Amazon Ads can tell a brand how much revenue was attributed to advertising, but that does not automatically show how much revenue advertising caused. A shopper may click an ad and buy, yet that same purchase may have happened without the ad.
That distance is whatโwe're trying to measure with Amazon incrementality. Instead of saying โwhich campaign got credit for the sale,โ incrementality says, โhowโmany more sales did advertising actually get me?โ
For advertisers with Sponsored Ads, Amazon DSP, Streaming TV, or external traffic, it matters because attributed ROAS only tells part of the story. That story can obscure where advertising is generating demand vs. whereโitโs just taking a piece of existing demand. This guide covers Amazon iROAS, the concept of counterfactuals, testing methods, AMC, and how you can turn incremental performance into better budget decisions.
Quick Answer: What Is Amazon Incrementality?
Incrementality on Amazon is the extra sales, orders, customers, or brand impact driven by advertising relative to what would have been expected if there were no advertising.
The comparison point is called the counterfactual: a credible estimate of performance in a no-ad scenario.
Amazon iROAS then compares incremental revenue with the advertising investment required to create it:
iROAS = Incremental Revenue รท Advertising Spend
Attributed ROAS answers, "How much revenueโreceived advertising credit?" Incrementality answers a more difficult question: "How much more revenue existed becauseโthe advertising ran?"
IAB's commerce-media measurement guidance similarly focuses on a believable counterfactual, controls for bias, and distinguishes actual signal from noise.
ROAS vs iROAS vs ACoS vs TACoS
Amazon advertisers use several metrics to judge campaign performance, but each answers a different business question.
ROAS and ACoS focus on ad-attributed sales. TACoS compares advertising spend with total Amazon sales, making it useful for comparing paid and organic performance. New-to-brand metrics help assess acquisition. Amazon's long-term sales metrics extend the view beyond immediate conversions by estimating the expected 12-month sales value associated with new-to-brand shopper engagement. Amazon describes LTS and LTS ROAS as ad-attributed measures, so they should not be treated as experimental proof of incrementality.
What each Amazon advertising metric actually answers
| Metric | Formula or definition | What it answers | Causal proof? | Best use |
|---|---|---|---|---|
| ROAS | Attributed ad sales รท ad spend | How much attributed revenue came from each ad dollar? | No | Campaign reporting |
| iROAS | Incremental revenue รท ad spend | How much additional revenue did advertising create? | Depends on test design | Budget decisions |
| ACoS | Ad spend รท attributed ad sales | What percentage of attributed revenue was spent on ads? | No | PPC efficiency |
| TACoS | Ad spend รท total sales | How does ad spend relate to total Amazon sales? | No | Account efficiency |
| NTB | New-to-brand orders or sales | Is advertising attracting customers new to the brand? | No | Acquisition analysis |
| LTS ROAS | Long-term sales estimate รท spend | What longer-term value may follow NTB engagement? | No | Long-term acquisition analysis |
| Brand Lift | Exposed vs control response difference | Did advertising change measured brand outcomes? | Stronger for measured study outcomes | Upper/mid-funnel measurement |
SalesDuo's guide to good ROAS on Amazon explains attributed return in more detail, while its Amazon advertising benchmarks can provide directional context for conventional campaign metrics. SalesDuo cautions that benchmarks are reference points, not universal performance rules.
The fundamental difference? Attribution is executional, based on outcomes; isโincrementality an attempt to measure the causal effect or caused lift?
Why Last Click ROAS Can Mislead Amazon Brands
A campaign may report excellent ROAS, but its incremental revenue contribution may be relatively small.
Consider branded search. A buyer who conducts a direct brand search is already mildly qualified. A Sponsored Products or Sponsored Brands ad can be credited with a conversion after the click, but the user could have also clicked on the organic result.
Brand-defense campaigns add a layer of complexity. Turning off branded ads can reduce paid cannibalization, while also allowing you to win ad space away from competitors. So the question thatโhas value isnโt simply whether branded PPC converts. Itโs whether fluctuations in branded spend affect total branded sales, organic sales, customer acquisition, or competitor stealage.
