Proving Incrementality: Are Your Amazon Ads Creating New Sales or Capturing Existing Demand?

Last Updated: June 15, 2026

By Justin W. Boggs, Chief Revenue Officer

What Incrementality Actually Means on Amazon — and Why Most Brands Are Measuring the Wrong Thing

Most brands think they're measuring ad performance.

They're not. They're measuring confirmation bias — a dashboard that validates every dollar spent because the attribution model counts every sale that touched an ad, regardless of whether the ad caused it.

Here's what that looks like in practice. A buyer types your brand name into the search bar. They were already coming. They click a Sponsored Product on the way to the cart. Your dashboard logs it as an ad-driven sale. But you didn't create that demand — you just paid a toll to access it. That's not advertising performance. That's your own existing demand billed back to your ad budget.

That's where ACoS and TACoS tell completely different stories. ACoS measures what percentage of ad-attributed revenue went back to ads. It looks clean. It says nothing about whether any of that revenue was new. TACoS measures ad spend against total channel revenue — and that ratio asks the honest question: is this channel actually growing, or is it recirculating demand you already had? The answer lives in disciplined TACoS optimization.

The Difference Between Captured Demand and Created Demand

There are exactly two types of sales on Amazon.

Sales you created — buyers who didn't know your product existed, found it through advertising, and converted. And sales you captured — buyers who already wanted your product, were already heading to your listing, and clicked an ad on the way there.

Captured demand isn't worthless. But it isn't advertising.

It's an expensive tollbooth on a highway you already own. The buyer was going to make the trip. You just charged yourself for the toll.

Created demand is the only category that justifies advertising investment at a P&L level. It's the shopper who searched a generic category term, found your listing, and converted for the first time. That's a new customer. That's incremental revenue.

And it's the only metric that compounds — because it builds organic velocity, reviews, and rank signals that don't disappear when you pause the campaign. Sustainable Amazon advertising management requires campaigns built specifically to isolate this kind of high-efficiency, non-branded traffic.

The Problem With Branded Keyword Bidding

Branded keyword bidding is the most common source of non-incremental spend on Amazon — and the hardest to argue against internally, because the numbers always look clean.

High click-through rates. Strong conversion rates. Low ACoS. The dashboard is happy. The P&L is not.

This isn't a theory. Research published by the National Bureau of Economic Research found that brand-keyword advertising generated virtually no incremental sales — with up to 99% of paid clicks simply cannibalizing organic traffic that would have converted anyway. Not a rounding error. Not an edge case. When a brand bids on its own name and the attribution model counts every resulting sale as a paid win, the report looks great and the P&L quietly bleeds.

The fix isn't pulling branded keyword coverage entirely. There are real defensive reasons to hold that ground. But stop counting those conversions as proof that advertising is working. Segment branded and non-branded campaigns. Measure them separately. Then run an isolation test — pause branded spend in a controlled window and watch whether organic sales hold or collapse. If organic holds, you've been paying for traffic you already earned. That's the tollbooth. And that's the number that belongs in the incrementality conversation.

Ad Type	Demand Source	Incremental Revenue Created	Risk of Cannibalization	P&L Impact
Branded keyword ad	Existing demand — buyer already searching your brand name	None to minimal	High — most clicks would have converted organically	Negative: ad spend converts organic sales into paid sales, inflating ACoS while eroding margin
Non-branded / category keyword ad	New demand — buyer searching a generic product category	High — buyer discovers brand for the first time	Low — organic rank unlikely to have captured this buyer	Positive: true customer acquisition that builds organic velocity, reviews, and rank signals
Competitor conquest ad	Contested demand — buyer actively comparing alternatives	Moderate — some buyers are genuinely persuadable	Low to moderate — depends on listing strength and price position	Variable: can generate incremental revenue but requires strong conversion rate to justify the cost-per-click
Retargeting / remarketing ad	Recaptured demand — buyer previously viewed listing but did not convert	Moderate — closes sales that may have been lost	Moderate — some buyers would have returned organically	Mixed: incrementality depends on how long the conversion window is and whether organic re-discovery was likely
Defensive branded ad (owned brand, protected position)	Existing demand — same buyer base as organic branded search	None — demand already belonged to the brand	High — spend is purely defensive, not generative	Neutral to negative on P&L: justified only to block competitor conquest, not to drive growth metrics

