In this lab attendees will learn how to unlock the more advanced capabilities available within Adobe Analytics. This lab will teach attendees to harness Adobe Sensei powers within Adobe Analytics to reduce time to insights, which in turn will increase analyst productivity. For this lab we will assume that we are analysts for an ecommerce company with a number of analysis goals.
Attendees of this lab should:
Please replace the blank in the username with the user number shown on the sticker attached to your lab machine.
Attribution IQ represents Adobe’s built in attribution models and attribution capabilities within Adobe Analytics. It supports the ability to use a number of weighted models to measure the impact that some data point has on driving an event of interest using a number of attribution models. Attribution IQ allows users to redefine how a variable can be given credit for some future event. The most common use case here is looking at how marketing drives conversions. In that case the variable would typically be Marketing Channels and the conversion event would be Orders. But realistically, the variable can be any eVar/prop and the “conversion event” can be almost any metric.
Attribution IQ creates a virtual instance of variables collected in Adobe Analytics and changes the persistence of those variables to be instance-based versus how they were initially defined. That means that if an eVar was set to Last Touch at a visitor level, Attribution IQ will ignore that setting, just for the attribution reporting, and look at just the Instances where the variable is set, versus the persistence. Imagine the following Scenario:
Using a visitor level last click eVar to capture the Campaign marketing channel would mean that Email would get 100% of the credit for the conversion because it’s the last touch.
If the Campaign marketing channel eVar was set to visitor level first touch, then Paid Search would get 100% of the credit for the conversion.
Attribution IQ supports the ability to ignore those settings and even though a variable is set as, let’s say last touch at a visitor level, it will look at each value that was set for that variable (for the reporting date range in the Analysis Workspace panel).
So, in the example above, assuming that the Campaign marketing channel was set as visitor level persistence with a last touch allocation, the initial value for the Campaign marketing channel was set to Paid Search, then it was overwritten by Display and then Display was finally overwritten by Email. So, Email received 100% credit for the conversion. Attribution IQ will ignore those rules and will look at each instance of an eVar (or any variable for that matter) being set and will include it in the attribution model (depending on the selected attribution model). This means that all of the touches in the example above are eligible to receive credit for the conversion, depending on the model selected. If a linear model is selected, Paid Search would get 33.333% of the credit, Display would get 33.333% of the credit, and Email would get 33.333% of the credit for the conversion.
Attribution IQ automatically creates a new dimension based off of the last touch marketing channel. The new dimension is called Marketing Channel and is basically a copy of Last Touch Channel but removes the confusion around the word “Last” in the name. Technically a user could use Last Touch Channel and have the same results, since Attribution IQ is looking at the setting of the variable instance, versus the persistence of the variable, regardless of how the allocation is defined.
OK, let’s start using Attribution IQ.
The director of marketing has asked us to measure the influence that marketing has on driving online orders. We can use Attribution IQ to help us with this task.
In this exercise we’ll learn how to use the Attribution Panel to build a panel with a number of visualizations tied to attributing credit to specific touches, for their influence on driving an event of interest. The attribution panel asks a user what they want to count as a touch event, what they want to count as a conversion event, and what models they want applied to assign credit. It also asks if they want the lookback window to be constrained to a visit or across visits at a visitor level, which will then have a lookback window constrained by the panel reporting window.
Start with a blank workspace project.
Within the blank project, drag the Attribution panel into the workspace project.
The Attribution panel will ask what to use as the success metric (this is the conversion event), channel (this is whatever you want to give credit to for driving the success metric), the models to be applied, the lookback window, and the reporting period.
Let’s assume we want to understand the influence of marketing at a channel level.
Custom: this is a Starter/Player/Closer model with customizable weights. Let’s give the following weights:
Let’s review the contents of the completed Attribution IQ panel.
As you can see, the Attribution IQ panel provides a number of ways to analyze and understand the influence that one thing had on driving something else. In this case, we analyzed the influence that Marketing Channels had on driving online orders, but again, the same analysis could easily be run to understand the influence that some other thing had on driving a different event.
Typically the Attribution IQ panel is used to analyze the relationships between the interactions that are helping to drive an event of interest so that an organization can choose which attribution model makes the most sense for their business moving forward. Studying the relationships between the touches and the event of interest will allow an organization to understand if there are primarily one interaction before a conversion, or if there are frequently multiple interactions/touches on a path to a conversion. If there is primarily only one touch prior to a conversion, a first or last touch model can typically be selected. If there are often many touches on a path to a conversion, a model that gives credit to all the touches on the path may be a better option.