Retargeting can create the same problem. Someone who viewed a product detail page or added an item to a cart is already relatively close to purchase. A retargeting campaign may report strong ROAS even if some of those shoppers would have returned without another ad.
Overlapping campaigns compound the problem. Aโsingle shopper may run into Amazon DSP, Sponsored Brands, Sponsored Display, and Sponsored Products all in one shopping trip.
Repeat buyers can also exaggerate attributed ROAS, as repeat purchasers may convert after an ad touch even when existing brand preference drove the conversion (i.e., the ad didnโt actually cause the conversion). Attribution-window bias is a type of observational bias that attributes causality to ads, converting events that occur in the reporting window without showing whether the ad caused them.
Attribution, multi-touch attribution, and incrementality are different questions.
Amazon continues to expand multi-touch attribution capabilities across its measurement products to better distribute conversion credit across eligible ad touchpoints. Its Multi-Touch Attribution metrics are available alongside traditional last-touch metrics in products including the Ad Console, Amazon DSP reporting, downloadable sponsored ads reports, the Amazon Ads API, and Amazon Marketing Stream. Amazon says the model divides credit for Amazon purchase conversions across multiple Amazon Ads touchpoints according to their estimated contribution to the shopping decision.
That provides another way to understand the path to purchase, but three questions should remain separate:
- Last-touch attribution: Which eligible interaction receives conversion credit?
- Multi-touch attribution: How should conversion credit be distributed across multiple touchpoints?
- Incrementality: Would the conversion have occurred without the advertising?
Better attribution could lead to betterโreporting in campaigns. It does not eliminate the need for a believable counterfactualโwhen the business question is causal.
When Amazon Advertisers Need Incrementality Measurement
Not all advertising decisions need to be tested. Incrementality is most useful when traditional reporting can't answer a critical budgeting question.
Branded search and brand defense
A brand enjoying a strong ROAS for branded PPC may want to know whether those ads generate incremental sales or capture demand from people already planning to buy. Testing changes in branded advertising, while monitoring paid, organic, and overall sales, may provideโadditional insight beyond ROAS.
DSP and retargeting
Retargeting often targets customers with purchase intent. So brands leveraging Amazon DSP may need more proof that attributed conversions are incremental sales, especially when audience overlap or frequency is high.
SalesDuo's Amazon DSP agency offering specifically connects DSP planning with Sponsored Ads, retail readiness, pricing, inventory, and broader Amazon strategy rather than treating programmatic media in isolation.
Sponsored Brands and upper-funnel campaigns
Sponsored Brands and other upper-funnel activity can influence a journey before the final conversion touchpoint. Incrementality helps brands determine whether those campaigns create additional demand, rather than judging them solely on immediate attributed return.
Streaming TV and CTV
Last-touch sales reporting may provide an incomplete picture of awareness-focused campaigns. Amazon Brand Lift measures outcomes such as awareness, preference, favorability, intent, and ad recall through survey responses from ad-exposed and control groups.
Product launches and conquest campaigns
A new product can show poor immediate ROAS because advertising createsโinitial demand and acquires customers. Conquest campaigns raise the reverse question: do relatively pricey competitor-centered clicks actually go after truly new demand?
Amazon Adsโ 2026 EOS launch case study reported that 81% of Amazon sales driven by Big Game-branded search were incremental, while iROAS during the Big Game outperformed EOSโs historical branded-search iROAS by 197%. The case illustrates how upper-funnel investment can create on-Amazon demand that last-touch reporting may undervalue.
External traffic
Amazon Attribution helps advertisers measure how non-Amazon channels including search, social, display, video, email, and affiliate or influencer activity lead to activity on Amazon.
It is useful for channel attribution, but attributed conversions are not automatically incremental conversions.
When incrementality testing is not worth running yet
A formal test may be premature when conversion volume is too low to distinguish lift from normal variation. Results can also be hard to interpret when inventory, pricing, promotions, or major seasonal events change throughout the measurement period.
Testing has little practical value when the result will not change a budget or campaign decision. The objective is better decision-making, not more complicated reporting.
