Your Google Ads dashboard says:

ROAS: 5.2x

Your Meta dashboard says:

ROAS: 4.4x

Your analytics platform reports thousands of purchases.

Business looks good.

But there is a question those numbers don't necessarily answer:

How many of those sales would have happened even if you had not shown the ads?

That question sits at the center of marketing incrementality.

And as advertising becomes increasingly automated by AI, it may become one of the most important questions businesses ask about marketing performance.

Google has been expanding its measurement ecosystem around first-party data, marketing mix modeling and causal experimentation, including Meridian and Meridian GeoX.

For business owners, however, the underlying issue is much simpler.

Traditional attribution asks:

Which marketing touchpoint gets credit for this sale?

Incrementality asks:

Would this sale have happened without the marketing?

Those are very different questions.

And the difference can change where you spend your next advertising dollar.

---

What Is Marketing Incrementality?

Marketing incrementality measures the additional business outcome caused by a marketing activity compared with what would have happened without it.

Suppose your advertising campaign generated 1,000 reported purchases.

That does not automatically mean advertising created 1,000 purchases.

Imagine that without the campaign, the business would have generated approximately 700 purchases anyway.

Then the estimated incremental effect is:

1,000 observed purchases − 700 expected purchases without advertising = 300 incremental purchases

Those 300 purchases represent the outcome the campaign appears to have created above the expected baseline.

The difficult part is obvious.

You cannot observe the exact same customer or market simultaneously in two realities:

Reality A: Advertising happened.

Reality B: Advertising did not happen.

Incrementality testing attempts to estimate that missing second reality—the counterfactual.

In practical terms:

What would probably have happened if we had not made this marketing investment?

---

Attribution and Incrementality Are Not the Same Thing

This distinction deserves special attention.

Attribution asks:

Who gets credit?

For example:

A customer sees a Meta ad.

Later searches the brand on Google.

Clicks a Google Search ad.

Purchases.

Depending on the attribution system, Google, Meta or another touchpoint may receive some or all of the conversion credit.

Incrementality asks:

Did advertising cause an additional purchase?

Perhaps that customer already knew the brand and would have purchased anyway.

If so, advertising may receive attribution while creating relatively little incremental demand.

This doesn't make attribution useless.

Attribution remains extremely valuable for:

Campaign optimization.

Daily budget management.

Creative analysis.

Keyword analysis.

Audience analysis.

Funnel analysis.

But attribution and incrementality answer different questions.

A useful distinction is:

Attribution = Where did the conversion get credited?

Incrementality = Did marketing create additional conversions?

Marketing Mix Modeling = How did channels contribute to business outcomes over time?

Businesses increasingly need these methods to complement rather than replace one another.

---

Why ROAS Alone Can Be Misleading

Suppose an ecommerce business spends:

$20,000 on advertising

and the advertising platform reports:

$100,000 revenue

Reported ROAS:

$100,000 ÷ $20,000 = 5.0x

That sounds excellent.

But suppose an incrementality experiment estimates that only $40,000 of that revenue was genuinely incremental.

Incremental ROAS becomes:

$40,000 ÷ $20,000 = 2.0x iROAS

The campaign didn't suddenly become “bad.”

Instead, management learned something different:

A portion of the revenue credited to advertising might have happened anyway.

That information can substantially affect decisions about:

Budget expansion.

Brand campaigns.

Retargeting.

Promotions.

Channel mix.

Customer acquisition.

This is why businesses should be careful about equating:

Attributed revenue

with:

Revenue caused by advertising

---

The Three-Layer Marketing Measurement Model

AI Growth Wire recommends thinking about modern marketing measurement through three layers.

Layer 1 — Attribution

What received credit?

Useful for tactical campaign management.

Layer 2 — Incrementality

What additional outcome did marketing create?

Useful for evaluating causal impact.

Layer 3 — Business Economics

Was the incremental outcome profitable?

Useful for investment decisions.

This final layer is essential.

Even if advertising creates incremental sales, those sales still need to make economic sense after considering:

Cost of goods.

Discounts.

Shipping.

Payment fees.

Returns.

Agency costs.

Platform costs.

Customer lifetime value.

A campaign can therefore have:

Positive incrementality

but still produce:

Poor economics.

---

What Is Google Meridian GeoX?

Google Meridian is an open-source marketing mix modeling framework designed to help advertisers measure the relationship between marketing investment and business outcomes.

Meridian GeoX extends causal measurement into geographic experiments.

Instead of changing advertising everywhere, a business can use different geographic markets as treatment and control groups.

