The Attribution Trap in Advertising: Incrementality vs ROAS

What Happened Because of Advertising?

There is a question every marketer eventually asks:

“Did my advertising work?”

And usually, we reach for ROAS. Spend ₹1 on advertising. Generate ₹5 in sales.

5x ROAS.

Looks great. But there is a small problem. We don’t actually know whether the ₹5 happened because of the advertising. Maybe the customer was already going to buy. Maybe they had the product in their cart. Maybe they searched for the brand five minutes earlier. Maybe they were already a loyal customer. Maybe a competitor’s ad would have had no impact either. And yet, we may still give the entire sale to the ad.

This is where incrementality becomes interesting. Incrementality asks a much harder – and much more useful – question:

What happened specifically because of the advertising that otherwise would not have happened?

That distinction changes how we think about advertising effectiveness.

The Attribution Trap

Digital advertising made it incredibly easy to connect an impression to a purchase.

A consumer sees an ad. They click. They buy. The system records a conversion. The campaign gets credit.

The problem is that “correlation is not causation.”

Consider a simple example.
A brand spends ₹10,000 and sees ₹50,000 in attributed sales.

ROAS = 5x

Looks like a great campaign.

Now run a control group.
Suppose similar consumers who did not see the ads would have generated ₹40,000 in sales anyway.

Then the advertising actually created only:
₹50,000 − ₹40,000 = ₹10,000 incremental revenue

So:
Incremental ROAS = ₹10,000 / ₹10,000 = 1x

The dashboard says 5x. The causal economics say 1x.

That gap is the attribution trap.

ROAS Measures What Happened. Incrementality Measures What Changed.

The Counterfactual Is the Whole Game

Incrementality is fundamentally about a counterfactual.

We observe:
What happened when someone was exposed to advertising.

But what we really want to know is:
What would have happened if that same person had not been exposed?

We can never observe both realities for the exact same person at the same moment. So we construct the closest possible approximation.

Treatment group: People who received the advertising.
Control group: Similar people who did not.

Then we compare outcomes.
If the treatment group generates 10,000 purchases and the control group would have generated 8,000, the incremental impact is approximately: 2,000 additional purchases.

Those 2,000 purchases are the economic value that advertising actually created.

That is incrementality.

ROAS Is Not the Same as Incremental Value

This becomes especially important when advertising targets consumers with high purchase intent.

Imagine two consumers.

Consumer A
Probability of buying without advertising: 90%
Probability of buying with advertising: 92%

Incremental lift: 92% − 90% = 2 percentage points

Consumer B

Probability of buying without advertising: 30%
Probability of buying with advertising: 60%

Incremental lift: 60% − 30% = 30 percentage points

Consumer A is easier to convert. Consumer B is more valuable to advertising.

This is the fundamental difference between harvesting demand and creating demand.

Demand Capture vs. Demand Creation

I think this is the simplest way to understand incrementality.

Demand Capture: The customer already has strong intent. Advertising helps capture the purchase.

Examples:
• Branded search
• Retargeting
• Cart abandonment
• High-intent audiences

Think:
Brand search → Ad → Purchase

Suppose 1,000 people search for your brand. Without advertising, 700 would purchase. With advertising, 750 purchase.

Incremental purchases: 750 − 700 = 50

Advertising generated 50 additional purchases, even though it was associated with 750 purchases. Most of the observed demand was already there.

In other words:
The easiest customer to convert is often the customer who needed the least advertising.

Demand Creation: Now take a different audience. The customer wasn’t necessarily going to buy. Advertising changes their behavior.

Examples:
• New product discovery
• Competitive conquesting
• Category expansion
• Cross-category adoption
• New customer acquisition

The incremental contribution can be much higher

Think:
Without advertising, 200 of 1,000 consumers purchase. With advertising, 350 purchase.

Incremental purchases: 350 − 200 = 150

Same number of consumers. Much greater causal impact. The advertising is actually changing behavior. This distinction matters because both campaigns can look successful in a traditional dashboard.

But they are doing fundamentally different jobs.

Capture converts existing intent. Creation changes intent.

And incrementality tells us how much of each is actually happening.

Why This Changes How We Think About ROAS

ROAS is still useful. The problem is treating it as the final answer.

Consider two categories:

If you optimize for ROAS, you invest more in Category A. If you optimize for incrementality, Category B is clearly more attractive. Why?