Brand maturity should also shape test priorities. Mature brands with strong organic demand should scrutinize branded search and retargeting first, because those tactics are more likely to capture existing intent; newer brands with a limited organic baseline may get more value from measuring prospecting and launch activity before aggressively reducing branded coverage.
How to Calculate Amazon iROAS
Amazon iROAS formula
Amazon iROAS, also called incremental ROAS on Amazon, is calculated as follows:
iROAS = Incremental Revenue รท Advertising Spend
Incremental revenue can be expressed as:
Observed revenue โ estimated counterfactual revenue = incremental revenue
The arithmetic is easy. Building a defensible counterfactual is the difficult part.
Worked Amazon iROAS example
Consider a hypothetical test:
| Item | Illustrative value |
|---|---|
| Observed revenue | $120,000 |
| Estimated no-ad revenue | $90,000 |
| Incremental revenue | $30,000 |
| Advertising spend | $10,000 |
| iROAS | 3.0ร |
First:
$120,000 โ $90,000 = $30,000 incremental revenue
Then:
$30,000 รท $10,000 = 3.0ร iROAS
These numbers are illustrative, not an Amazon benchmark.
Why calculating the counterfactual is the hard part
The same shopper, market, and time period cannot simultaneously be observed both with and without advertising. Measurement therefore requires a credible comparison.
That comparison could come from:
- a control audience;
- a matched geography;
- alternating treatment periods;
- a randomized experiment;
- a modeled baseline.
IAB distinguishes experimental approaches, model-based counterfactuals, econometric models, and hybrid proxies. Its guidance emphasizes that without a credible counterfactual, measured differences may represent correlation rather than campaign impact.
Amazon Incrementality Measurement Methods
The right method depends on the campaign, available data, business risk, and level of causal confidence required.
Geo holdout tests
Ads run in certain regions while comparable markets serve as controls. Treatment and controlโperformance differences can be used to estimate lift. Strong matching is important, as baseline differences between markets can skew results.
Audience holdouts
A similar group is deliberately not exposed to advertising and is then compared with the group that is exposed. Adequately designed randomized holdouts canโyield robust causal proof.
Switchback tests
Advertising is cycledโon and off in accordance with a predetermined schedule. Switchbacks are useful when there are no audience or geographic holdouts, but seasonality, day-of-week effects, promotions, and inventory changes can confound them.
Amazon Marketing Cloud analysis
Amazon Marketing Cloud is a secure, privacy-safe clean-room environment for custom analytics and audience building across pseudonymized Amazon Ads signals and advertiser inputs.
AMC can be used to examine journeys, frequency, overlap, and cohorts; however, descriptive AMC analysis on itsโown does not imply causality.
Amazon Brand Lift
Amazon Brand Lift uses exposed and unexposed control groups to measure changes in brand-related survey outcomes. Amazon reports absolute lift as the difference between qualifying response rates for the exposed and control groups.
Marketing Mix Modeling
MMM relies on historical volatility of marketing spend and business outcomes to calculate attribution shares between channels. IAB considers econometric modelsโsuch as MMM as part of the wider incrementality measurement toolkit.
Amazon Attribution
Amazon Attribution monitors non-Amazon marketing activity and its impact on on-Amazon shopping activity. Amazon now positions it as an analytics solution for measuring the impact of your non-Amazon campaigns in helping customers discover, consider, and purchase your products on Amazon.
NTB, TACoS, and other proxy metrics
New-to-brand performance (NTB), TACoS, overall sales, organic outcomes, branded-search patterns, and repurchasing may point in the right direction. They are good diagnostic signals, butโyou should not present them as causal evidence.