Why ACoS-Only Optimization Is the Wrong Lens

ACoS is the first slide in every agency deck. It's clean, it's simple, and it makes the program look like it's working.

That's exactly why it's dangerous.

ACoS measures one thing: ad spend divided by ad-attributed revenue. That's the whole equation.

It doesn't tell you whether that revenue was new. It doesn't tell you whether the buyer was already heading to your listing. And it tells you nothing about what's happening to organic performance while the campaigns run.

Retail media networks are projected to reach $140 billion in global ad spend. Brands pouring that kind of capital into Amazon are finally pulling up their P&Ls — and what they're finding is that a low ACoS doesn't mean the channel is healthy.

It means the attribution model is working. That's not the same thing. The brands now demanding proof of incrementality are the ones who stopped accepting a dashboard metric as a substitute for a real answer.

What ACoS Actually Measures — and What It Misses

Here's the mechanics. A buyer searches your brand name. They click a Sponsored Product ad. They purchase. That sale gets attributed to advertising.

ACoS looks great. But that buyer already knew you. They were already on the highway. The ad didn't create anything — it just collected the toll.

But the attribution gap goes deeper than one branded click. ACoS can't see organic sales. It can't see repeat purchases built on brand equity. It can't see the shopper who saw an ad, didn't click, and came back three days later through a direct search.

None of that shows up. So ACoS stays low, the report looks clean, and the channel quietly falls apart around it. The metric isn't lying — it's just only counting what it's allowed to count.

That's the gap between campaign management and channel management. When TACoS falls while total revenue holds or grows, something real is happening — organic velocity is building, converting without paid support, compounding over time.

That signal is invisible to ACoS. Because ACoS doesn't measure the channel. It only measures what the attribution model decides to claim. The operational proof lives in organic velocity — and that's a number ACoS will never show you.

How Agencies Use ACoS to Report Good Numbers on a Failing Channel

Here's how it works in practice. An agency bids aggressively on branded keywords. Branded searches convert well — they always do. ACoS drops. The monthly report shows efficiency gains.

The agency presents this as proof the program is performing. No one asks whether any of those sales were incremental. No one has to — the slide looks great.

As retail media ad spend scales toward $100 billion, brand manufacturers are starting to demand something most agencies have never had to provide: transparency on incremental conversion rates — not just campaign efficiency scores.

That demand is coming from one place. Brands are putting their advertising reports next to their P&Ls. And the numbers don't match. The retail media comes of age moment isn't a trend. It's an accountability reckoning — and agencies that have been running on dashboard optics are about to feel it.

The comfortable-lie of ACoS optimization is that it's never technically wrong. The numbers are real. The attribution is traceable. Everything checks out on the dashboard.

But a low ACoS on a stagnant channel isn't success — it's an expensive tollbooth running efficiently on a highway no one new is driving. Proving incrementality means tearing that tollbooth down. It means asking what's actually happening to total channel revenue. That's the question ACoS was never built to answer.

