How to Read Attribution Reports in Google Analytics

Using Google Analytics for Attribution Modeling — Part 2

Hello everyone, welcome back to part 2 of our comprehensive guide on attribution analysis in Google Analytics.

Last week, we introduced the series and offered you a few core concepts to get started.

If you missed it, here is the link to help you catch up.

A Comprehensive Guide on Attribution Analysis in Google Analytics

This week, the real fun begins.

We are going to start going through the reports under the “Multi-Channel Funnels” and the “Attribution” section of Google Analytics and show you how you can best conduct analyses using those reports (overview of the report list presented in the graphic below).

At the end of this article, I want you to be extremely familiar with the reports presented today and comfortable doing analysis independently through them — as we will gradually help you integrate those reports into your reporting workflow in the upcoming weeks.

Let’s begin.

The Multi-Channel Funnels section gives you an overview of your customer behaviors across different channel/source/medium.

We are going to start from the top to bottom based on the navigation of Google Analytics and provide you with an in-depth view of each of the reports in this section.

Overview — Giving you a basic idea of how your channels are interacting in conversions

In the graphic above, I have divided the entire overview section into four major components, now let’s review what each of those parts means.

Section 1: Setting

The first part is the “setting section”, which dictates how you want to view your multi-funnel conversions.

There are three options here:

1. Conversions: Here you can select which conversion you want to analyze. I would recommend focusing on one or a few at a time.
2. Type: Here you can select between All and Google Ads. Unless you are analyzing the performance of your Google Ads campaigns, selecting all is the way to go here.
3. Lookback window: As explained above, this is the time you want Google to start collecting data related to this conversion. The default 30 is fine, and the variation of this really depends on the average sale cycle of your product/service — I would recommend changing around the number and see which one works the best for you.

This section will reoccur repeatedly throughout this guide, so we will always refer back to this in the future.

Section 2: Overview Graph

Nothing is special here so I am going to glance through it.

Just like the “conversion analysis” section of Google Analytics, this shows you the total number of conversions you have for the period that you have selected, depending on which conversions you selected specifically in section 1.

Section 3: Detailed Overview

This is where things start getting interesting. Based on your Google Analytics setting, you might see two or four numbers here, depending on whether you have Google Adwords integrated with your Google Analytics.

The first number is always “total conversions”, which shows you the total number of conversions of your website, adjusted by the filter in section 1.

Then, if you don’t have Google Ads integrated, you will see one number, which is Assisted Conversions, which is the number of conversions that are “assisted” by other channels, meaning that this is a multi-funnel conversion.

If you have Google Ads integrated, this “assisted conversion” is instead three numbers:

1. Click Assisted Conversions: This is basically “assisted conversions” if you have not integrated with Adwords, describing all of your conversions that have a conversion path more than 1 one visit.
2. Rich Media Assisted Conversions: This number describes all conversions in which the users have viewed a rich media on the Google Ad Platform along the conversion path. Rich media include video views on Youtube Ads, and interactive flash or video ads on the Google Display Network.
3. Impression Assisted Conversions: This number describes all conversions in which the users have had an impression on one of the ads in the Google Ads network along the conversion path.

By clicking on each of the numbers, you will be able to change the time graph in section 2 to any of those above and analyze them in detail.

Section 4: Multi-Channel Conversion Visualizer

Now we have a brief idea of the assisted conversions, it is now time to see how each of the specific channels contribute to your conversion, and how they overlap — here comes the multi-channel conversion visualizer.

This section comes into two parts: 1) the table that displays conversion from multiple channels, and 2) the Venn diagram that displays overlap between the channels.

By clicking on the checkbox next to the table, you will be able to change the Venn diagram to examine the path overlap between channels.

Overall, the more each of the channels overlap (appear together) in a conversion path, the more you should focus on the synergy between those channels.

For example, here is the most recent Venn diagram for Google merchandise store.

As you see, the overlap between organic search and referral are significant.

From this data, we can identify a synergy between referral channel and organic search channel — meaning that focus on both SEO and PR will have a very significant effect on the conversion rate of the website.

Assisted Conversions — Help you identify the “unsung hero” of conversions

The “Assisted Conversion” tab is perhaps one of the most useful section in this article because it helps you identify the “true” conversion value of a channel.

Section 1: Setting

This is the same as before, so we are going to skip it. An additional interaction option is available, which is covered above.

