How to Measure Incrementality on Amazon Ads Using AMC

The Number That Changes Everything

Here’s a scenario I encounter regularly with Amazon brands that are scaling ad spend. They’re running Sponsored Products on their brand keywords, generating strong ROAS — say 8x or 10x — and the dashboard looks healthy. Everything seems to be working.

Then we run an incrementality test. We withhold ads from a small group of their customers for two weeks and measure what happens. Sales in that holdout group drop slightly, but not nearly as much as the ROAS number suggested they would. The people in that group who didn’t see the ads still bought, just through organic search instead of clicking the sponsored ad.

The ‘8x ROAS’ campaign was actually driving very little incremental revenue. Most of those buyers would have found the product organically anyway. The ad was taking credit for sales it hadn’t caused.

This is the incrementality problem, and it affects most Amazon advertising accounts to some degree. Sponsored keywords on branded terms, retargeting audiences who were already planning to buy, DSP campaigns reaching existing loyal customers — all of these can produce strong attributed ROAS numbers while delivering very little genuine incremental growth. Amazon Marketing Cloud gives you the tools to find and fix this, and this article shows you exactly how.

Attribution vs Incrementality: The Critical Difference

Most brands running Amazon ads are measuring attribution, not incrementality. They sound similar but they answer completely different questions.

The simplest way to put it: attribution tells you which ads were in the room when a sale happened. Incrementality tells you which ads actually caused the sale.

Both are useful. Attribution helps you understand the customer journey and optimize budget allocation across touchpoints. Incrementality tells you whether the whole system is generating genuine growth or just taking credit for organic demand that already existed.

Introducing iROAS: The Metric That Replaces Bare ROAS

Incremental ROAS (iROAS) is the incrementality-adjusted version of ROAS. Instead of measuring all revenue attributed to your ads, it measures only the revenue that genuinely would not have happened without the ad.

The iROAS formula

A worked example makes this concrete. A home goods brand ran an Amazon DSP campaign to drive awareness for a new product line and used AMC to set up a holdout group:

• Test Group (saw the ads): Generated $250,000 in sales during the campaign period
• Control Group (did not see the ads): Generated $175,000 in sales organically during the same period
• Incremental Revenue: $250,000 minus $175,000 = $75,000
• Incremental Lift: ($75,000 / $175,000) x 100 = 42.8% lift
• If ad spend was $30,000, iROAS = $75,000 / $30,000 = 2.5x

The campaign’s standard ROAS might have shown $250,000 in attributed sales divided by $30,000 in spend, giving an 8.3x ROAS. The true incremental ROAS was 2.5x — a very different picture. Both numbers are accurate within their own frame. The iROAS number is more truthful about what the ad actually achieved.

The Cannibalization Problem: Where Most Amazon Brands Lose Budget

Ad cannibalization happens when your ads generate sales that would have happened organically anyway — you’re paying for credit rather than for growth. On Amazon, this shows up in predictable patterns, and knowing where to look for it saves significant budget.

Where cannibalization is most common on Amazon

1. Branded keyword Sponsored Products

When a shopper types your exact brand name into Amazon search, they’re already looking for you. They would find your organic listing even without a paid ad. Bidding on your own branded keywords and paying CPC for those clicks often means you’re buying credit for sales that your organic ranking would have captured for free.

A well-documented geo-holdout test approach involves pausing branded search in a geographic test market for 2-3 weeks and measuring what happens. In many cases, 80-90% of traffic and sales that would have been attributed to branded ads return through organic channels when the paid ad is removed — revealing that most of the branded search spend was buying credit, not driving incremental growth.

2. Retargeting over-saturated audiences

Retargeting audiences — people who already visited your product page — have high intent. They often convert at high rates. But if those people were planning to buy anyway and you retarget them 12 times in a week, you’re spending CPM budget to reach an audience that didn’t need convincing. The high ROAS looks great. The incremental contribution is minimal.

3. DSP campaigns reaching existing repeat customers

Past purchasers who regularly replenish your consumable product may buy again regardless of whether they see a DSP ad. If your campaign audience is heavily weighted toward loyal existing customers, your attributed ROAS benefits from their predictable purchase behavior without your ad actually influencing it.

4. Retargeting in overlapping campaigns

When Sponsored Products, Sponsored Display, and DSP retargeting campaigns all target the same users simultaneously without exclusion logic, those users see dozens of your ads per week. Many of them were already planning to buy. Every campaign attributes the sale to itself. The incremental contribution of any individual campaign is much lower than it appears.

How Amazon Marketing Cloud Measures Incrementality

AMC gives you two approaches to measuring incrementality, and they serve different purposes. Understanding which to use when makes the analysis much more actionable.