The objective is to estimate:

What would sales in the treatment markets have looked like if the advertising intervention had never occurred?

The difference between that estimated baseline and actual performance helps estimate the incremental effect of advertising.

This is particularly useful for businesses trying to answer questions such as:

“Does increasing paid social spend actually create more sales?”

“What happens when we reduce advertising in selected markets?”

“Does connected TV create incremental revenue?”

“Should we increase investment in a particular advertising channel?”

---

How Does a Geo Incrementality Experiment Work?

Suppose a business sells nationally across the United States.

Instead of increasing advertising everywhere, it divides comparable geographic markets into groups.

Treatment Markets

Advertising investment is changed.

Control Markets

Advertising continues under normal conditions.

The business then compares outcomes.

The objective isn't simply to determine whether treatment-market sales increased.

It is to estimate whether those markets performed better than they probably would have without the advertising change.

That distinction separates a causal experiment from a simple before-and-after comparison.

---

Three Types of Geo Experiments Businesses Can Consider

Heavy-Up Test

Increase advertising investment in selected markets.

The question:

Does spending more create additional conversions or revenue?

This can be particularly useful before increasing budgets nationally.

---

Go-Dark Test

Turn off or substantially reduce advertising in selected markets.

The question:

How much business disappears when advertising disappears?

This can provide powerful evidence but may be commercially uncomfortable because management must deliberately reduce advertising activity in some markets.

---

Holdback Test

Keep selected markets from receiving a particular marketing intervention.

The question:

How much better do exposed markets perform compared with the expected baseline?

The appropriate design depends on geographic scale, conversion volume, advertising spend and the business question being tested.

---

A Practical Example: Should We Increase Meta Ads From $100,000 to $150,000?

Imagine a US ecommerce brand spends:

$100,000 per month on Meta Ads.

Platform-reported performance looks strong.

Management wants to increase the budget to:

$150,000.

The question shouldn't simply be:

“What is our current ROAS?”

A better question is:

“Will the additional $50,000 create enough incremental revenue to justify the investment?”

The business could test increased investment in selected geographic markets while maintaining business-as-usual spending in comparable markets.

Suppose the experiment eventually estimates:

Additional advertising spend: $20,000

Incremental revenue: $70,000

Estimated incremental ROAS:

$70,000 ÷ $20,000 = 3.5x

Management now has evidence that relates directly to the decision it is trying to make:

Does incremental budget create incremental business?

That is much stronger than assuming historical platform ROAS will remain constant as spending increases.

---

Why Incrementality Becomes More Important as AI Controls Advertising

Advertising platforms increasingly automate:

Bidding.

Audience expansion.

Search matching.

Creative combinations.

Budget allocation.

Campaign optimization.

Landing-page selection.

Google's AI Max evolution is one example.

AI Growth Wire previously examined how businesses can test Google Ads AI Max using experiments and planning tools rather than simply switching on additional automation and assuming it improves profitability.

The same principle applies to measurement.

As advertising platforms make more decisions automatically, businesses need an independent question:

Is the automation generating additional profitable demand?

Otherwise, companies risk optimizing aggressively toward metrics defined and measured inside the same advertising ecosystem.

AI-powered optimization and causal measurement should therefore evolve together.

---

First-Party Data Comes Before Sophisticated Measurement

Businesses may be tempted to jump directly into incrementality experiments or marketing mix modeling.

But measurement quality starts much earlier.

Before advanced analysis, verify:

Are purchases tracked correctly?

Are qualified leads distinguished from raw leads?

Are offline sales connected where relevant?

Are duplicate conversions removed?

Are revenue values accurate?

Are refunds handled correctly?

Are CRM stages meaningful?

Can customers be consistently identified where permitted?

Are appropriate consent processes in place?

If these foundations are weak, sophisticated modeling can create false confidence.

This is closely related to the challenge discussed in AI Growth Wire's guide to ChatGPT Data Agent for business analytics.

AI can dramatically accelerate analysis, but it cannot make fundamentally unreliable source data trustworthy.

The principle applies across analytics systems:

Better analytics begins with better data—not a more impressive analytics tool.

---

Don't Run a Before-and-After Test and Call It Incrementality

Suppose sales were:

$1 million in August

and:

$1.2 million in September

while advertising spend increased.

It is tempting to conclude:

Advertising generated $200,000 of additional revenue.

That conclusion may be wrong.

September may naturally be stronger.

Competitor activity may have changed.

Pricing may have changed.

Organic traffic may have increased.