Because Category A has strong organic demand. Category B has stronger advertising-created demand.

This is why I think incrementality is effectively a new way of looking at the ROAS of different parts of a brand.

Not simply: What is the ROAS of my brand?

But: What is the incremental ROAS of each category, product, audience and channel?

Every Brand Has an Incrementality Curve

A large brand rarely has one advertising job.

Some categories are already well known. Some are emerging. Some have strong brand loyalty. Some are highly competitive. Some require education. Some depend heavily on discovery.

The incremental value of advertising will therefore vary.

A simple portfolio might look like:

Core category
10x ROAS → 1x incremental ROAS

Growth category
5x ROAS → 3x incremental ROAS

New category
2x ROAS → 1.8x incremental ROAS

Traditional reporting says: Core is the winner.

Incrementality says:
Growth is where advertising changes the most behavior.

That can completely change budget allocation.

Measurement Evolution: From Attribution to Causality

This is really the evolution of advertising measurement. The questions have progressed:

Did someone see the ad?
↓
Did they click?
↓
Did they buy?
↓
Can we attribute the purchase to the ad?
↓
Would they have bought without the ad?

That last question moves us from attribution to causality.

And that distinction matters because marketers don’t spend money to receive credit. They spend money to change outcomes.

Eventually, From Measurement to Decisioning

And this is where incrementality becomes more than a measurement framework. If we can measure incremental impact, we can eventually optimize for it.

The evolution could look something like:

Attribution → Incrementality → Incremental Optimization

First, understand what happened. Then, understand what happened because of advertising. Then, allocate advertising dynamically toward the opportunities where it is most likely to create additional value.

That is the point where measurement becomes decisioning.

Optimize for Incremental Optimization.

AI Makes This Even More Interesting

AI could make incrementality much more granular. Today, experiments can tell us whether advertising created incremental value for a particular population or campaign.

But imagine systems that continuously learn:
– Which products create incremental demand?
– Which audiences need advertising to convert?
– Which customers would buy anyway?
– Which categories are being under-invested in?
– Which impressions are genuinely changing behavior?
– Which creative creates incremental consideration?
– Which placements create incremental purchases rather than simply harvesting existing intent?

Instead of optimizing toward predicted conversion probability, the system could eventually optimize toward something closer to:
Predicted incremental probability of conversion.

That’s a very different optimization target.

An AI system could learn that:

– Customer A has a 95% probability of buying anyway.
– Customer B has a 25% probability without advertising and 55% with advertising.
– Category X has high organic demand.
– Category Y is highly responsive to advertising.
– Placement A mostly captures existing intent.
– Placement B creates new demand.

The system could then shift spend toward the opportunities with the highest expected incremental value.

Incrementality at Scale

Historically, incrementality has often depended on carefully designed experiments. That will remain important. But AI could make the output much more actionable.

Instead of measuring one campaign at a time, imagine continuously estimating:
• Incremental ROAS by category
• Incremental ROAS by audience
• Incremental ROAS by product
• Incremental ROAS by placement
• Incremental ROAS by customer lifecycle
• Incremental ROAS by creative

The question then becomes less about reporting what happened last quarter and more about deciding where the next dollar should go.

That is where incrementality becomes strategically powerful.

The New Advertising Question

Advertising doesn’t have one ROI.

– Different categories have different levels of organic demand.
– Different customers have different propensities to buy.
– Different products need different amounts of persuasion.
– Different channels play different roles.

And therefore, different parts of the brand have different incremental economics. The future isn’t about finding the ROAS. It is about understanding the incremental ROAS curve.

I don’t think ROAS is going away. It shouldn’t. ROAS tells us the economic output associated with advertising.

But incrementality tells us something ROAS alone cannot:
How much of that output would not have existed without the advertising?

That distinction becomes increasingly important as targeting gets better. Because the better we become at finding people who are already likely to buy, the better our ROAS can look – even when the causal impact of advertising is small.

The ultimate goal isn’t to find the consumers most likely to purchase. It is to find the consumers whose behavior advertising can actually change. And perhaps that is the real future of advertising measurement:

Don’t optimize for where you get the most credit. Optimize for where you create the most change.

That is incrementality.

What happened specifically because we advertised — that otherwise would not have happened?

That may be the most important number in advertising.