You Might Like These Articles

Here, almost no one answered correctly. The answers varied between “everyone would claim” to only Facebook and Twitter would claim as the ones generating the install.

The right answer, would have been: GoogleLiftoffFacebook, and Twitter would have claimed the install.

Yes - the same install would have appeared in 4 different platforms.

What is more interesting is that if we would have replaced Applovin, Unity, and Vungle with: Amazon Ads, Apple Search, Pinterest, and/or Snapchat - ALL would have claimed the attribution.

Some of the largest media vendors, those accountable for up to 80% of budgets spent are Self Attributing (or Self Reporting , SANs, SRNs).

These platforms do not depend on the MMPs to validate “who generated the install” , but only utilize the MMP to receive ALL install events, while claiming the attribution on their own reporting themselves. (i.e. “I showed the user an ad. An install happened. I generated the install”).

Most Self Reporting platforms do not share impression or click data with MMPs.

The MMPs role is to deduplicate the claims and declare a clear “winner” based on who touched the user last, alongside the attribution model in respect to the weights of this or that impression or click.


Now that you understand the complexity of Attribution, Lets ask the real question: Who CAUSED the Attribution.

To demonstrate this, lets try and add context to the user journey:


Most people recognized that Liftoff would win the attribution in this scenario. A click event is a clear sign of intent, and the fact that the click contained a device identifier, enables an MMP to match between the click and the install in a deterministic way rather than adding probabilistic logic to the process of attribution.


Who Claimed the Attribution

The quiz was somewhat more difficult, when we asked - “who CLAIMED the attribution?

In this example, it is much less straightforward to determine who should get 100% credit for the install.

One could claim that ALL vendors contributed to the install in some way.

One could claim that only Twitter should receive credit.

One could claim that the seeing a video served by Vungle, clicking the end card and making the decision to install 3 days later, had the most “weight” towards the user intent.

One could also claim that Facebook and Google, had no right to claim the attribution.


Last Touch Attribution:

Last Touch attribution is not a great way to measure causality.

It’s a great proxy allowing a marketer to run optimization in real time.

Measuring causality takes a completely different approach from attribution as mobile marketers know it today.



INCRMNTAL is an incrementality testing platform providing ammunition to the digital marketer. The platform uses machine learning and AI to provide incrementality and cannibalization actionable insights acting as a strategic tool to unlock the value of your marketing budget.

If you want to learn more, visit INCRMNTAL or book a demo today!


Winning, Claiming, And Causing Are NOT The Same

(And why Marketing is anything but deterministic...)

Assuming you set up your attribution well. Assuming you’re working with some of the best media vendors, those with direct access to publishers, those who are well known in the market as not opening their gates to bad publishers - does that give you safety that your marketing mix is adding value ?


Mobile Attribution today works using “Last Touch Point”. It’s the only way, as the mobile app eco-system is guarded by the operating systems walled gardens.

But it is not as straight forward, as Attribution providers (MMPs) use various ways to select who should win the attribution based on the parameters arriving with impressions or click events.

That means that a click that carries more data gets a stronger “priority” in the attribution model vs. a click with less data.


Who Won the Attribution

We ran a quiz asking people - “who WON the attribution?” using current last-touch-point attribution mechanics, where various events had different parameters included. Some with device identifiers (IDFA), some only with IP, some with no identifier data. 

Did you Like this Article ? Share It!