Amazon Incrementality Measurement Methods Compared
| Method | How it works | Best fit | Causal confidence | Data need | Main caution |
|---|---|---|---|---|---|
| Geo holdout | Treatment vs control markets | DSP, launches, external media | High when well designed | Moderate-high | Market comparability |
| Audience holdout | Exposed vs withheld audience | DSP, retargeting | High | Moderate-high | Control contamination |
| Switchback | Alternate treatment by period | Sponsored Ads tests | Moderate | Moderate | Time/seasonality effects |
| AMC | Analyze journeys and audiences | Advanced advertisers | Diagnostic alone | Moderate-high | Correlation โ causation |
| Brand Lift | Exposed/control survey | Upper-funnel media | High for measured outcome | Eligibility dependent | Not direct sales lift |
| MMM | Model historical contribution | Large multi-channel brands | Model dependent | High | Model/data quality |
| Attribution | Track external touchpoints | Search, social, creators | Low for causality | Low-moderate | Attribution โ incrementality |
| Proxies | TACoS, NTB, sales trends | Smaller advertisers | Low | Low | Directional only |
How to Measure Amazon Ads Incrementality: Design a Reliable Test
The best incrementality tests begin with a business decision, not a dashboard.
Step 1: Define the decision and hypothesis
Avoid vague questions like, โIs DSP effective?โ
Make it something actionable:
We will increase the prospecting DSP budget if the measured incremental return exceeds our profitability threshold.
The measurement now has a clear decision rule.
Step 2: Choose the treatment
Indicate what will be altered: the exposure, the budget, the audience, the geographical scope,โthe campaign status, or the advertising period.
Determine the test unit (e.g., audience, geography, campaign, ASIN set, time frame) and specify inclusion and exclusion criteria before launch. Also describe which exposure, spend, sales, and information on retail conditions will be gathered so treatment and control groups can be compared uniformly.
Avoid changing bids, prices, promotions, creatives, and targeting altogether, as this makes the experiment's results harder to interpret.
Step 3: Select a credible control or counterfactual
A control should represent what would reasonably have happened without the treatment.
Before choosing comparable audiences, periods, or regions, review historical sales, traffic, customer mix, category conditions, and promotional activity.
Step 4: Set the primary KPI
Choose the main outcome before viewing results.
Depending on the objective, that might be:
- incremental sales;
- incremental orders;
- incremental new customers;
- contribution profit;
- a specific Brand Lift outcome.
This reduces the temptation to search afterward for whichever metric looks strongest.
Step 5: Control for inventory, pricing and promotions
Advertising is only one factor influencing Amazon sales.
Document major changes in:
- inventory availability;
- Featured Offer status;
- pricing;
- coupons;
- deals;
- product availability;
- listing suppression;
- conversion rate.
IAB specifically identifies pricing, promotions, seasonality, competitor activity, and concurrent campaigns as factors that can bias causal measurement when they are not properly controlled.
Step 6: Predefine duration and stopping rules
Set theโdesired measurement window in advance of the test start. Don't quit justโbecause one early result looks good.
If the results are still noisy, or the test is interrupted by changes inโthe inventory or promotion, extending, redesigning, or rerunning the test may be more defensible than forcing a conclusion.
Before starting, set a minimum threshold for the amount of data needed to produce a result, and separate campaign types that are materially different. Do not lump branded defense, retargeting, and prospecting together when estimating lift, as each has a different baseline and incrementality curve. Ifโyour experiment does not achieve the required data amount, state the outcome as inconclusive.
Step 7: Measure lift and uncertainty
Calculate the gap between performance observed and the counterfactual,โbut also consider how stable that estimate is.
An estimated iROAS of 2.5ร is not particularly actionable if the uncertainty around the estimate is such that the actual result could be below the brand's profitability threshold.
Step 8: Decide what result will trigger action
Set possible actions in advance: scale, maintain, reduce, restructure, or retest.
That keeps Amazon advertising incrementality measurement tied to budget decisions.
How Amazon Marketing Cloud Supports Incrementality Analysis
Amazon Marketing Cloud can provide deeper visibility than standard campaign-level reports. SalesDuo's existing AMC guide covers its broader features and use cases, so this article stays focused on incrementality.
Audience overlap
AMC can help advertisers determine whether consumers are seeing them across their various marketing campaigns. This comes in handy when you want to know whether a few campaigns are targeting essentially the same audience.
Customer journey and path analysis
Teams donโt have to analyze each campaign as a single touchpoint but can look at a series of advertising engagements.