Method 1: Holdout Studies in AMC

A holdout study is the gold standard for incrementality measurement. In AMC, you create a control group of shoppers who are eligible to see your ads but are intentionally suppressed from receiving them. After the study period, you compare the purchasing behavior of the exposed group versus the control group.

What makes AMC’s holdout studies particularly valuable is the precision of the control group definition. You can specify exactly which audiences to hold out — for example, ‘shoppers who match my DSP audience targeting criteria but will not receive DSP impressions during this period’ — giving you a genuinely comparable test and control population.

The output is a clear incremental lift percentage and iROAS calculation based on real Amazon purchase data, not modeled estimates.

Setting up a holdout study in AMC: practical steps

1. Define your test audience. Identify the campaign or audience type you want to test. Be specific — for example, ‘users in our DSP in-market prospecting audience for the Beauty category.’
2. Create the holdout group. In AMC, create a suppression audience that includes a randomly selected 15-20% of your target audience. These users match all your targeting criteria but will not receive ads during the test period.
3. Set the test duration. Run the test for at least 3-4 weeks to generate statistically significant data. Shorter periods may be affected by random variation or weekly sales cycles.
4. Maintain everything else constant. Don’t change bids, budgets, or targeting during the test period for the campaigns being measured. Any structural change resets the baseline.
5. Query the results in AMC. After the test period, run an AMC query comparing purchase rates, revenue, and new-to-brand rates between your exposed group and the holdout group.
6. Calculate iROAS. Incremental Revenue = (Exposed Group Revenue) minus (Control Group Revenue scaled to exposed group size). Divide by ad spend on the exposed audience to get iROAS.
7. Make budget decisions based on the output. An iROAS below 1.0 means you’re spending money on sales that would have happened anyway. Anything above 2.0 generally indicates strong incremental performance worth scaling.

AMC’s Customer Path Reporting has expanded significantly — it now supports a lookback window of up to 5 years, allowing brands to analyze long-term purchase cycles, replenishment patterns, and repeat customer behavior in ways that weren’t possible before. This extended horizon matters especially for incrementality in consumable categories where purchase cycles span months.

Method 2: Path-to-Purchase Multi-Touch Analysis

While a holdout study is the most rigorous incrementality test, AMC’s path-to-purchase analysis provides ongoing incremental insight without requiring the revenue trade-off of withholding ads from a control group.

A path-to-purchase query shows you the sequence of ad touchpoints that appeared before every conversion in your account. By examining which touchpoints appear early vs late in customer journeys — and which ones appear in journeys that resulted in genuine new-to-brand purchases vs repeat purchases — you can identify which campaigns are driving incremental acquisition versus capturing existing demand.

What to look for in path-to-purchase data

• Campaigns that appear as ‘first touch’ frequently: These are your true demand-creation campaigns. They introduced the shopper to your brand before they were in the market. These tend to be incrementally valuable even if their direct ROAS looks weak.
• Campaigns that appear only as ‘last touch’: These often capture demand that was already there — particularly branded search campaigns. They may be partially cannibalizing organic.
• New-to-brand rate by touchpoint sequence: Paths that contain a DSP awareness ad first tend to have higher NTB rates than paths that start and end with branded search. This is a proxy indicator of incrementality.
• Campaigns that appear in high-value customer journeys: A campaign that’s present in journeys that ended with a Subscribe and Save signup is more incrementally valuable than one that only appears in single-purchase journeys, even if the ROAS looks similar.

Amazon’s Long Term Sales (LTS) Metric: Built-In Incrementality Proxy

In October 2024, Amazon introduced the Long Term Sales (LTS) metric in the Ads Console — a built-in measurement that estimates the incremental sales value a brand can expect over the next 12 months, based on how effectively current ads are moving new-to-brand shoppers further down the purchase funnel.

LTS takes into account the downstream value of non-purchase engagements — detail page views, branded searches, add-to-carts — treating them as indicators of future purchase intent rather than failed conversions. A shopper who viewed your product page three times without buying is still moving toward purchase. LTS quantifies the likely value of that progression.

LTS ROAS vs Standard ROAS: LTS ROAS measures the 12-month estimated incremental revenue from an ad against ad spend. A campaign with a low standard ROAS but a high LTS ROAS is one that’s acquiring new customers who will buy again — exactly the kind of incremental growth that matters for brand building. Viewing LTS ROAS alongside standard ROAS gives you a more complete picture of which campaigns are truly growing your business.

Branded Search Incrementality: A Specific Approach

Branded search cannibalization is so common and so specific that it deserves its own testing framework. Here’s how to measure it precisely for Amazon Sponsored Products.

