Attribution Models

…explained within Digital Marketing

It’s difficult to know which attribution method is the most accurate. These are the most frequently used. The first-click attribution, last-click attribution, linear attribution, and the time-decay attribution. Let’s take a closer look at each.

Attribution

To discuss attributions within the world of digital marketing, we first need to define what
attributions are. “In marketing, attribution, also known as multi-touch attribution, is the
identification of a set of user actions that contribute to a desired outcome, and then the
assignment of a value to each of these events.” – Wikipedia.

In marketing we are attempting to understand what types of advertising brings people to us
and converts them into purchasing customers. We can use Google Analytics
to help us gauge people’s behaviors and then adjust accordingly. More on that later!

First, let’s look at Mr. Johnson who wants to buy Christmas gifts for his children. He sees
multiple ads on television promoting Target.com as a reliable source for helping him to find gifts for
his children. He sees banner ads for Target while browsing on the internet, and sees Target
advertisements in some of his favorite magazines. Not to mention the large red Target billboard
sign he sees every day on his way home from work.

When he gets his paycheck, he goes online to start shopping and notices an email from Target
offering him a discount if he buys today. So, since his exposure to Target has been ruminating in
his mind for the past several weeks, he decided to act on the discount and order a
bunch of toys online for his children, which he then has the option to pick up at the store.

Attribution Methods

Within Marketing, it’s difficult to know which attribution method is the most accurate. It seems
fair-minded to say that credit goes to the last-click interaction just before the customer makes
his purchases. But the accuracy of what influenced the final choice is of more value. Do people have control over their decision making, or is there another science at play here, influencing the situation to help arrive at an intended behavior?

It would be simplistic to give full credit to the last-click attribution, but other methods are to be
considered here as well. For example, consider first-click attribution, linear attribution, and time-decay attribution (there are other methods as well, found at: Overview of Attribution Modeling.)

To illustrate, say that a person is shopping and over a brief period has been exposure to clicking on an email link, clicking on a banner ad, and then clicks on a Google Search Ad where he makes his final purchase. Last-click attribution would give all the credit to the Google Search Ad. First-click attribution would give all the credit to the email link, linear attribution would give equal credit to all three, and time-decay would also give credit to all three but would put emphasis on the most recent click, or last-click attribution.

Attribution Tools

First option

Google Analytics has two tools that allow for careful examination of attributions. First, running a sensitivity analysis against the different attributions, allowing you to do a side-by-side analysis of the distinctive styles of attribution. Then calculating the ROAS (return on ad spend) and if each of the attribution channels is consistent, you can have more confidence in the end results. If on the other hand there is a large discrepancy, then you may need to use the second attribution tool.

Second option

The second data-driven attribution (DDA) tool within Google Analytics uses algorithms to
calculate ROAS of multiple attribution channels through what they call counterfactual
comparisons. It’s like an A/B test in that they create a survey group of so many clients
and expose half of them to the same ads, yet for the other half they withhold those same ads. The drawback is that this requires a large dataset to conduct a successful survey. Data hidden within platform activities offline such as those within Facebook or Google, make it impossible to connect these two data sources together.

Do your A/B testing

The best way to conduct the effectiveness of advertising within attributions is to use
a carefully controlled group experiment. Expose one group to your campaign marketing ads and
withhold these ads from your control group. Upon completion look at their visits and
activities and determine if your ads had any significant impact on their decision making. Results won’t be accurate unless carefully controlled.

Google now has a new ‘ghost ad’ method they are using for such scenarios which may be a nice alternative. You can find more information about this at thinkwithgoogle.com

With dynamic data, things change constantly in this age of information. Let’s take a closer look!

How does data-driven attribution work?

With the advent of algorithms, smart “machine learning” algorithms will use a comparison of all its current touchpoints, to determine what it believes is occurring and at what frequencies. It will predict what should happen based on a string of factual events that it’s monitoring, very quickly, with authority and accuracy. Known as counterfactual, this approach will determine the course that a conversion touchpoint will happen at. Assigned credit goes to the modeled attribution deemed the successor.

Learn more about the methodology of data-driven attribution through other online sources.

Specialized analytics equipment and tools

You will need large pools of data to achieve calculations with accuracy. A lot of firms or agencies are not going to have this amount of data to drive the attribution conversion models they’re hoping to have on board and in house for their benefit. These models need to be able to observe something like 500-700 events that cause a conversion within a month’s time.

Since data-driven attribution requires such a high percentage of collectible data, only the heavy hitters will be cutting the trail. And because of this, not all marketers are going to have Data-driven attribution models in their arsenal. In general, an account must have approximately six hundred conversions within a 30-day time limit for the Data-driven model to work. Fortunately, Google Analytics is available to all of us thus giving us access to this same kind of data that can be beneficial to an industry’s livelihood. If you don’t have the data, you’re not going to be able to harvest the accurate attribution models that you’re hoping to obtain.

How is data-driven reporting sustained?

Like the way freshwater feeds into a riverbed to keep it flowing, data-driven attributions require a continuous fresh supply of new and accurate data. If that data source dries up or is not able to produce the quantity required, you will not be able to produce any predictions. You need that fresh supply of data!

Find additional industry articles on our Trends & Blogs page website at Bar-None byDesign.com

Web Design – Blog Post
Author: Bradley Rupprecht – Nov. 28th, 2022