Attribution modeling is no longer a time-consuming, expensive process. If you are reviewing online marketing performance by campaign or channel – and you should be – it’s time to start consistently using attribution models to make well-informed marketing and investment decisions.
First, let’s take a step back and define attribution modeling. Simply put, an attribution model is the set of rules used to assign credit for conversions to touchpoints in a visitor’s journey. More frequently, we are seeing that a visitor’s experience spans multiple site interactions and it is important to consider everything that ultimately drove their conversion.
Google Analytics offers several out-of-the-box attribution models. In order to compare the results, I will use a single shopping experience as an example.
Consider a woman who has been shopping for a new watch for several months. At some point in her research, she clicked through on a display advertisement that took her to a new website: watch-you-see.com. After another week of research, the woman settled on a specific model and she does a Google search to find the best price. A click through on a paid search advertisement results in her second session to watch-you-see.com. Two days later, she receives an email from watch-you-see.com that sends her to her third session on the site. The next day, she goes directly to watch-you-see.com and purchases the watch for $100.
Single touch models assign full credit to one touchpoint – in Google Analytics, this includes last interaction, last non-direct click, last Google Ads click and first interaction models.
The last interaction model assigns full credit to the last touchpoint in her experience. In this case, the entire transaction and $100 value would be attributed to the direct channel as a result of her fourth and final session that the transaction occurred in.
The last non-direct model ignores the final direct session and attributes the entire transaction and $100 value to the email channel – the session that occurred before her final transaction. In this model, the only transactions attributed to direct would be those that had no other sessions driven by another channel. This is the most common model used and it is what populates the transactions and revenue in Google’s out of box reporting.
Last Google Ads Click
The last Google Ads click model assigns the conversion to the last (and only in this case) Google Ads click. The paid search channel would receive 100 percent of the credit.
The first interaction model would assign the conversion to the marketing channel that drove the very first session on the site – in this case, 100 percent of the credit would go to the display channel.
Google offers three multiple-touch models out of the box: Linear, time decay and position-based. All multiple-touch models split the credit across multiple touch points in the visitor’s experience.
Using the linear model, the order and value are split across all four channels evenly: Display, paid search, email and direct would each get credit for 0.25 transactions and $25 in sales.
With the time decay attribution model, channels that drove the sessions closest to the conversion receive more credit than those that drove earlier sessions. The half-life for this attribution model is 7 days. In this case, paid search, direct and email sessions all occurred within seven days of the transaction and would receive 1x credit. The display session that occurred 9 days prior to the transaction would receive 0.5x credit. One transaction divided by 3.5x = 0.286, so display would be attributed 0.143 of the order and $14.30 in value and paid search, direct and email would each receive 0.286 of the order and $28.60 in value.
The position-based model credits the first and last touchpoints the most. The default attribution is 40 percent to the first and last interactions and 20 percent spread out across the remaining touchpoints. In our example, direct and display would each receive credit 0.40 transactions and $40 in value, while paid search and email would split the remaining credit resulting in 0.10 orders and $10 in value each.
Customizing Attribution Models
Google Analytics allows us to copy some of these models and make some basic changes to save for later use.
Lookback Window: The lookback window is the period between an interaction and a purchase. The default Lookback Window is 30 days but can be changed to anything between 0 and 90 days.
Interaction Type: The Interaction Type default is Click, but this can be changed to Impression (saw an ad but did not click), direct visit (visited the website and converted immediately after seeing an ad) and rich media (users who convert after interacting with rich media, such as YouTube).
Credit Adjustment by Engagement Type: Using the credit adjustment by engagement type, instead of distributing credit evenly across channels, you can distribute it based on time spent on the site or scroll depth during a visit.
Custom Credit Rules: The custom credit rules allow you to determine the exact distribution of credit by things like position path or a number of other filters.
Decay Half-Life: You can adjust the half-life to a shorter or longer time period based on your typical user experience with the time decay model.
Google Analytics 360 – Data-Driven Attribution Model
The data-driven attribution model (DDA) is a huge perk exclusively available for Google Analytics 360 clients. Unlike any of the models we reviewed above, the DDA model uses your actual data to generate a custom model that assigns credit to the touchpoints throughout a visitor’s journey. This model uses not only Google Analytics data, but any information available through accounts that are linked to it: Google Ads, Google Display Network, and Campaign Manager and uses an algorithm to generate the attribution rules.
What Attribution Model Should You Use?
If you’re a Google Analytics 360 client, the data-driven attribution model is the most actionable option available. After activating this feature, you’ll quickly have data that attributes performance based on user activity over the previous 90 days.
If you don’t have access to that model, you should consider the goal of your marketing campaign while reviewing performance. If you’re attempting to increase brand awareness in order to find new customers, use the First Interaction model to see what channels are more likely to drive visitors that ultimately come back and convert. If your goal is to increase conversions with existing customers, use a multi-touch model for a better understanding of what drives them back to the site vs what makes them convert. The Attribution Comparison Tool in Google Analytics allows you to compare the results of multiple out of box and custom attribution models to ensure that you are not undervaluing a marketing channel or campaign.