We've all had moments where we're looking at our content analytics and desired user outcomes and thought: "So which piece of content is actually driving user decisions...?"
The answer of course is easier said than done. Which is where attribution modelling comes into play.
What is attribution modelling?
Attribution modelling is the act of creating a process in order to determine specific activity contribution with regard to a user's actions.
For example, if your company produces an ad, a blog, and a whitepaper - you want to know how much each piece influences a user's decision making process.
If the user interacts with the ad, then the blog and then the whitepaper before taking a desired action such as making a purchase - what contribution did each of those pieces of content make in the overall purchase decision?
The answer is - it depends what attribution model you use.
Common Attribution Model Types
Attribution models come in many shapes and sizes. Many people debate the effectiveness of each model, as each model has its pros and cons depending on your individual situation.
If you use Google Analytics, you are likely familiar with the following models:
Last Non-Direct Click: The last non-direct touchpoint a user interacted with.
Last Google Ads Click: The last Google Ad a user interacted with.
First Interaction: The first touchpoint a user interacted with.
Last Interaction: Attribution will go to the last touchpoint a user interacted with.
Linear: All touchpoints a user interacted with receive equal attribution.
Time Decay: The most recent touchpoints a user interacted with receive higher attribution. Attribution decreases the older the touchpoint.
Position Based: The first and last touchpoints a user interacted with receive 40% attribution each, and all touchpoints between them receive the remaining 20% split equally.
Custom: You decide how attribution value is assigned to each touchpoint.
Data-Driven Attribution: Google uses a complex algorithm called the Shapley Value solution to assign incremental credit to various marketing activities along a user's journey. (This feature is only available on Google Analytics 360)
Google is updating its analytics software continuously, but often times change is quite slow. As of writing this, Google has Cross Device Reporting in a BETA release for advertising features to help with more accurate attribution.
If you use Facebook, you are likely familiar with the following model:
Data-Driven Attribution: Facebook uses a form of A/B testing to determine incremental attribution value. By using a control and multiple tests, Facebook calculates which customer journey performs best by comparison.
Essentially, Facebook takes all of the decision making out of your hands, you simply need to keep developing strategies and testing them to continuously improve your marketing. Facebook will keep picking the winning strategy.
If you use Adobe or other services, you have likely seen variations of what are listed above. Simple changes to the curve of the attribution, but not any more insightful.
A/B Testing is the act of comparing 2 more more variations of a single element in your marketing. It is a core principle of optimizing your marketing efforts.
A/B Testing relies on attribution to determine which funnel/customer journey is performing best out of the test. You use these results to optimize your funnel/customer journey to better serve your customers and ultimately improve the impact of your marketing and drive greater user decisions.
Issues With Attribution Modelling
Suffice it to say, there is no one model that can give 100% accuracy when it comes to attribution. The reason being, we have no tangible way to determine the weight of each piece of content a user came into contact with, nor do we have complete access to a user's interaction history.
Adding to this are activities such as:
- Public relations
- Other earned media
- Word of mouth/dark social
- OOH advertising
- Previous campaigns
- Various branding activities
And much more...
It is nearly impossible (unless each consumer specifically tells you) to determine which marketing touchpoint contributed what to the overall user decision.
For example, we may be testing a new series of blog topics to determine which one drives registrations on a website. Before and during this test we have been running a PR campaign, OOH advertising, and digital advertising. Who is to say the PR campaign didn't drive the most value, and the blogs simply acted as a CTA as they were the first touchpoint for the user after searching out the company?
Remember, this is a very simplistic example. In reality, even a small company's marketing efforts will be enormously complex. Being able to assign perfectly accurate attribution is impossible.
Which attribution model should I use?
It depends. It really depends on what your business goals and objectives are. You may wish to use linear attribution on one campaign and custom attribution on another.
The key to success comes in the form of continuous testing and optimization. Both Google and Facebook have tools that assist in this process a great deal (Data-Driven Attribution). However, they can be quite expensive or difficult to use.
You can do it yourself to a degree through A/B Testing to test specific campaign elements. This process can take a long time, but it is worth it to optimize your operations.
The long and the short of it is; you need to test constantly. Not using an attribution model or using one at face value without any thought behind it or testing to confirm its accuracy is a waste of time and money and will impede performance.