Once a model has been selected as the model that the organization will use for measurement purposes, this model can be applied to a metric directly from within a table with that metric in it. Let’s assume that based on the research that was performed in the Attribution panel, a linear model has been selected as the model that makes the most sense for the business. In this exercise we’ll be adding an Attribution IQ model to a metric within a table for analysis purposes.
Let’s start with a new Blank Panel with a table in it.
Add the Marketing Channel dimension to the table.
Now add the Visits metric to the table.
Now add the Online Orders metric to the table.
Now if we mouse over the Online Orders metric in the table and click on the gear, we get the configuration screen for that column in the table. At the bottom of the configuration screen is a checkbox where there’s the ability to choose “Use non-default attribution model”. Select that checkbox. This will allow us to specify the attribution model we want applied to the Online Orders metric in that column.
By selecting the “Use non-default attribution model” checkbox, a pop-up appears that asks what attribution model to apply to this metric. In this case, let’s select Linear and click Apply.
The Marketing Channel table now has the Visits metric as well as the Online Orders metric with a Linear attribution model applied. From here we can compare how our marketing channels are performing in relation to other KPI’s. As we’ll learn in a future exercise, we can also create calculated metrics that use Attribution IQ models as part of their calculations.
As discussed several times in this lesson, almost any data point collected in Adobe Analytics can take advantage of the Attribution IQ capabilities. Let’s say that management is interested in understanding the influence that pages have on driving downloads on the site. We’ve decided that Attribution IQ could help answer that question easily.
Let’s add a Freeform Table visualization to the panel.
Add the Page dimension to the table.
Add the Visits metric to the table.
Add the Downloads metric to the table.
You’ll notice that the table shows 0 downloads for each page. This is because the download event hasn’t occurred on any of the pages that can be seen in the table. But, if we apply an attribution model that looks at the touches leading up to an event, we’ll be able to see the top pages that visitors view prior to a Download event. This could help us understand what pages help or assist in driving a Download event. Mouse-over the Downloads metric and select the gear to open the Column Configuration screen.
Within the Column Configuration screen, select the “Use non-default attribution model” checkbox.
In the attribution pop-up, let’s select a Time Decay model with a 1 hour half-life and constrain it to a Visit lookback window. This will only give credit to pages that were viewed within the last 60 minutes prior to a Download event, within a visit. The Time Decay model will give incrementally less credit to each page prior to the download event, with the most credit going to the page just prior to the download event.
Click Apply to apply the Time Decay model to the Downloads metric and the table should look like below. This now allows us to see the top pages that helped to drive downloads in a visit.
In the last exercise we modified the Downloads metric to use the Time Decay attribution model with a 1 hour lookback window within a Visit. In that exercise we created the table above. Let’s say that management now wants to know what percentage of traffic each page is driving to download, still using the Time Decay attribution model. This would require a calculated metric. In this exercise we’ll build a calculated metric that calculates the percentage of traffic being driven to download, from each page, using the same Time Decay model we built in the last exercise.
In the left side-rail, under Components, click on the + sign to the right of Metrics to bring up the Calculated Metric Builder.
The Calculated Metric Builder will be presented where you can start building a calculated metric.
From Components in the left side-rail, drag the Downloads metric into the drop-zone.
From Components in the left side-rail, drag the Visits metric and drop it in the drop-zone beneath the Downloads metric. This will create the formula of Downloads divided by Visits.
Now we want to apply the Time Decay model to the Downloads metric in this formula. To do that we need to click on the gear icon to the far right of the Download metric.
Select the Use non-default attribution model check-box, which brings up the Column Attribution Model pop-up window.
From within the Column Attribution Model pop-up window, select the Time Decay model with a 1 hour Half-life and a Visit lookback window.
The Calculated Metric Builder should now look like the below. You can see that the Downloads metric has the Time Decay model applied, with a 1 hour Half-life and constrained to a Visit.
Change the format of the metric to be Percent, give it a name of Visit Download Rate, and save the metric.
Now we want to add that newly created metric to the table.
Drag the Visit Download Rate calculated metric over to the Page table.
In this lesson we learned how to use the new Attribution IQ capabilities to quickly build an Attribution IQ panel, apply Attribution IQ models to metrics in any table, apply Attribution IQ models to non-marketing events, as well as harness the power of Attribution IQ from calculated metrics. These capabilities are very powerful in that they allow analysts to learn what model makes the most sense for the analysis goal of the moment, as well as apply the models in a way that allows them to most effectively understand what’s working and what’s not.