Section 2: Graph

The graph section contains a graph of your total assisted conversions for the time period you have selected, and provide you with a couple of metrics for you to better understand your assisted conversions:

• Assisted Conversions: The total number of conversions that have at least one assisted channel.
• Assisted Conversion Value: The total dollar (or whatever currently your Google Analytics is) value of the conversions that are considered “assisted”.
• Last Click or Direct Conversions: This metric may sound confusing, but it is simply the total number of conversions of your website based on the filter you have selected in section one.
• Assisted / Direct Conversions: This is the percent of assisted conversions compare to the total number of conversions. The higher this number is, the more I recommend you look into attribution modeling for your website — as you are missing a LOT of assisted contribution from channels in your analysis. When broken down by channel, this number carries a different meaning — it is an illustration of whether a certain channel is best for delivering information (if this number is high) to users or making them convert (if it is low).

On the right side of the graph, you can actually switch between three modes: Day of Conversion (display number of conversions by day), Days before conversion (displaying distribution of the number of days it takes for a conversion path to convert), and Path Position (displaying the distribution of the length of the path in terms of number of visit nodes).

Because we will cover Days before conversion and path position intensively very soon, we are just going to stop here and acknowledge their existence for now.

Section 3: Data Table

Now let’s move onto the data table, which displays the datapoints introduced above by different channels.

Usually, you will see the “direct” channel sitting at the top of your list, and I am going to ask you to ignore that.

It is not that the “direct” channel is not an important part of your conversion journey, is just that we cannot tell ANYTHING about your users’ experience if they come through the direct channel — therefore we cannot gain insight from analyzing it.

Let’ use Google Merchandise Store’s Data to show you how a sample analysis can be conducted.

First of all, we want to identify the channel/s that has the most assisted conversions, in this case, referral.

From the table, we can see that even though organic search resulted in more direct or last click conversions, referral assisted more conversions than organic search.

This discovery further validated our finding in the “overview” section, that the referral channel and the search channel works very well together in delivering conversion value for Google merchandise store.

Secondly, we want to identify channels that have high assisted / last click ratio — and that’s the paid search and display channels.

From the data here, we can conclude that those two channels are best served as the “intermediary” channels, deliver information about your product to your users, instead of forcing them to convert if they come from those channels.

Finally, let’s look at the channels that have the lowest assisted/ last click ratio — and that’s the search channel.

What this means is that search is a very good channel to make your customers convert in the end when they are ready to purchase. If I suspect correctly a lot of those searches come from branded keyword such as “Google Merchandise Store”.

Here, merchandise store may consider some ads on branded keyword advertising discount to facilitate the final purchase behavior of its users and increase its conversion rate.

Top Conversion Paths — Detailed Analysis of popular paths

Now let’s move onto the “top conversion path” section of the multi-funnel analysis feature.

This section gives you a more detailed view of all of your conversion paths and shows you which one is currently the most popular one by your users.

However, one major problem I have identified in this section is that “direct” channel is really stealing the show here.

Let’s take Google Merchandise store for example.

All of the top conversion paths are essentially one single funnel + few direct visits.

This is troublesome because I usually consider those conversion paths not real assisted conversion paths in the first place since users only obtained your information from one measurable source.

Therefore, I will show you a trick here to remove the influence of direct channels in our analysis.

What are we going to do is to remove all conversion paths that contain “direct” as a node.

In the data table for the top conversion path section, select “advanced”, and type in a filter with the information displayed below

Click “apply” at the bottom of the filter, and now you will get a new set of data (much smaller set) without any direct channels in it.

Notice here that the conversion number and conversion value will be greatly deflated since we removed ALL conversion paths that have “direct’ as a node in them.

But since the removal of “direct” is systematic, the distribution of those channels should remain the same or at least a very good proxy of your top conversion paths.

From the table above, we once again have seen the interaction between organic search and direct as one of the primary driver of conversions in Google merchandise store.

Furthermore, we also can see that paid search and organic search also has some synergy in terms of conversion — which is not surprising since they are both searches.

Time Lag and Path Length — Additional Attribute Analysis of your Conversion Paths

Now let’s briefly go over the remaining two sections under the “multi-channel funnels” feature.

The time lag section shows you the distribution of the time length of your conversions, whereas the path length section shows you the distribution of the node length of your conversions.

In both sections, which are laid out almost identically, you can see the number of conversions of each of the time lag/path length, and their contributions to both raw conversions, and conversion value.

One thing I found to be most interesting to analyze in those two sections are the conversions/ value ratio.

