Attribution Modeling Training - Learn How to Measure Beyond Last Clicks in AdWords

Attribution and Deficiencies of Measuring AdWords Conversion Performance

Bar Graph of Attribution models in AdWords

Attribution by definition concerns who or what gets the credit for the results. In the realm of PPC and digital advertising, this is referring to what is the cause of the results such as sales or new service leads. Attribution is a difficult subject that usually produces more questions than answers. Even with the robustness of Google Analytics and its myriad attribution models, it still begets questions like “Great, I can see how much assistive value each step is providing, but it won’t show that in AdWords, so what difference does it make in day to day PPC management?”

The great news is Google has heard our cries and they have produced an answer: attribution model selection at the conversion level. For the first time, when setting up conversions in AdWords, whether by importing it as a goal from Google Analytics, or directly using an AdWords conversion code, the user can now choose what attribution model from a drop-down menu. The five attribution models available are: last click (standard option), first click, linear, time decay and position based. For the sake of simplicity, this article will refer to the value created by each campaign within an AdWords account. Let’s break down what these attribution models are and when you (or a client) should consider using the different variants.

AdWords Conversion Attribution Models

Drop down menu showing attribution models available in AdWords conversion interface

Last Click Attribution Model

Think of the last click model as similar to scoring a goal in hockey. The last player to flick the puck into the goal gets the credit, but other players may have assisted that player to make the point. This is the standard model and has been since the dawn of PPC advertising with Google AdWords. Thanks to exposure to concepts like cross-device conversions, in-store conversions and others, we realize now that attribution is much more complicated than what was the last step a customer makes.

As the default model, historically the last click attribution methodology overestimates the value of the last touch point in a customer’s purchase funnel, the funnel made up of each touch point made including a brand campaign, a service campaign, and another service campaign. By not divvying up the value of a multi-touchpoint customer funnel, it undervalues all other steps in the customer funnel as value creators. This model states “the last campaign is all that matters, all other campaigns created no value” which is most certainly not true.

Why use this model? This is the standard model, it is how the overwhelming majority of individual advertisers and agencies measure and gauge their performance and those metric trends over time. Changing the model can skew analysis and take time for adjustment as advertisers are more comfortable and familiar with this model. Also, this model is the most straightforward and thus is a perfect model for beginners who have yet to train on the more advanced models, to be covered next.

First Click Attribution Model

The first click model runs diametrically counter to the last click model as it fully attributes the value of a sale or lead to the first click or touch point in a customer funnel. The theory follows that the original campaign to bring awareness and consideration for a web user is all that matters as everything else piggy-backs off this initial campaign.

Like the last click model, the first click model again overestimates a singular touchpoint and undervalues (doesn’t count any value) to the other touchpoints. Again, the problem lies in a bias toward one piece of the puzzle to the detriment of all others. The first click may create the initial awareness and consideration, but until the user sees their funnel through to a sale or lead, the value still isn’t fully realized.

Why use this model? The first click model can be quite helpful to determine which campaigns are raising awareness and consideration for users, and is especially helpful in adding value to campaigns that normally would not receive it. Generally speaking broader, more general terms are searched first, and users then return using a branded search, remembering the name of the business earlier on in the customer funnel. By giving the campaign(s) with the more general terms the value, they are noted for their assistive performance for a change.

Linear Attribution Model

As the term linear sounds, this model is a proportional model that divvies up the value of each campaign equally. If two campaigns were involved in the customer purchase funnel, then 50% of that customer’s value will go to each campaign. If five are involved, 20% each, and so on.

The first major differentiator of this model vs. those covered so far is it does not give 100% of the value to any one touchpoint. Instead, it gives credit across the board equally. By doing so, all campaigns are credited, noting that all campaigns have a synergistic effect on creating value for the advertiser.

Why use this model? The linear model is the simplest non-singular touchpoint attribution model. This means it is the most easy to understand model that involves crediting more than one touchpoint with the value of a single customer’s action(s). The fewer the number of touchpoints, the more value for each, the greater the number of touchpoints the less value each receives.