Metric	What It Measures	What It Ignores	Who Benefits from Reporting It
ACoS (Advertising Cost of Sale)	Ad spend as a percentage of ad-attributed revenue only	Organic sales, repeat purchases, brand-equity conversions, and total channel health	Agencies reporting campaign-level efficiency to justify their retainer
TACoS (Total Advertising Cost of Sale)	Ad spend as a percentage of total channel revenue — paid and organic combined	Nothing — it captures the full picture of channel economics	Brands and operators accountable for P&L outcomes, not just campaign dashboards
Branded keyword conversion rate	How efficiently branded search terms convert within the paid attribution window	Whether those buyers were already going to purchase without the ad	Agencies that need strong conversion metrics to validate aggressive branded bidding
Click-through rate (CTR)	The percentage of impressions that result in a click on a sponsored placement	Whether clicks represent new demand or existing demand redirected through a paid touchpoint	Agencies optimizing for visible engagement signals rather than incremental revenue
ROAS (Return on Ad Spend)	Revenue generated per dollar of ad spend within the attribution window	Organic sales displacement, non-incremental captured demand, and contribution margin after fees and COGS	Brands and agencies focused on topline revenue optics rather than bottom-line channel profitability

How TACoS Reveals What ACoS Hides

TACoS and ACoS aren't measuring the same thing differently. They're measuring completely different questions.

ACoS asks: how efficiently did ad spend convert to ad-attributed revenue? TACoS asks: what percentage of the entire channel required advertising to exist at all?

Those are not the same question. And only one of them matters at the P&L level.

Most agencies never ask the second question. ACoS-only reporting doesn't require them to.

A brand can run a perfectly clean ACoS for twelve straight months while organic sales quietly decay, channel dependency on paid spend quietly deepens, and contribution margin quietly erodes.

The report never shows it. The dashboard stays green. And the agency keeps collecting credit for a channel that's slowly hollowing out.

U.S. Census Bureau data on e-commerce as a percentage of total retail consistently shows over 15% of total U.S. retail transacting online — and that share keeps expanding every quarter.

More brands are pouring capital into Amazon. More of them are competing for the same positions with the same ad types.

Brands that can't separate created demand from captured demand are paying a toll on every sale. They're just not seeing the invoice.

What a Healthy TACoS Trend Actually Looks Like

A healthy TACoS trend doesn't start low. It moves lower over time as total revenue grows.

That's the pattern. Advertising seeds new demand. That demand compounds into organic rank, reviews, and repeat purchases that no longer need ad spend behind them.

When TACoS is falling while the channel is growing, organic velocity is building. That's the direction that matters — not a snapshot number frozen in a monthly report.

There's one context where a high TACoS is the right outcome: a new product launch.

When organic rank is zero and reviews are thin, advertising carries the full weight of discovery. Every sale it drives is genuinely incremental — the channel has no existing demand to capture, so the spend is building something from scratch.

That's a completely different situation from an established product bleeding margin through non-incremental branded bidding. Amazon new product launch strategy justifies that elevated spend. An entrenched ASIN with years of organic history does not.

But that same elevated TACoS on a mature product with established organic rank is a completely different signal.

It means advertising has stopped creating demand and started replacing it. The channel is compensating for organic decay — and it's getting more expensive to hold position every month.

The trend you want on any established ASIN is a TACoS that falls as total revenue holds or grows. That movement proves the investment built something that doesn't need continuous ad spend to survive.

Reading TACoS as a Signal of Organic Sales Velocity

TACoS is a velocity signal, not just an efficiency metric.

When TACoS falls while total channel revenue grows, organic sales are accelerating. When TACoS rises while revenue stays flat, advertising is compensating for organic decay.

And the P&L is absorbing that cost invisibly — behind a clean-looking ACoS number that tells you nothing about it.

Retail media ad spend is projected to reach $140 billion globally — making it the fastest-growing ad channel in the world.

As that capital floods in, brand manufacturers are being pushed to demand proof of incrementality — not just efficiency scores on individual campaigns. The gap between those two things is exactly where margin disappears.

TACoS is what closes that gap. It's the only metric that forces the question the channel actually needs answered.

When you read TACoS as a velocity signal, you're not managing a campaign anymore.

You're managing a channel.

A falling TACoS on growing revenue means the tollbooth is coming down — organic sales are compounding, rank signals are strengthening, and advertising is doing its actual job. That's not just a better metric. It's a fundamentally different accountability model for what advertising is actually for.