Email campaigns may have improved.

Inventory availability may have changed.

A proper incrementality experiment attempts to estimate what would have happened without the marketing change.

That's why control groups and counterfactual modeling matter.

---

What Does a Business Need Before Running a Geo Experiment?

Geo experiments require more than advertising-platform access.

Businesses generally need sufficient:

Geographic diversity.

Conversion volume.

Historical data.

Marketing spend.

Reliable geographic KPI data.

Operational consistency.

The experiment also needs enough time and statistical power to distinguish an actual marketing effect from normal business variation.

That means sophisticated geo experimentation won't be appropriate for every small business.

A local service company spending $2,000 per month on Google Ads probably has more immediate priorities.

It should first make sure:

Calls are tracked.

Qualified leads are recorded.

Sales are connected to lead sources.

Revenue values are accurate.

Campaign waste is controlled.

Incrementality becomes more valuable as marketing scale and measurement ambiguity increase.

---

The Incrementality Readiness Ladder

AI Growth Wire recommends progressing through five measurement levels.

Level 1 — Conversion Tracking

Can you reliably measure what customers do?

Level 2 — Business Outcome Tracking

Can you distinguish a lead from a qualified lead, sale or profitable customer?

Level 3 — Cross-Channel Reporting

Can you understand performance across channels rather than relying only on individual advertising dashboards?

AI Growth Wire's guide to AI reporting automation explains how businesses can automate recurring management reporting while maintaining accuracy controls.

Level 4 — Incrementality

Can you determine whether marketing caused additional outcomes?

Level 5 — Budget Optimization

Can you use causal evidence to determine where the next marketing dollar should be invested?

Do not jump directly to Level 5 when Level 1 is unreliable.

---

Attribution ROAS vs Incremental ROAS: Which Should You Use?

You generally shouldn't choose one and abandon the other.

They answer different questions.

A mature measurement strategy uses different methods for different decisions.

---

The I-M-P-A-C-T Framework for Marketing Measurement

AI Growth Wire recommends evaluating important marketing investments using six questions.

I — Instrument

Is tracking reliable?

M — Measure

What happened?

P — Prove

What happened because of marketing?

A — Analyze

Why did the result occur?

C — Calculate

Was the incremental outcome profitable?

T — Take Action

What should change in the next budget cycle?

This prevents businesses from stopping at:

“The dashboard says ROAS is good.”

Measurement should eventually produce a business decision.

---

How Should SMBs Start With Incrementality?

Don't begin by testing every advertising channel.

Choose one economically important uncertainty.

Ecommerce

“Does branded paid search create incremental sales or mostly capture existing demand?”

Lead Generation

“Does increasing paid social investment create additional qualified opportunities?”

Retail

“Does connected TV investment increase store sales?”

Subscription Business

“Does retargeting create additional subscriptions?”

Established Brand

“What happens to revenue when advertising is reduced in selected markets?”

The experiment should answer a business decision, not demonstrate that the company knows how to run an experiment.

---

A Practical Measurement Plan

Phase 1 — Audit the Data

Verify:

Conversions.

Revenue.

CRM outcomes.

Offline conversions.

First-party data.

Geographic information.

Attribution settings.

Phase 2 — Define the Decision

Write one question.

For example:

“Should we increase paid social investment by 30%?”

Phase 3 — Define the Business KPI

Avoid vanity metrics.

Prefer:

Revenue.

Qualified opportunities.

Gross profit.

New customers.

Subscriptions.

Store sales.

Phase 4 — Design the Experiment

Determine:

Treatment markets.

Control markets.

Historical period.

Test duration.

Budget intervention.

Minimum detectable effect.

The experiment needs enough statistical power to detect a commercially meaningful change.

Phase 5 — Protect the Experiment

Avoid unnecessary differences between treatment and control markets that could distort the comparison.

Phase 6 — Analyze

Evaluate:

Incremental conversions.

Percentage lift.

Incremental revenue.

iROAS.

Confidence intervals.

Phase 7 — Make a Budget Decision

Scale.

Maintain.

Reduce.

Retest.

An experiment has business value only when it informs action.

---

How to Calculate Incremental ROAS

A simplified calculation is:

Incremental ROAS = Incremental Revenue ÷ Incremental Advertising Spend

Example:

Additional advertising spend:

$30,000

Estimated incremental revenue:

$90,000

iROAS:

$90,000 ÷ $30,000 = 3.0x

Now add business economics.

