As a marketer, or a data scientist working in the marketing field – imagine if you were able to get each individual users’ journey before finding your product. You could utilize iOS FaceID to deterministically know that this person is the same person across all touch point. You’d be able to know about every ad they saw, engaged with, as well as the ones they did not. Location data would signal that they saw a billboard with your brand on it, exactly at 17:02 on a Tuesday. You would even know that they spent over 20 seconds around the billboard, as you would have real time traffic information, and data would show that the user was in a standstill traffic jam at that point. Imagine that you could derive the users’ proximity to other users’ who are already using your product. 

What if you could also identify this users’ inclination to usage, and what if you could have all of this data over all users?

If you had this enormous amount of data, you would win the digital advertising game. Using this data on a large sample of users, you could identify the exact weight and value of each and every ad engagement. You’d be able to understand how many ad interactions does it take before a user “converts”, as well as understanding variables such as location, affinity towards other products, and so on. You could build the perfect contextual, as well as the perfect user based targeting model.

Owning such a data set could allow you to reach a perfect conversion funnel: 
1 Impression = 1 Click = 1 Conversion = 1 Action

Now put yourself in the shoes of the user. Imagine a company owning so much of your data, to answer the question: Why did you download their app? Or why did you purchase these specific shoes?

Multi-touch attribution was a dream of mine (speaking as CEO @ INCRMNTAL) before understanding that gathering an enormous amount of user level data is a complete overkill when it comes to something as simple as understanding: what impacts performance?

Putting aside everything that could happen if personally identifiable data would leak, there was never any reason for data to be so freely available. I have claimed in the past that the first 20 years of digital advertising were like the wild wild west. It was all new, and therefore, there were barely any regulation to how, and how much personally identifiable data company could gather and store for infinity.

Our approach with incrementality measurement  had to completely break the paradigm of user based measurement. With incrementality measurement - We don’t know why did you download a certain app, or why you bought shoes. Heck – maybe you got the shoes as a gift.

What incrementality measurement tries to do is ask a different question: 
Were there any more installs than expected ? did we actually sell more shoes? 

If the answer is “no” – than the channel or campaign you’ve just started is cannibalizing your own marketing performance.

If there were more installs or sales:
why were there more installs of your app? Or why did shoe sales increase? 
Incrementality wouldn’t try and ask this about a user (singular) , but look at the data from a causality standpoint.

Could we explain it by seasonality patterns? Ah no? then it might be because you increased ad spend over a specific campaign, as we can attribute an increase in installs / sales every time you increased the ad spend of this campaign, while attributing a decrease in installs / sales every time you scaled it down.

The more data we have about your own marketing activities, the more confidence we can build in the results understanding the impact of your marketing activities over your marketing performance.

With incrementality measurement, users can sleep well at night. no one is tracking their sleep. 
Unless they use a sleep tracker app.