TACoS Range	What It Signals	Organic Velocity Direction	Recommended Action
Very high TACoS, low organic rank	Advertising is carrying full discovery weight — expected during launch phase	Flat or not yet established	Maintain aggressive ad investment; monitor for early rank signals and review accumulation
Declining TACoS, growing total revenue	Organic velocity is building — advertising seeded demand that now converts without paid support	Accelerating	Sustain disciplined ad coverage; protect rank gains with listing integrity and competitive pricing
Stable low TACoS, stable total revenue	Channel has reached an efficient equilibrium — organic rank is holding, spend is proportional	Holding steady	Shift focus to channel defense — MAP enforcement, unauthorized seller removal, review quality
Rising TACoS, flat or declining total revenue	Advertising is compensating for organic decay — paid spend is replacing demand, not creating it	Decelerating	Diagnose the root cause — listing quality, competitor pressure, pricing erosion, or suppressed rank — before increasing spend
Low ACoS but rising TACoS	Campaign-level efficiency is masking channel-level deterioration — the metric gap in action	Declining beneath the surface	Reframe reporting around total channel revenue; audit organic performance independently of ad attribution

How to Test and Prove Incrementality on Your Amazon Channel

Knowing ACoS lies isn't enough. The real question is harder: how do you prove, with actual data, whether your ad spend is creating new demand — or just collecting a toll on demand that already existed?

It's not a single test. It's a discipline — structured methods that isolate variables, create clean measurement windows, and force the attribution model to tell you what it would rather hide. Causal measurement research from Amazon Science confirms that campaigns built around causal isolation — not last-touch attribution — produce materially different performance reads. The same discipline that builds keyword architecture builds the testing framework that proves whether that architecture is actually doing anything new.

Every test here measures the channel, not the campaign. Each one asks the tollbooth question directly: if the ad disappeared, would the sale still happen? If yes — you're paying for a road you already own. If no — the advertising is doing real work.

Geo-Split and Time-Based Testing

Geo-split testing divides your market into comparable geographic regions. Advertising runs in some. Others go dark. The control regions show you what organic demand looks like without paid support. The test regions show you what incremental lift the advertising actually produces — cleanly, without the attribution model burying the answer.

Time-based testing works the same way with a different lever. Pull advertising spend on a specific ASIN for a defined window. Watch what happens to revenue. If organic sales hold after you remove paid support, the advertising wasn't driving new demand. It was intercepting demand that already existed — and charging your brand a toll to reach it.

Both methods only work if branded and non-branded campaigns are cleanly separated. The FTC's native advertising guide for businesses requires that sponsored results be clearly distinguished from organic placements — which means the platform already creates the structural separation you need. Use it. Branded spend in one bucket. Non-branded in another. Blend them and the attribution becomes impossible to read.

Using Campaign Pause Tests to Isolate Non-Incremental Spend

Campaign pause tests are the most direct method. The approach is simple: pause branded Sponsored Products for a controlled window. Measure what happens to total channel revenue, organic rank, and TACoS.

If total revenue drops sharply when branded campaigns pause, that's worth understanding. But if revenue holds — if organic sales stabilize at roughly the same level they sat before the ads ran — you've just found a tollbooth. You were paying for access to a highway you already owned. That's the data point a low ACoS can obscure. The numbers looked healthy. The margin was leaving anyway.

Pause tests also expose keyword-level dependencies. Some non-branded keywords will show real incremental lift — organic rank isn't strong enough yet to capture that traffic without paid support. Others will show no measurable difference when the ads stop. Every dollar on those terms was non-incremental. Pulling that spend and reallocating it toward keywords that actually drive new demand is how campaign structure earns the channel its margin back.

