AnalyticsRefining Your Attribution Model to Gain Clarity on Media Performance

Refining Your Attribution Model to Gain Clarity on Media Performance

To begin quantifying attribution, start by determining the model by which you’re most comfortable attributing success and understanding the factors that impact the value of each touch. Once values are agreed upon, you’re off to the races!

You’ve started to look at all your media channels together and you’ve begun to review the touch-points along the way to a conversion event. You’re beginning to see what impact display is having on search performance or how email drives initial site visits, but display ads are driving the end conversion. You’re even beginning to see which particular sites and placements are showing more influence.

But how do you quantify that attribution? How do you create reliable framework for determining efficiency? And most importantly, how do you effectively compare performance across channels for the purpose of optimizing spend allocation?

In order to begin to quantify attribution, you must first determine the model by which you are most comfortable attributing success. There are four generally accepted attribution frameworks:

  • Last Click Attribution: 100 percent of the conversion event value to the placement delivering the last click. While this is the most commonly employed model, it’s arguably the most flawed.
  • First Click Attribution: 100 percent of the conversion event value to the placement delivering the first click. Basically the same as last-click, but delivering credit for the initial engagement and not the last.
  • Equal Weighting Attribution: Every touch-point leading to the conversion gets an equal share of the credit. We’ll call this the “Good Parent” model, because while we love each of our kids equally, should we do the same with our media channels?
  • Custom Credit Attribution: This model takes multiple factors into account and determines the proper credit share to give to each touch-point. This model becomes the basis for most algorithmic-based models and hones in on the unique levers, which drive user action forward.

In order to get a customized model, we must first understand the factors that impact the value of each touch. These factors include:

  • Engagement Factor: What did the user do? Did they click an ad or simply view it? Did they watch an in-video banner? Determine your perceived value of each engagement type and rank-score them against each other.
  • Media Factor: What unit delivered goods? Was it a rich media unit? A standard banner? What size was the banner? Was it a text ad? Determine your perceived value of each media unit and rank-score them against each other.
  • Time Factor: How long was it between touch-points? How long was it from the specific touch-point to conversion event? Define the parameters around time decay to determine the impact of time on the value of each placement.

It is at this stage that attribution models get stuck. The reason is because we begin to bring in the human element through assigning values to each factor.

What’s important to realize here is that here is no “right” answer to what the assigned value is. The value must be worked out and agreed upon across all parties involved (most notably, the client) for the model to move forward. However, once values are agreed upon, you’re off to the races on attributing success across channels!

attribution-placement-model

In this example, the time factor starts out as all touches are of equal value, but touches that are 2 to 5 days prior to the conversion are 25 percent less valuable, while touches greater than 5 days out are 50 percent less valuable.

The media factor starts with a standard banner view as the baseline, but states a paid search ad is 25 percent more valuable and an oversized banner is 10 percent more valuable. The engagement factor states clicks are four times as valuable as a view. This attribution model also states that all factors are equally relevant.

Given each of these factors, touch one receives 22.1 percent of the conversion credit, whereas touch three, on the strength of it being paid search and within two days of the conversion event, receives 54.1 percent of the conversion credit.

But, what would happen if you changed your weighting? In the above example, if we change the Time Factor so that all touches that are more than 2 days prior are worth one third of those within 48 hours (going on the premise that any touch older than 2 days is “stale”), this would increase Touch Three’s factor to 60 percent, and Touch One’s Factor to 20 percent while decreasing Touch Two to just 20 percent. In turn, this will increase Touch One’s credit to 23.2 percent, while decreasing Touch Two’s credit to 22.2 percent. Touch Three would increase its credit to 54.6 percent (nearly a 10 percent lift it share).

How you weight the value of each factor will impact each touch’s credit.

For example, you may decide that a homepage takeover’s Media Factor is worth 10 times a standard banner because it is high impact creative. This will greatly skew attribution in favor of the home page banner, while diminishing the value a standard banner, regardless of when the event took place or whether a user engaged.

Good luck and happy attribution.

Resources

The 2023 B2B Superpowers Index
whitepaper | Analytics

The 2023 B2B Superpowers Index

8m
Data Analytics in Marketing
whitepaper | Analytics

Data Analytics in Marketing

10m
The Third-Party Data Deprecation Playbook
whitepaper | Digital Marketing

The Third-Party Data Deprecation Playbook

1y
Utilizing Email To Stop Fraud-eCommerce Client Fraud Case Study
whitepaper | Digital Marketing

Utilizing Email To Stop Fraud-eCommerce Client Fraud Case Study

1y