From the very early days of Haus, we’ve made our viewpoint on the pros and cons of attribution models pretty clear. While tools like multi-touch attribution (MTA) were once the best game in town, they always had a key flaw: They measured correlations between marketing exposure and outcomes, but didn’t measure whether a marketing exposure caused an actual lift in outcomes.

In recent years, incrementality has emerged as a causality-based alternative to attribution models. Instead of measuring conversions that would have happened with or without a given marketing intervention, incrementality tests use control groups to isolate the true impact of your campaign. (Take a gander at our Incrementality School series for more information — there’s no homework, we promise.)  

Let’s get more detailed and dive into the key differences between attribution and incrementality. We’ll explore how each approach works, when you might use each one, and some advantages and disadvantages of both. 

What is attribution and how does it work?

Attribution models credit a conversion or sale to the marketing touchpoints a customer interacted with on their journey. For instance, if a user clicks your Meta ad right before converting, that Meta ad will get credit for the conversion. There are many types of attribution models, each with its own way of dividing up credit:

• Last Touch Attribution: The last ad or touchpoint a customer clicks before converting gets all the credit. So if a customer sees several ads but clicks a Google search ad just before buying, that search ad gets full credit for the sale.
• First Touch Attribution: You guessed it: The first touchpoint gets credited in a first-touch attribution model. More specifically, it’s credited as the touchpoint that introduced the customer to your brand or product.
• Linear or Multi-Touch Attribution (MTA): Credit gets spread across multiple interactions, either equally or using a more complex model driven by machine learning.

Modern attribution models don't just rely on deterministic tracking (like cookies). They're increasingly using statistical modeling and first-party data because privacy changes have limited traditional tracking methods. 

How does attribution work in practice?

Let's say Clara sees five different ads across Meta, Google, and email over two weeks before buying a subscription to an online fitness program. Under last-touch attribution, the last ad she clicked (maybe a Google remarketing ad) gets all the credit. Under a linear multi-touch model, each of the five ads would get 20% of the credit for her purchase.

What are the advantages and disadvantages of attribution?

Attribution models are popular because they offer near real-time feedback and detailed insights into the customer journey. But here's the thing — they're not inherently causal. Just because a click happened before a conversion doesn't prove that the click caused the conversion.

Popular ad platforms have gotten remarkably good at serving ads to users who would have converted anyway. But if these ads still get credit for conversions, it’s natural to assume that many platform-reported metrics are inflated. So if you rely purely on attribution, you may end up spending too much on ads. 

Our Industry Survey found that only 42% of respondents trust their first/last-touch attribution models, making it one of the least trustworthy measurement solutions available. Given its focus on correlational data, it’s not hard to see why. 

How does incrementality differ from attribution?

Incrementality moves beyond mere attribution and asks a deeper question: How many conversions or sales wouldn't have happened without a specific marketing action or spend? It's all about uncovering the causal impact of your marketing activities, not just the correlation between exposure and outcomes.

You typically measure incrementality by running experiments, which most often take the form of randomized controlled trials (RCTs). In these tests, one group of people sees a marketing action while a similar group doesn't. The difference in outcomes gets attributed to the marketing action. 

A drug trial in the medical industry is a helpful point of comparison. In these cases, a pharmaceutical company gives a medication to one group and a placebo to a similar group and tracks outcomes. If both groups have vastly different outcomes, the medication is likely effective.

How does incrementality testing work in practice?

Say an online retailer wants to know if their new Instagram campaign is actually increasing sales, or if customers would've bought anyway. They split their market into two similar geographic regions: Region A sees the Instagram campaign (the "test" group), while Region B doesn't (the "control" group).

If Region A sees 12% higher sales than Region B after the campaign, they can reasonably say this incremental lift came from the campaign — assuming everything else stayed consistent.

Methods and real-world challenges

Incrementality testing has grown beyond simple geographic split tests. Methods like "synthetic control" (where you mathematically construct control groups from multiple sources to match the test group as closely as possible) make results more accurate, especially when simple 1:1 matches aren't possible.

We favor this approach because it gives you insights into true causality — crucial for optimizing spend and justifying marketing investments to executives.

But there are downsides too:

• Cost and Complexity: Experiments are more resource-intensive than standard attribution and can require turning off spend in parts of your market.
• Frequency and Detail: You can't easily run thousands of experiments at once or get high-frequency data; each test gives you a "snapshot" under specific conditions.
• Applying Results: The results of one experiment might not translate perfectly to other periods, regions, or campaigns.

For these reasons, we don’t advise teams to build their incrementality practice entirely in-house. Partnering with a trusted incrementality testing partner is often the most cost-effective solution, leading to more precise, actionable results. Read our advice for choosing an incrementality testing partner.

