A Comprehensive Guide on Attribution Analysis in Google Analytics
Using Google Analytics for Attribution Modeling — Part 1
Update: You can find the second part of this guide in the link below :D. Thanks for reading.
How to Read Attribution Reports in Google Analytics
Using Google Analytics for Attribution Modeling — Part 2
Last week, I wrote an in-depth guide that explained in depth the core concept of attribution modeling. In the guide, I not only explained to you what attribution modeling is, but also showed you several examples of attribution models used by Google Analytics and us at Humanlytics.
However, we did not cover in detail HOW to do attribution modeling in practice in Google Analytics — a gap that will be made up thru this article.
For the next few weeks, I am going to fill in the gap by provide you with a comprehensive overview of how to analyze, setup, and conduct attribution modeling using Google Analytics.
More specifically, here are the key issues we are going to cover:
- Key concepts that you need to know about attribution modeling in Google Analytics before conducting any analysis.
- How to read multi funnel reports and attribution modeling reports in Google Analytics.
- How to setup custom attribution models in Google Analytics to better existing models or satisfy your unique needs.
- How to apply insights you have gained via attribution modeling into your business, and improve your conversion rate with those insights.
As you may have already realized, we are covering a lot in this guide — and it cannot be done in one simple article.
Therefore, we are going to divide the entire topic into three or four articles, published throughout this month, so you do not lose your insanity reading a 6000-word long article, and I do not lose my insanity writing it.
In the first week, we are going to start with a brief introduction of a few key concepts (such as conversion path, interaction type) related to attribution modeling in Google Analytics.
By going over those key concepts ahead of time, I hope that you will be prepared for our content next week, which is an in-depth dive of all multi-channel funnel and attribution modeling report in Google Analytics.
Without further ado, let’s get started.
Attribution Reporting in Google Analytics
In Google Analytics, there are actually two sections that can tell you about your attribution performance — the Multi-channel Funnels, and the Attribution section, both under the “conversion” tab in the navigation.
The “multi-channel funnels” section gives you an overview of the behaviors of your customers across different channels, and how those channels work together to make your customers convert.
On the other hand, the “attribution” section utilizes attribution modeling to help you understand the value of each of your channels as a whole.
While the “multi-channel funnels” feature technically does not use attribution model for analysis, it provides valuable information under similar veins, so we will cover this section in depth as well in this article.
Let’s start by going over a few key concepts in multi-funnel attribution and attribution modeling
A conversion path describes a series of visits user/s make to your website that ultimately resulted in the conversion of those users.
A conversion path has three primary elements: a lookback window, a conversion, and nodes represent each visit.
A lookback window describes how far back you want to trace visits of your users.
By default, Google Analytics gives you a lookback window of 30 days. This means that it will only take into consideration visits made by your users 30 days before the ultimate conversion.
Why is this window needed?
On one hand, it is of very little relevance to you whether the user visited your website half of a year ago — since they most likely forget about your website and their previous visits are unrelated to your conversion.
On the other hand, to get user visit data for all users of all times is a very computationally intensive task for Google Analytics to do, and will make the feature very slow and buggy.
Therefore, for both considerations, a lookback window is needed to reasonably limit the query so it can be delivered fast and relevant.
On the Google Analytics interface, you can choose a conversion window anywhere ranging from 0–90 days. This number defaults to 30 days when you are using the Google Analytics Multi Funnel API, and cannot be changed.
The next important element of a conversion path is the conversion itself.
Unless restricted, Google Analytics will calculate the conversion path of ALL of your conversion goals in your Google Analytics account.
This default option is not entirely helpful since it mixes all of your goals together, ranging from goals that represent the ultimate purchase to goals that represent merely an engaged session.
Therefore, you should ALWAYS analyze conversion path by an individual goal, or a group of goals — and lucky the interface allows you to do that fairly easily (more on that later).
Each of the “visit node” represents each visit your user/s conduct on your website between the beginning of your lookback window and the occurrence of the conversion.
The amount of visits in this timeframe is called a “path length”, and the time between the first visit and the last visit (conversion) is called the “time lag” of that specific conversion path.
A large portion of your visits will only have one visit (having visit length of 1), meaning that the user did not conduct any other visits to your website prior to the conversion (those visits also have a time lag of 0).
This phenomenon, however, is getting increasingly rare as users’ decision journey has become a lot more iterative in recent years — requiring multiple visits (maybe one for gathering information, one for comparison, and one for conversion) before the ultimate decision to buy.
The visit node that resulted in the final conversion is called the “last click/interaction” conversion node, whereas all visits before that ultimate visit are considered “assist” conversion nodes. Keep those terms in mind — they will come into play when we go over the reports.
In the most perfect world, we want to know, for each unique conversion path, how is a specific visit node doing in terms of engagement and impression. We may also want to know what is the specific, page by page user experience of each of those nodes, and whether those nodes resulted in any other conversions.
However, achieving such functionality will require a very high computing power and time, making it really difficult for Google Analytics to provide all of those functionalities.
Therefore, the Google Analytics only offers a limited range of dimensions for each of the node paths, including sources, channels, and mediums of the visit (detailed list below).
Conversion Paths - Dimensions & Metrics Reference | Analytics Multi-Channel Funnels Reporting API |…
The length of conversion paths in number of interactions. The value is a histogram across a range of possible values…
Finally, it is also worthy to note that within your conversion path, your user might have already accomplished the conversion objective in the middle of the path without you knowing since Google Analytics does not take into account the conversions occurred within the path during its calculation.
In most cases, this will not impact the overall validity of your result — just an interesting detail I found about the conversion path.
If you have Google Ads integrated with your Google Analytics account, you will see an additional option in some of the settings that are called “interaction type”.
This is because, with Google Ads integration, Google Analytics will also take into account any actions on the Google Ads platform as part of the conversion path, which includes 1) ad impression on a specific keyword (impression), 2) rich media view thru an ad (rich media), and 3) a click onto the website.
For those of you who are confused about the terms, a “click” interaction is basically any session visits onto your website, whether they are from Adwords or not, while “rich media” and “impression” interactions are only for Google Ads.
Now, there is a fourth interaction — direct.
Why do we want to isolate direct visits? Because they are almost impossible to attribute. When a user conducts a direct visit, they only thing we know is that they typed our url into their browser (or they disabled cookies), and nothing else.
For this reason, direct visits are almost useless to us when analyzing multi-channel funnels, so a lot of marketers end up decide to remove them from the mix for a cleaner analysis.
Here, I usually recommend unchecking the “impression” box and the “direct” but keep everything else. This is because if you select “impression”, and “direct”, your conversion path will be filled with them, diluting all of the more important “nodes” along your path.
However, if you do not have Google Ads integrated with your account, you do not have the luxury of removing “direct” clicks and visits from your analysis — but we can do it with custom attribution models — a topic to be covered later in this series.
Alright, that’s enough background information for this week.
Next week, we are going to actually dig into the report and figure out what each of the reports presented by Google Analytics can tell us about the multi-channel behavior of our customers.
Strap up and prepare for a ride, since it is going to be a long guide next week (it’s already written and almost 3k words).
See you next week o7.
To receive more contents like this, please subscribe to our newsletter below.
For more reading;