Who This Approach Is Not For

This approach isn't for brands that need their advertising reports to look clean on a slide. Incrementality testing surfaces uncomfortable data. It shows that a meaningful share of current ad spend is non-incremental, that branded campaigns are capturing demand instead of creating it, and that the ACoS the previous agency was optimizing was never measuring what it appeared to measure.

And it's not for brands whose leadership wants to call every campaign-level shot. Running a controlled pause test means letting the data reach a conclusion — even when that conclusion means pausing spend that feels safe. Brands that require sign-off at every tactical step, or that treat the agency as a button-pusher rather than an accountable channel partner, will find reasons not to run the test. Without the test, the tollbooth stays standing.

And it's not for brands shopping for short-term numbers to justify an existing budget. Proving incrementality is a long-game discipline. It requires patience through controlled windows, willingness to sit with data that complicates the story, and a commitment to managing the channel for P&L outcomes — not for the monthly report everyone in the room already wants to believe.

Test Method	How It Works	What It Isolates	Minimum Duration	Risk Level
Geo-Split Testing	Divide your market into comparable geographic regions — run ads in some, go dark in others — and measure the revenue difference between test and control regions	Whether organic demand exists independently of paid support in a given market	Long enough to capture a full purchase cycle for the ASIN's typical buyer	Moderate — requires comparable region selection and clean segmentation to avoid confounding variables
Time-Based Pause Testing	Remove paid support on a specific ASIN for a defined window and track whether organic sales hold, decline, or accelerate without advertising	Whether advertising is creating new demand or intercepting demand that already existed and converting it to a paid click	Long enough for organic rank signals to stabilize after the removal of paid support	Low to moderate — results are clean if campaigns are properly isolated before the pause begins
Campaign Pause Testing (Branded)	Pause branded Sponsored Product campaigns specifically, then monitor total channel revenue, organic rank, and TACoS across the pause window	Non-incremental branded spend that captures existing demand rather than creating new purchase intent	Sufficient window to separate short-term fluctuation from structural organic demand patterns	Low — branded campaigns are the most likely source of non-incremental spend, making this the highest-value first test
Branded vs. Non-Branded Segmentation	Separate all campaigns into two clean buckets — branded keywords in one, non-branded in another — and measure TACoS and organic velocity independently for each	Whether branded spend is propping up existing loyalty versus whether non-branded spend is reaching genuinely new buyers	Ongoing — this is a structural campaign discipline, not a one-time test	Low — the separation itself creates the clean measurement environment required for every other testing method
Keyword-Level Incrementality Audit	Pull spend data by keyword, pause low-performing non-branded terms, and measure whether organic rank and total channel revenue change after removal	Which specific keywords are generating genuine incremental traffic versus which are collecting tolls on demand that organic rank already captures	One to two purchase cycles per keyword cluster to avoid seasonality distortion	Moderate — requires disciplined holdout methodology to prevent misreading short-term rank fluctuation as a structural signal

Frequently Asked Questions About Amazon Ad Incrementality

Most brands get the argument fast. The part that stalls is the execution.

The questions below cut to where that happens — the definitions, the tests, and the metrics that still feel slippery after years of watching the wrong number.

These aren't entry-level questions. They're what comes up after a brand has already decided ACoS isn't telling the whole story — and now needs to build the framework that does.

What is incrementality in Amazon advertising?

Incrementality measures whether an ad actually caused a sale. One question: would this buyer have purchased without seeing the ad?

If yes, the ad didn't create demand. It collected a toll on demand that already existed.

The research makes the problem concrete. Brand-keyword campaigns can cannibalize up to 99% of organic traffic — meaning nearly every paid click would've converted anyway. The campaign reports a win. The channel didn't earn one.

Incrementality closes the gap between what the dashboard claims and what the P&L actually reflects.

How do I know if my Amazon Sponsored Products ads are generating new sales or just capturing existing demand?

You test it. There's no proxy metric that answers this — you need a controlled measurement window.

Run a geo-split or a time-based pause test on your Sponsored Products campaigns. If total channel revenue holds when paid support drops, those ads weren't creating demand. They were intercepting it.