Suppose gross margin is:

40%

Incremental gross profit before advertising:

$90,000 × 40% = $36,000

Subtract incremental advertising spend:

$36,000 − $30,000 = $6,000

Suddenly the 3.0x iROAS looks less spectacular.

This is why businesses should eventually connect incrementality to profit, not simply revenue.

---

Five Marketing Incrementality Mistakes to Avoid

1. Treating Platform Attribution as Causal Proof

A platform recording a conversion does not automatically prove that advertising caused it.

2. Testing Too Small a Sample

An underpowered experiment may fail to detect a commercially meaningful effect.

3. Changing Too Many Variables During the Test

Major pricing, promotion or operational changes can complicate interpretation.

4. Ignoring Profitability

Incremental revenue can still be unprofitable revenue.

5. Automating Decisions Too Quickly

AI systems may increasingly recommend budgets using sophisticated measurement signals.

Businesses should still understand the assumptions behind consequential investment decisions.

---

Where Does AI Fit Into Marketing Measurement?

AI can increasingly help businesses:

Prepare data.

Detect anomalies.

Build models.

Explain results.

Generate dashboards.

Investigate performance.

Recommend actions.

The recently launched ChatGPT Data Agent for business analytics is another example of this broader shift toward conversational business analysis.

But AI does not remove the need for experimental discipline.

A beautifully explained incorrect model is still incorrect.

The future of marketing analytics therefore isn't simply:

More AI.

It is:

AI + better data + causal measurement + human commercial judgment.

---

How Incrementality Fits With Automated Business Reporting

Incrementality experiments shouldn't exist in isolation.

Once a business starts measuring incremental outcomes, those results should eventually become part of management reporting.

For example, instead of a weekly report showing only:

Meta Ads ROAS: 4.5x

leadership might eventually see:

Attributed ROAS: 4.5x

Latest tested iROAS: 2.8x

New-customer CAC: $72

Contribution margin after advertising: 14%

This creates a much richer view of marketing economics.

Businesses already building AI-powered weekly reporting workflows can gradually incorporate causal-measurement results alongside traditional advertising KPIs.

The objective isn't to eliminate platform reporting.

It's to prevent platform reporting from becoming the only version of marketing performance management sees.

---

AI Growth Wire Verdict

Marketing measurement is moving beyond the question:

“Which ad got the conversion?”

toward a much more important question:

“What additional business did our marketing investment actually create?”

For smaller advertisers, that doesn't mean immediately implementing sophisticated econometric models.

The correct progression is:

Track accurately → Measure business outcomes → Understand attribution → Test incrementality → Optimize investment

For larger ecommerce brands and businesses already spending heavily across multiple advertising channels, however, incrementality deserves considerably more attention.

A 5x attributed ROAS can look impressive.

But the more valuable number may be:

How much additional profitable revenue appeared only because the advertising existed?

That is the number that helps a business determine where its next marketing dollar should go.

---

Frequently Asked Questions

What is marketing incrementality?

Marketing incrementality measures the additional sales, conversions or other outcomes caused by marketing compared with what would likely have happened without that marketing activity.

What is the difference between attribution and incrementality?

Attribution determines which marketing interaction receives credit for a conversion. Incrementality attempts to determine whether marketing actually caused an additional conversion.

What is incremental ROAS?

Incremental ROAS, or iROAS, compares incremental revenue caused by a marketing intervention with the incremental advertising spend required to create it.

What is Google Meridian GeoX?

Meridian GeoX is Google's open-source framework for designing and analyzing geographic incrementality experiments. It helps businesses estimate the causal effect of changing marketing investment across geographic markets.

What is a geo incrementality test?

A geo experiment divides geographic markets into treatment and comparison groups. Marketing activity changes in treatment markets while comparison markets help estimate what would have happened without the intervention.

Can geo experiments measure advertising outside Google Ads?

Geo-experiment methodology can be used to evaluate marketing interventions across channels when the business has suitable geographic data and a properly designed experiment.

Do small businesses need incrementality testing?

Not necessarily. Businesses should first have reliable conversion tracking, sufficient data and enough marketing scale for the result to influence a meaningful investment decision.

Does a high ROAS prove that advertising is incremental?

No. ROAS describes the relationship between attributed revenue and advertising spend. Some attributed customers may have purchased without the advertising, which is why causal measurement requires a different approach.

When should a business start measuring incrementality?

Incrementality becomes particularly useful when advertising spend is substantial, multiple channels claim overlapping conversions, management is considering significant budget changes, or the business wants to distinguish attributed demand from genuinely incremental growth.