Let’s take the data presented above as an example.

In Google Merchandise store, over 63% of the conversions have a delay of 0 days (direct conversion), but those conversions only contributed to 44% of the revenue of the company.

For all conversion paths longer than 0 days, this ratio is almost all reversed, indicating that the conversion paths that are longer actually contributed more value to the business than direct conversions.

Now let’s look at path length, in which the path length between 1–2 has a worse conversion/value ratio compare with paths that are longer than 2.

Does this mean that direct conversion is doomed? Not really, since over 50% of the conversions are direct conversions, and contributing almost 50% of the total value to the company.

However, it does mean that we, as marketers, should not only focus on getting customers to convert right away, but instead creating a more drawn-out journey so they can convert (perhaps again) in the future — only with this we can capture the remaining 50% of the value.

Now let’s introduce attribution modeling into the mix, and go over the “Model Comparison Tool” in Google Analytics

Now let’s go over the model comparison tool in Google Analytics.

The setting section here is not much different from that of the multi-funnel comparison section, however, now you have a new role of options — to choose which model you want to use for your analysis.

Here, you can choose up to three models to compare, and I would highly recommend picking at least two since a lot of insights we can gain here are via comparison of models.

Your selection of models will impact the layout of the data table section below, which will look somewhat like the graphic above.

There are a few choices you can make here in regards to what you want to display on your data table.

First of all, you can add a secondary dimension for a more in-depth analysis of your attribution models.

Secondary dimensions may include but are not limited to the location of the users, whether they are new users, and so on.

Here, we are not going to cover secondary dimension analysis in detail since it will make the analysis a lot more overwhelming if it is not already is — but feel free to play with it yourself after you get more comfortable with the section.

Secondly, you can choose three metric combinations to display for each of the models you have selected.

I almost always recommend the final option, which is conversions & value.

This is because CPA and ROAS are only, by default, calculatable for channels that are related to Google Ads, making those metrics not so meaningful since you cannot compare it across different channels.

Finally, you can choose between two options when comparing across models.

Here, either option are reasonable choices and will provide you with slightly different insights, so I recommend do a separate analysis for each of those options.

Now let’s go into the report and look at what insights we can gain from comparing different attribution models.

Let’s first look at the “first interaction” model, which attributes all conversion credits to the first visit node the user visits.

Here, ignoring direct channel, referral and organic search reigned top as the two best channels in bringing the first impressions for the conversion paths. However, we do realize, from comparing conversion values, users that first came through organic search delivered a lot more value than those first came through referral — illustrating the importance of SEO for the first impression.

Then, let’s look at the last non-direct click model.

First thing, you will see, is that even though the number of “direct” conversions has decreased, they still exist!

Wait, I know what you are thinking, you thought this removes all direct hit in this model?

This question troubled me for a good while until I realized that the remaining “direct” conversion ONLY have “direct” as a channel in its conversion path — everything else has been moved to the “real” conversion channel.

From Google Merchandise Store’s Data, we can see that the number of referral conversions has shot up drastically when we do not take “direct” conversion into account as the last conversion.

It has shown us that “referral” is actually one of the unsung heroes when comes to conversion contribution, with its importance drastically undervalued by the default last interaction model.

Finally, let’s look at the position-based attribution model, which attributes the very first and very last interaction of the conversion equally while giving much less credit to the middle channels.

From this model, we see that “direct” channel shot up significantly as one of the most important attribution channels, most likely due to its heavy presence throughout most of the conversion paths, and its importance as the final channel people visit before conversion.

However, because “direct” channel is very hard to analyze, we can’t really tell much more than this.

Alas, only if we can create a custom model in that is position based, but do not take into account the “direct” channel as first or last click/interaction (hint: look at the image below).

Hey, look at the graph below, we created a position based model without counting direct click!

From the data, we reaffirmed our belief that referral and organic search are two of the most important channels for Google Merchandise store.

The position based model gave even more credit to referral and organic search compare with the last non-direct click model above!

Want to know how I did it? Let’s take a break and come back to it next week!

Next time, we are going to cover two major topics about attribution analysis in Google Analytics

1. How to setup custom attribution models so you can control your analysis at your fingertips.
2. How can we convert everything we learned today and next week into real, concrete analytics workflow in your business.

See you next time!

To receive more contents like this, please subscribe to our newsletter below.

You can also join our facebook community via the following link to get more updates and discussions around the topic of digital marketing and receive exclusive contents from us.