Attribution vs. Incrementality
Time to put the past in the past

When Apple killed the IDFA, the marketing industry reacted like someone unplugged life support. People catastrophized it (including me) as if it was going to ruin their marketing capabilities.

Truth is – attribution was broken long before privacy showed up to the crime scene. 
Cookies, Identifiers, 28 days window of attribution were all symptoms this old. 
We didn’t lose measurement when privacy happened. We just lose the illusion of precision.

The Illusion of Precision

For years, marketers worshipped attribution dashboards like sacred scrolls. Every chart had exact numbers, presented at full confidence, even though this was fiction. 

Attribution presented numbers as if it was measuring with confidence, but it was confidence theater. Attribution made marketers feel like scientists while they were really just cheering for whichever ad took the last click. It measured activity, not impact.

That’s how you end up with 12 platforms claiming credit for the same conversion and everyone celebrating their own brilliance.

The Birth of Attribution (and Its Original Sin)

Attribution was invented when the web was simple. A user saw an ad, clicked, bought. Easy. Attribution was invented in a world where we connected to the internet using a dial up modems, waiting twenty minutes to download an MP3, only to be stopped when Mom wanted to use the phone. 

Attribution was linear, it was trackable, and it gave marketers the comfort of thinking they’d finally cracked human behavior.

Except humans aren’t linear. We’re unpredictable, distracted, emotional, inconsistent, and occasionally shopping at 2AM for reasons that have nothing to do with marketing.

And while technology evolved: supercomputers were placed in our pockets, internet became over 10,000 times faster (584mbps vs 56kbps), Mom was able to use the phone while doom-scrolling, and…who downloads an MP3 these days??

Attribution assumes we’ll continue behaving like computers – Click > Buy > Click Buy. Incrementality starts by admitting we don’t.

Attribution counts clicks. Incrementality measures value.

Privacy Didn’t Break Measurement – It Exposed It

When privacy rules started trickling down (GDPR in 2017, ATT in 2021, Cookies deprecation in 2005 to date…) – the panic in the industry was prevalent. Attribution models built on user IDs started losing their confidence. 

Suddenly matching between a click and device using meta data such as “define type” (i.e. click came from a Chrome browser, Conversion came from a Chrome browser, MUST BE the Same Person!)

The loss of IDs didn’t kill attribution measurement. It just helped us see how fragile it really was. If your entire system collapses because you can’t follow people around the internet anymore, that’s not privacy’s fault. That’s a design flaw. Privacy didn’t kill attribution. It simply turned on the lights and showed us the cracks.

Attribution is obsessed with visibility – “Who touched the user last?”
Incrementality is obsessed with causality – “What actually caused the lift?”

That’s not semantics. It’s the entire difference between knowing what happened and knowing what worked.

Attribution asks, “Which ad got the credit?”
Incrementality asks, “Would this result have happened anyway?”

It’s the difference between counting clicks and measuring impact.

So What Is Incrementality, Really?

Incrementality is the science of understanding what would have happened if you did nothing.

It’s the counterfactual – the parallel universe where the campaign didn’t run, and you can see what truly changed because of your marketing. It’s not about tracking users. It’s about measuring impact.

Incrementality measures causal impact — the true contribution of your marketing efforts, stripped of noise, overlap, and wishful thinking.

Then AI Showed Up and Broke Everything Again

Just when we started adjusting to privacy, along came AI. Dynamic creatives, predictive audiences, personalized everything – changing every millisecond. Great for marketing. Terrible for attribution, especially in a privacy first world. 

You can’t run deterministic tracking on systems that change faster than you can export a CSV. AI broke the idea of “fixed journeys” completely, but it gave birth to a whole new paradigm in measurement. An AI method of measurement to measure incrementality using the day to day changes marketers do already.  

AI allowed us to use existing methodologies (i.e. Causal Impact), and innovate on top of those with technologies we didn’t think were previously possible, such as. 

• Causal Discovery - learning the relationships between changes to one another
• Reinforced Learning – rewarding the modeling for good predictions and penalizing it for bad ones (i.e. “bad algorithm!”)

Attribution was built for static cause-and-effect.
AI made cause-and-effect dynamic, chaotic, and sometimes unknowable.

So now we’re flying blind, and most marketers are too busy staring at pretty dashboards to notice they’re upside dow.