Attribution vs. incrementality: the key differences

At the core, it's the difference between correlation and causation.

Attribution tracks customer interactions and tries to figure out which touchpoints "matter" by linking them (statistically or through rules) to conversions. It can help you map journeys, analyze channels, and access frequent, detailed reports. But its conclusions are always questionable: Correlation doesn't equal causation.

Incrementality goes further, isolating the actual, causal effect of marketing — the incremental impact. It answers: "What would've happened if we hadn't run this campaign?" It's the gold standard for answering business questions about marketing ROI. 

Here's the fundamental difference: If a customer was always going to buy your product but happened to click an ad out of habit, attribution might give credit to the ad. Incrementality testing might show that for this group, the ad had no incremental effect on conversions.

How do privacy changes impact attribution models?

The loss of third-party cookies and broader privacy regulations have significantly weakened many attribution models, especially those relying on deterministic user-level tracking across platforms. So marketers are increasingly relying on first-party data and more robust statistical inference — but this adds noise and uncertainty.

In contrast, incrementality experiments (when properly set up) are less affected by tracking limitations. They only need random assignment and accurate outcome measurement, not complete individual-level journey data.

When should you use each approach?

Both attribution and incrementality have their place in modern measurement. Large organizations with big budgets and multichannel activity will typically use both, often along with additional models like Marketing Mix Modeling (MMM).

• Attribution works well for fast, ongoing insights and for channels where experiments would be too hard or expensive to run frequently. It can flag performance changes that need further (causal) investigation.
• Incrementality is best when you need to justify investment, evaluate new channels, or settle resource allocation debates where understanding true cause and effect matters most.

Can incrementality tests improve attribution insights?

The best organizations use incrementality to calibrate their attribution and MMM models. For example, they might run a few robust experiments per year, then tune their ongoing attribution models so these models' results align with causal ground truths from the experiments. This way, attribution estimates become more reliable for day-to-day optimization, while incrementality makes sure bigger strategy moves are rooted in reality.

The bottom line

Attribution and incrementality are both essential for modern marketing departments, but they're not interchangeable. Attribution tells you where activity happens and how it correlates with conversions, while incrementality tells you what would truly change if you changed your marketing spend or tactics. Attribution is like checking footprints on a path; incrementality is like controlling the weather to see if people would still walk that way.

Smart organizations blend both approaches: using attribution for frequent optimization and incrementality for strategic decisions, calibrating the former with the robust truth-seeking of the latter. As privacy changes and complexity grow, understanding — and acting on — the difference between what looks influential and what's truly causal matters more than ever.

Attribution and incrementality FAQ

What is the concept of incrementality?

Incrementality seeks to determine how many conversions or sales would not have happened without a specific marketing action or spend. It aims to uncover the true causal impact of marketing activities, rather than just showing correlation between exposure and outcomes. 

Incrementality is typically measured through experimentation, most notably randomized controlled trials (RCTs), where one group is exposed to a marketing action while a similar control group is not. The difference in outcomes between these groups can then be attributed to the marketing action.

What is the difference between attribution and incrementality testing?

The core difference lies in correlation versus causation:

• Attribution tracks customer interactions and attempts to assign credit to touchpoints by linking them to conversions. It's good for mapping journeys and providing frequent, detailed reports, but can only show correlation. 
• Incrementality testing isolates the actual causal effect of marketing by answering "What would have happened if we hadn't run this campaign?" Through controlled experiments, incrementality provides evidence of true cause and effect, while attribution can only infer relationships between touchpoints and conversions.

What is an example of an incrementality test?

Baby brand Lalo hypothesized there was a bigger audience that they could reach, but their current marketing was reaching folks who were going to convert anyway. So they designed upper-funnel incrementality tests on Meta and TikTok. 

The test results showed that running upper-funnel campaigns to reach new prospects wasn't just effective — it also delivered surprisingly efficient sales for the business. The team has since layered upper-funnel tactics into their evergreen Meta and TikTok strategies. Read their full story here. 

What is incremental attribution in Meta?

Incremental attribution in Meta refers to measurement approaches that help advertisers understand the true causal impact of their ads on the platform. Meta offers tools that help advertisers measure lift (the incremental impact) of their campaigns through randomized controlled experiments. These experiments compare the behavior of people who see ads with similar people who don't. 

The platform has developed various methodologies to address the challenges of measuring incrementality in digital environments where traditional tracking is becoming more limited due to privacy changes. These solutions help advertisers determine not just which ads were seen before conversions, but which ads actually caused conversions that wouldn't have happened otherwise.