Attribution 101: Finding the right measurement model for your business

Jeremy Freedman / March 2019

These days, it seems like everyone is talking about attribution — what’s the best model to use? How do you make the most of the data you have to make smart business decisions?

Before you can even address those questions, however, you’ll need to have a solid grasp on what attribution is and why it matters.

Attribution is the process of assigning a value to individual actions that lead to a certain outcome. The marketing interactions and conversions collected by a platform allow advertisers to evaluate the customer journey and determine how media will be valued. This process can be challenging for any company, whether you’re a small, growing business or a large global brand.

We’ll start by demystifying the process and help you determine which model is best for your business.

The conversion journey

Let’s take you through a simple measurement and attribution journey that’s fairly common and based in real life: the journey to getting married.

Scale Up Inline 1

Step one usually starts with a swipe on a dating app. Step two might be going on a coffee date and step three is a movie date. Step four is the proposal (yes, it’s pretty quick in this particular model, but sometimes that’s the way it goes), and finally, the conversion: marriage.

So this is the path our user has taken — but which step should get credit for the conversion? The dating app? The coffee date? The proposal itself?

Let’s go through this and demonstrate how perspective is the key to all of this.

Scale Up Inline 2_V2

If you ask the people behind the dating app, they would say that it’s first-touch attribution: the app is what drove this conversion. If that swipe hadn’t happened, these two people never would have met, dated, or gotten married.

If you ask the person who did the proposing, they might say it’s last-touch attribution — they bought the ring, they popped the question, their partner said yes, so that deserves all the credit. Without a proposal, they never would have gotten married.

But if you ask a data scientist, they’d say that all of these different steps deserve some element of credit for this conversion. If any of them had been missed or skipped, the marriage wouldn’t have happened.

Steps toward better attribution

For quite a while, the last-touch attribution model was all we had to go on. We simply didn’t have a way of seeing how each touch contributed to the ultimate conversion. But a number of tools and capabilities now exist to give us a more holistic view, or what’s known as data-driven attribution models.

Scale Up Inline 3

First-touch, linear, position based, and time decay are all rules-based attribution models. They’re about aligning philosophically within an organization and deciding that this model is the lens you want to apply to the conversion journey.

But data-driven attribution is a wholly different way of looking at it. Powered by machine learning, it takes all the guesswork and philosophizing out of the equation, and looks at conversions objectively based on trends. Going back to the example of the couple’s conversion journey, it will take into account every marriage on record and find patterns — in cases where the couple didn’t go to the movies, were there fewer weddings in the end? If so, then that movie date in the middle starts to get a bit more credit for the conversion.

Measurement platforms with data-driven attribution capabilities, such as Google Analytics 360 or Google Campaign Manager, use machine learning to look at every conversion and all the elements that led to it, to determine which steps needed to be there and which were less important.

Regardless of which attribution model you use, your goal should be to get as big a scope as possible on all the information in order to drive better, faster decisions. Attribution will allow you to answer key questions such as: how much does each touchpoint contribute towards driving a conversion? Do you have all the information about what’s driving those conversions? Are you looking at that information through the right lens?

If you can find the answers to those questions, your measurement tools and your business will have a happy and successful marriage.

Demystifying machine learning — and why you should start using it now