Cohort 2.0 significantly improves the capabilities available within the Cohort Table visualization. It enabled the ability to apply segments to both the inclusion as well as return metrics within the visualization. It also enables the ability to specify that the cohort table visualize Churn versus retention, or that the calculation is rolling across the date range. Latency is a new capability that allows an analyst to understand what occurred prior to an event of interest. Lastly, the Cohort Table now allows for dimensions to be applied against the inclusion metric to understand their future influence on the return criteria. In this lesson we’ll focus on learning all of these new capabilities.
In this exercise, let’s assume that we’ve been asked to determine if visitors come back to the site in the days after they place an order. We can technically use a standard cohort table to measure this.
Add a new panel by clicking on the + below the bottom panel in the project.
From within the new panel you can select the type of visualization you’d like to start with. Let’s select the Cohort Table visualization.
From within the Cohort Table configuration screen, we have a number of options. We will go over all of these as part of this lesson. Let’s just start with a basic Cohort Table and build from there.
Let’s assume that we’re interested in analyzing behaviors of visitors in the last week. Change the date range of the panel to Last Week.
Let’s also set the granularity of the cohort table to Day. This will allow us to measure the day-over-day retention of visitors after they place an online order.
Add the Online Orders metric as the Inclusion Criteria and the Visits metric as the Return Criteria.
This will build a Cohort Table that shows, day over day, when visitors have a return visit, after they have an online order event. When we click Build, it will build the Cohort Table represented below.
We just built a standard Cohort Table that doesn’t show of any of the capabilities available in Cohort 2.0. Let’s edit this Cohort Table to take advantage of the Cohort 2.0 functionality. Click on the pen to the right of the Inclusion and Return details for the Cohort Table to bring up the Cohort Table configuration screen again.
Let’s assume that management has asked how often a visitor comes back in the days following an online order and either starts the checkout process or completes the checkout process and places another online order. That is a perfect use case to take advantage of the segmentation capabilities released with Cohort 2.0. All we need to do is segment the Return Criteria by visitors who touched any of the checkout pages. In the Components menu in the left side-rail, click on the arrow to the right of the Page dimension to view the Page dimension items. Multi-select the following pages:
Drag those selected pages and drop them into the Return Criteria Segment drop-zone in the Cohort Table configuration screen.
Clicking Build will rerun the cohort analysis to look at visitors who had an Online Order event and then returned in the next 6 days and either started the checkout process or completed the checkout process.
That’s interesting. A fair amount of visitors are coming back and at least starting the checkout process in the days following an online order. Our management also finds this very interesting and now wants to know how many of those visitors are coming back and either starting a checkout or completing an order every day after an Online Order event. We can take advantage of the new Rolling Calculation capability released with Cohort 2.0. Click the pen to edit the Cohort Table configuration again and select the Rolling Calculation checkbox.
Click Build to build out the Cohort Table with the criteria that a visitor had to have placed an Online Order and then either started or completed the checkout process every day after that initial Online Order event. This means the count in the +2 Days column is a count of visitors who had an Online Order event on Day 0, either started or completed the checkout process on +1 Days AND either started or completed the checkout process on +2 Days.
Again, an interesting view of the data. There are quite a few visitors following this behavior day over day. Management wants to see the inverse of this. They want to see how many people don’t follow that behavior. How many visitors do we lose (as it relates to this behavior) day over day. The churn capability that was released with Cohort 2.0 can be used to answer that question.
In this exercise we’ll cover how to use the Churn Analysis capabilities that were released as part of Cohort 2.0. Following along with the analysis we performed in exercise 2.1, let’s assume that management wants to see the inverse. They want to see how many people don’t follow that behavior. How many visitors do we lose (as it relates to this behavior) day over day. The Churn capability that was released with Cohort 2.0 can be used to answer that question.
Click the pen to edit the Cohort Table configuration again and select the Churn radio button to change the Cohort Table Type from Retention to Churn.
Click Build to build the Churn Cohort Table
This provides an interesting view into the number of visitors we lose, as it relates to repeat business.
In this exercise we’ll cover using the advanced feature of Latency Analysis within the Cohort Table. The Latency capability allows us to analyze what happened before an event of interest as well as after. It adds negative Days (- Days) to the Cohort table so you can see what occurred prior to an event of interest. This ties nicely to the analysis that we’ve been doing so far. Using this capability we can analyze the behaviors of Visitors before they have an Online Order event.
Click on the pen to edit the Cohort Table showing the churn analysis from the last exercise.
Within the Cohort Table configuration screen change the Cohort Table Type back to Retention from Churn.
Remove the Return Criteria Segment of Visitors that touched any of the checkout pages.
Select the Advanced checkbox and ensure that the Latency Table radio button is selected.