Frequency analysis
Frequency analysis can help identify how often audiences are exposed and whether performance patterns change as exposure frequency increases.
New-to-brand and cohort analysis
Advertisersโcan segment customer groups and compare behavior across journeys or cohorts, enabling them to form more informed acquisition and retention hypotheses.
What AMC cannot prove by itself?
A correlation discovered in AMCโis not automatically causal.
If shoppers exposed to multiple campaigns convert at a higher rate, part of that difference may reflect pre-existing purchase intent or audience selection rather than ad exposure. AMC provides the analytical environment; the counterfactual and research design determine how confidently you can interpret the result as incremental.
Brands that need hands-on clean-room expertise can also review SalesDuo's comparison of Amazon Marketing Cloud service providers.
How to Turn iROAS Into Amazon Budget Decisions
Incrementality becomes useful when it changes how the next advertising dollar is spent.
High ROAS + high iROAS
The campaign is receiving strong attributed returns and appears to be driving real incremental revenue.
Potential action: Scale cautiously and monitorโdiminishing returns, inventory, and profitability.
High ROAS + low iROAS
The campaign looks very efficient in attribution reporting but isn't generating much incremental revenue. Brandedโsearch and aggressive retargeting are the usual suspects.
Potential action: Test lower branded spend, separate brand and non-brand traffic, or reduce retargeting exposure.
Low ROAS + high iROAS
The campaign receives limited direct attribution but may be creating additional demand that last-touch reporting does not fully capture.
Potential action: Consider whether prospecting, Sponsored Brands, video, or DSP activity deserves more budget despite weak last-touch reporting.
Low ROAS + low iROAS
The campaign performs poorly under both attributed and incremental measurement.
Potential action: Reduce, rebuild, or move spend unless a separate strategic objective justifies the investment.
Compare iROAS with contribution margin
There is no universal "good" Amazon iROAS.
A brand still needs to consider:
- COGS;
- Amazon fees;
- discounts;
- promotional costs;
- contribution margin;
- inventory constraints;
- longer-term customer value.
Incremental revenue is not the same as incremental profit.
ROAS vs iROAS Budget Decision Matrix.
| ROAS | iROAS | Interpretation | Potential action |
|---|---|---|---|
| High | High | Efficient and incremental | Scale cautiously |
| High | Low | May be harvesting demand | Reduce or test |
| Low | High | Attribution may undervalue impact | Evaluate profitable scale |
| Low | Low | Weak on both measures | Reduce or redesign |
| Unclear | Unclear | Evidence insufficient | Improve test design |
Translate the diagnosis into a specific lever: scale or cap spend, suppress weak audiences or tactics, separate brand from non-brand, adjust DSP exposure or frequency, test creative, or shift budget to a stronger incremental opportunity.
If the diagnosis reveals spend that is producing weak business value, the next step is identifying wasted ad spend on Amazon and correcting campaign-level inefficiencies. SalesDuo's PPC service specifically reviews bids, budgets, placements, search terms, ACoS, TACoS, and ROAS as part of account diagnosis.
Common Amazon Incrementality Measurement Mistakes
| Mistake | Why it creates a problem |
|---|---|
| Calling attribution incrementality | Credited sales could still have occurred without advertising. |
| Treating AMC correlation as proof | Observed relationships are not automatically causal. |
| Using a weak control | Existing differences can be mistaken for advertising lift. |
| Testing with too little data | Normal variation can overwhelm the effect being measured. |
| Allowing control contamination | Control audiences exposed elsewhere weaken the counterfactual. |
| Stopping tests early | Short-term fluctuations can create misleading conclusions. |
| Ignoring inventory | Availability changes can alter sales independently of advertising. |
| Ignoring pricing or promotions | Commercial changes may explain measured lift. |
| Treating NTB as incrementality | A new-to-brand order indicates customer status, not causality. |
| Using generic iROAS benchmarks | Profitable thresholds differ by economics and objective. |
To protect against these failures, define all treatment/control eligibility criteria, the minimum amount of data required, the length of the test, and the conditions under which you will stop running the test before launching. Log inventory, Featured Offer, pricing, promotions, seasonality, competitor activity, andโsimultaneous campaign changes, and rerun or qualify results when these elements cause results to become non-comparable. TreatโAMC, NTB, TACoS, and attributed ROAS as diagnostic signals unless the measurement design provides a meaningful counterfactual.