The geo-holdout approach for branded search

8. Select two comparable geographic markets. Choose two regions that are similar in population size, purchasing behavior, and historical sales for your brand — for example, two mid-sized U.S. metropolitan areas with similar Amazon purchasing patterns.
9. Run branded search in the control market, pause it in the test market. For 2-3 weeks, maintain all other advertising (DSP, non-branded search, Sponsored Brands) but remove branded keyword Sponsored Products in the test geography.
10. Monitor organic search volume and sales in both markets. Use Brand Analytics to track branded search volume by geography. Check whether organic sales fill the gap left by removing paid branded ads.
11. Calculate the organic recovery rate. If you remove $20,000 in branded search spend and organic sales in that market decline by only $3,000, your organic recovery rate is 85%. The incremental value of branded search in that market was only 15% of what was being attributed.
12. Make a budget decision based on the data. If recovery rate is above 70-80%, reallocate a significant portion of branded search budget to non-branded or DSP prospecting campaigns that are more genuinely incremental.

This approach doesn’t mean eliminating all branded search spend. Branded keywords provide some value as defensive coverage — preventing competitors from appearing in that slot when a customer searches your brand. But that defensive value is worth much less than what most brands currently spend on it, and AMC incrementality testing helps you find the right level.

Incrementality Methods on Amazon: Choosing the Right Approach

There are several ways to measure incrementality on Amazon, ranging from simple proxies to rigorous holdout studies. The right choice depends on your budget, data access, and how much precision you need for the decision at hand.

For most established Amazon brands, the most practical starting point is a combination of: AMC path-to-purchase analysis to identify which campaigns look potentially non-incremental, followed by a focused AMC holdout study or geo test on those specific campaigns to confirm. You don’t need to test every campaign simultaneously — start where the budget is largest and the cannibalization risk is highest.

What Does Good iROAS Actually Look Like?

One of the most common questions after running an incrementality test is ‘is this good?’ The answer depends on what you’re testing, but here are practical benchmarks.

For context: a beauty brand geo holdout study comparing Amazon, Ulta, and Sephora showed Amazon delivering 2.8x iROAS — categorized as strong incremental performance — which led the brand to shift 30% of Sephora’s budget to Amazon, resulting in 22% more incremental sales overall. The iROAS test directly informed a meaningful budget reallocation decision.

Turning Incrementality Data Into Budget Decisions

Incrementality measurement is only valuable if it changes what you do with your budget. Here’s how to translate the findings into concrete actions.

Campaigns that typically show low incrementality — and what to do

• Heavy branded search Sponsored Products: Reduce branded search budget. Test a 30-40% reduction. If branded sales stay strong (likely through organic), permanently reallocate that budget to non-branded or DSP.
• Retargeting audiences with very high frequency: Add frequency caps, shorten lookback windows on the lowest-intent retargeting segments, and reduce bids on audiences showing very high organic purchase rates in the control group.
• DSP campaigns targeting existing loyal customers: Add past purchaser exclusions to your prospecting campaigns. Move loyal customer outreach to a separate lower-CPM program with replenishment messaging.

Campaigns that typically show high incrementality — and what to do

• DSP in-market prospecting (non-branded category audiences): Scale these. They’re finding people who weren’t already planning to buy from you, which is genuine incremental growth.
• Brand+ awareness campaigns with CTV: High first-touch incrementality in path-to-purchase analysis. Scale with appropriate creative quality.
• Competitor ASIN targeting in DSP: Reaching people who were about to buy from a competitor is highly incremental. Expand these audiences with careful creative that clearly communicates your differentiation.
• Lookalike audiences based on high-LTV customers: If the lookalike test shows meaningful iROAS, this is the most scalable incremental acquisition channel available in DSP.

Building an Ongoing Incrementality Program

A single holdout test is useful. A regular cadence of incrementality testing is what separates brands that continually improve their media efficiency from those that drift toward cannibalization over time.

The goal over time is to build an internal benchmark of iROAS for each campaign type, so that when you launch a new format or consider scaling an existing one, you’re comparing against data rather than intuition.

Frequently Asked Questions

Final Thoughts

Incrementality measurement is one of those topics that sounds technical until you understand what’s actually at stake. If 25% of your ad spend is going toward campaigns that are taking credit for organic sales rather than creating new ones, that’s a significant slice of budget that could be moved to campaigns that genuinely grow your business.

AMC makes that analysis possible in a way that was genuinely unavailable to most brands just two years ago. Holdout studies, path-to-purchase analysis, LTS metrics, and the iROAS framework together give you a clear picture of where your ads are earning their spend and where they’re not.

The brands that build this discipline into their measurement practice — running regular incrementality tests alongside standard attribution, reviewing iROAS alongside ROAS, and making budget allocation decisions based on causal data rather than correlation — consistently outperform peers who optimize purely for attributed efficiency.

It’s not the easiest thing to get started. But it’s one of the highest-ROI investments a well-established Amazon brand can make in their measurement infrastructure.