Time Decay Attribution Model

The time decay attribution model is a more complex multi-touchpoint model. To explain, think of this model as being akin to carbon dating a fossil to determine its age: a half-life decay model is used with a radioactive isotope. Depending on how much of the fossil decays, we are able to estimate its age.

Time decay attribution uses a similar model where a touchpoint receives ½ the value that it could have received, every 7 days that precedes the final touchpoint. For example, if the first campaign touchpoint occurred 14 days ago, the second campaign touchpoint occurred 7 days ago, and the third and final campaign touchpoint occurred today, the distribution of the value would occur as follows: the final touchpoint receives 4x as much value as the first touchpoint (½ x ½ = ¼ of full value) and 2x as much as value as the second point, which in turn received 2x as much as the first touchpoint. So if a customer’s purchase is worth $700, the first touchpoint receives $100 of credit, the second touchpoint $200 of credit, and the final touchpoint $400 of credit.

The model is complex, but it touches on a concept that the further removed a touchpoint is in time from the final touchpoint of a customer funnel, the less impact it has. Brand awareness and recognition is constantly hammered into potential customers to keep it at the forefront of the mind. As time goes by, recall becomes less strong for a brand. In this way, time decay considers this phenomenon in its attribution.

Why use this model? This model is great for businesses with a short business cycle consideration. Examples could include emergency dental or plumbing services or purchases with a small price tag. In these types of businesses, a touchpoint that occurred two weeks ago may have a slight impact on getting the customer to consider a purchase, but its influence wanes exponentially as time passes.

Position Based Attribution Model

The position based attribution model is a hybrid mixture of the last click, first click and linear models rolled into one. The position based model grants 40% of the value of a customer to both the first campaign and the last campaign while the remaining 20% is distributed equally among the “middle” campaigns. This model emphasizes the importance of the first click which starts the customer funnel and the last click which completes the customer funnel as equally important components in value creation. But the model also does not discredit the smaller, yet still important stepping stones in between the first and last click for seeing the customer funnel through to the last click.

An example of a $100 purchase that involved 4 touchpoints would credit $40 to the first campaign, $40 to the last campaign, and split up the remaining $20 by giving the 2 middle campaigns $10 each. A good sports analogy can be found in football. The running streak for a football team begins when either a turnover of the ball or a reception of a kick from the other team is equivalent to the first click of a customer funnel. Without that starting point, the customer funnel does not begin. As the football team continues to make first downs and move across the field, these are the “middle” touchpoints that continue to follow the customer’s consideration and decision making process, and keeps the brand front and center in his or her mind. Finally, the touchdown or field goal is the last click, that last moment when value is ultimately created. By emphasizing value at the beginning and end while still recognizing the importance of the “in between” touchpoints, this attribution model is a good hybrid of the last click, first click and linear attribution models.

Why use this model? For many businesses with several touchpoints in their customer’s purchase decision making process, such as auto dealers and travel agents, this model provides a closer estimation of the value each campaign contributes to creating value than the standard last click model.

AdWords Attribution Modeling Tool & Conclusion

Attribution is a complex, multi-faceted conundrum that requires a lot of thought and mapping out the pros and cons for specific businesses and industries. What is the best attribution model for your business or client? One way to answer this is by testing the data itself using the attribution modeling report found under tools > conversions > attribution.

Example of an actual client’s Attribution Modeling Comparison Report

Example client attribution modeling comparison report showing differences in assignment of conversion value creation among campaign ad clicks

If this model is not helpful or difficult to interpret the results and apply them as a rule for finding the right attribution model, you can also answer the following questions:

  1. Do I care more about the first click, the last click, or other?

  2. Is time an important variable, such as short or long customer funnels?

  3. What has been the prevailing attribution model thus far? How will I reconcile comparing the previous model to a different, new model?

  4. How much time do I have to actively interpret the attribution results on an ongoing basis?

Dependent on how these questions are answered, you may find the last click model is sufficient enough. But maybe you will learn that a more robust model is needed to encapsulate each touchpoint such as a time decay or position based model.

To learn more about Google AdWord’s attribution models visit their support article here.

If you enjoyed this article, you can sign up to receive more news and helpful tips in your inbox every week!
Leave your questions or comments on this article below. We are also taking recommendations on what to write about next, so submit your how-to requests below!