Branded campaigns are the most likely offender. As retail media ad spend climbs toward $140 billion globally, the pressure to prove incrementality — not just report efficiency — is only intensifying.

The test isn't optional anymore. It's how you find out what you're actually paying for.

Why is TACoS a better metric than ACoS for measuring advertising incrementality?

ACoS measures one thing: how efficiently ad spend drove ad-attributed revenue. That's it.

It says nothing about organic sales. Nothing about whether total channel revenue is growing. Nothing about whether the advertising built anything durable.

TACoS — Total Advertising Cost of Sale measured against total channel revenue — shows the full picture. When TACoS falls while total revenue grows, organic velocity is accelerating. That's the signal that advertising created something that doesn't need continuous spend to survive.

A low ACoS on flat revenue doesn't mean the channel is healthy. It means the attribution model is doing its job. Not the advertising.

What happens to my organic sales velocity if I pause non-incremental Amazon ad spend?

It depends on which ads you pause — and that's exactly the point.

Pause genuinely incremental spend — campaigns driving non-branded traffic where organic rank can't carry the load yet — and organic sales will dip. That's the signal the advertising was doing real work.

Pause non-incremental spend — branded campaigns capturing demand that was already converting — and organic sales will hold. That's the tollbooth test. If revenue doesn't move when the ad disappears, you were paying for a road you already owned.

Disciplined campaign structure separates both categories. Then it puts the budget where the work is actually incremental.

How does branded keyword bidding create non-incremental ad spend on Amazon?

Branded keyword campaigns target shoppers who are already searching for your brand by name. They weren't lost. They were already heading to your listing.

When you run Sponsored Products against your own brand terms, you're paying to intercept buyers who would've clicked the organic result for free. The ad collects the toll. The buyer makes the same trip.

The research confirms it directly. Brand-keyword campaigns can produce virtually no incremental sales — with up to 99% of those paid clicks cannibalizing organic traffic anyway. The campaign attributes the sale to advertising. The P&L absorbs a cost that never needed to exist.

What are the risks of ignoring incrementality when managing Amazon ad budgets?

The risks are structural. And they compound.

Non-incremental spend consumes budget that could be building genuine demand. It inflates ACoS benchmarks, which then justify higher bids on the same non-incremental terms. And it keeps branded cannibalization invisible — the channel looks healthy in campaign reports while contribution margin quietly erodes behind it.

Retail media is already the fastest-growing ad channel globally. Competition for placement rises every quarter. Brands that can't prove their spend is incremental are paying more each month for the same non-incremental traffic — and falling further behind the brands that can.

The Tollbooth Is Optional

The tollbooth isn't a structural feature of Amazon advertising. It's a default setting nobody turned off.

ACoS-only reporting keeps it standing. It makes branded cannibalization invisible. It makes non-incremental spend look productive. And it lets the channel quietly erode inside numbers that appear healthy.

The longer that goes unexamined, the more margin a brand surrenders to a cost it never had to pay.

Tearing it down requires exactly what most agencies are unwilling to build: a measurement framework that asks whether the sale would have happened anyway.

That's the question TACoS optimization forces into the room. A declining TACoS on growing revenue isn't just a metric win. It's proof that the channel is compounding — that organic velocity is real, that advertising dollars are creating demand instead of collecting a toll on demand that already existed.

That's what the P&L is supposed to reflect. Most agencies never show you that number because they're not accountable for it.

Marketplace Valet was built on one conviction: Amazon is a channel to be managed with discipline — not a lever to pull until the numbers stop making sense.

The brands that win long-term aren't the ones with the biggest ad budgets. They're the ones that prove, with actual data, that their advertising is doing real work — creating demand instead of collecting a toll on demand that already existed.

The tollbooth is optional. It's only still standing because no one ran the test that would expose it. The only question left is whether you're willing to.