The IAB guidelines reiterate the importance of these details: the credibilityโof incrementality relies on the counterfactual, the mitigation of bias, and the isolation of the marketing effect from other factors.
Need Help Proving What Your Amazon Ads Actually Drive?
Amazon incrementality helps brands get a clearer signal on whether their spend is driving profitable growth or just getting credit for demand that would have purchased anyway. High-quality Amazon advertising measurement is more than just ROASโreporting. It ties campaign performance, customer journeys, new-to-brand purchase behavior, retail conditions, profitability, and credible testing to actual budgetโdecisions.
SalesDuo's Amazon advertising agency service connects advertising execution with full-funnel measurement and marketplace performance.
Brands focused mainly on Sponsored Ads can explore SalesDuo's Amazon PPC agency, while brands using programmatic display, video, or Streaming TV can review its Amazon DSP agency support.
The objective is not simply to identify what advertising received credit for. Book a 1:1 growth call with SalesDuo to determine which investments are creating enough profitable incremental growth to deserve the next dollar.
Frequently Asked Questions on Amazon Incrementality
1. What is Amazon incrementality?
Amazon incrementality is the additional sales, orders, customers, or brand impact caused by advertising compared with what would likely have happened without the advertising. A credible incrementality analysis therefore requires some form of counterfactual rather than relying only on attributed conversions.
2. What is iROAS on Amazon?
Amazon iROAS, or incremental return on ad spend, compares incremental revenue with advertising spend. The formula is incremental revenue รท ad spend. Its usefulness depends on how reliably the advertiser estimates the revenue that would have occurred without the advertising.
3. How is iROAS different from ROAS?
ROAS attributes sales to advertising. iROAS estimates return using only the revenue judged to be incremental to advertising. A campaign may therefore demonstrate a strong ROAS and a weaker iROAS when a substantial portion of its associated revenue comes from purchasers who would likely have bought anyway.
4. Does Amazon Ads ROAS prove incremental sales?
No. ROAS is the ratio of attributed ad sales to advertising spend. Attribution is the process of deciding which advertising interaction gets credit under a prescribed attribution methodologyโit doesnโt define what the absent advertising situation would be. Incrementality necessitatesโa valid counterfactual or similar causal technique.
5. How do you measure Amazon ads incrementality?
Define the business decision first, establish a test treatment and credible control or counterfactual, choose the KPI in advance, control major factors such as price and inventory, run the test for an appropriate period, estimate incremental lift, calculate iROAS, and use the result to make a predefined budget decision.
6. Can Amazon Marketing Cloud measure incrementality?
AMC enables an advertiser to conduct custom analysis across pseudonymized Amazon Ads signals and advertiser inputs to enable incrementality analyses. It allowsโthe examination of journeys, audiences, overlap, cohorts, and frequency. However, evaluating in AMC does not always imply causality; how confidently lift can be consideredโincremental depends on the measurement design.
7. Is TACoS a good proxy for incrementality?
TACoS can be a useful directional metric because it compares advertising spend with total Amazon sales rather than only ad-attributed sales. However, a change in TACoS does not prove advertising caused a change in total or organic sales. It should support diagnosis rather than replace causal testing.
8. What is a good iROAS for Amazon advertising?
There is no universal Amazon iROAS benchmark advertisers should use. The target return depends on contribution margin, cost of goods sold (COGS), Amazon fees, promotions, inventory, customer value, campaign objective, and measurement accuracy. The relevant cutoff is the incrementโat which additional advertising yields a worthwhile return.
About the Author
Meet Srushti P. Borle, an SEO Content Writer Intern at SalesDuo who enjoys turning complex ideas into clear, easy-to-read content. She focuses on creating smooth, engaging content that readers can understand and enjoy. Outside of work, she loves reading, exploring new ideas, and finding simple ways to present detailed topics.
