers. For many brands, the result is the same total budget producing meaningfully more growth.