Based on how the Cohort Table is currently configured, it will build a Latency Table that shows Visitors who had an Online Order on Day 0 and how many of them had Visits (the Return Criteria)in the days before as well as after that Online Order event.
Click on Build.
As you can see, there are Visitors that visit prior to an Online Order event as well as Visitors that visit after an Online Order event. In viewing the output, it looks like close to 20% of the Visitors that have an Online Order event visit the site 1 day prior. That’s interesting information from a marketing perspective.
Now let’s dig a bit deeper to see if they’re interacting with Paid Search in the days leading up to that Online Order event. Click on the pen to edit the Cohort Table configuration.
Drag the Paid Search element from the Marketing Channel dimension over from the Components menu on the left side-rail and drop it into the Segment drop-zone for the Return Criteria. Because we have the Cohort Table configured as a Latency Table, this will result in us being able to see the Visitors that interacted with Paid Search in the days prior as well as after an Online Order event.
Click Build to build the Latency Table.
From this view we can easily see that at least some Visitors that have an Online Order event interact with Paid Search in the days leading up to that Online Order. We can also see that some Visitors also interact with Paid Search in the days after an Online Order event.
This is interesting. But it only shows this for Paid Search. Let’s say we now want to understand how far in advance Visitors are interacting with various Marketing Channels prior to a conversion. We could technically do this one Marketing Channel at a time within the Latency Table. But this is a job better suited for the Custom Dimension Cohort capabilities that were released with Cohort 2.0. Let’s use them here.
Custom Dimensions allows us to replace the time-based dimension for the Inclusion Criteria with any other dimension that makes sense. This will allow us to breakout performance by dimension elements in a dimension over time. What we’re trying to see here is a Cohort Table that breaks out visits by Marketing Channel with a view into the Online Orders that occur in the days after.
Click the pen icon to get back to the Cohort Table configuration screen.
Under Advanced, select the Custom Dimension Cohort radio button. This will allow to drag a dimension into Custom Dimension Cohort drop-zone
In order to do this analysis we need to swap the Inclusion and Return Criteria. We want Visits as the Inclusion Criteria and Online Orders as the Return Criteria.
Now drag the Marketing Channels dimension from Components in the left-rail and drop it into the Custom Dimension Cohort drop-zone.
This configuration will create a Cohort Table that breaks out visitors by Marketing Channel and allow us to understand the number of Online Orders that occurred for those visitors in the subsequent days after they touched a particular Marketing Channel.
As you can see, the capabilities released with Cohort 2.0 are very powerful. They allow analysts to dig much deeper into what Visitors are doing both before as well as after an event of interest. Not covered as part of this lesson is the ability to build segments based on the visitors that have an event of interest followed by or proceeded by another event of interest. But this is a capability that the Cohort Table supports.
In this lesson we’ll be covering the Fallout visualization and some of the more advanced capabilities it supports. The Fallout visualization can be used to measure how visitors navigate through a process on the website or within a mobile app. It can be used to measure that process within a visit or across visits at a visitor level. It can also be segmented by anything you know about your visitors, to allow you to understand how different segments navigate through that process. In addition, it can be used to understand where visitors navigate to after any step in the process. Both the visitors that make it to the next step in the process, as well as those who don’t.
The Fallout visualization allows an analyst to drag the data points that make up each step of a process onto the visualization. In this case let’s say that management has asked how visitors flow through the checkout process and most importantly where there are problems in the process. This is a perfect use case for using the Fallout visualization.
Let’s start by adding a new panel to our project. Minimize all of your open panels and then Click on the + sign underneath the last panel, to add a new blank panel.
This will open the Panel Builder screen, where you can start with a blank freeform table as well as a number of other visualizations, including the Fallout visualization.
Select the Fallout Visualization.
The Fallout Visualization starts with a first step of All Visits by default. But it can be toggled off as we’ll cover later in this lesson.
The checkout process involves hitting key pages on the site. Let’s drag over those key pages into the Fallout so we can see how visitors traverse that process and where they fallout of the process. Click on the arrow to the right of the Page dimension under dimensions in the Components left-rail menu.
That will allow us to view all the page names in the Page dimension. Type “purchase” into the search bar to show only the pages with the string “purchase” in their name.
Drag the page named “purchase step 1” and drop it into the Add Touchpoint drop-zone within the Fallout Visualization.
This will add the purchase step 1 page to the Fallout Visualization. From here we can see that 22.2% of all visits touch the “purchase step 1” page and that 77.8% don’t.
Ok. Let’s add the next step in the checkout process. Drag the page named “purchase step 2” and drop it into the Add Touchpoint drop-zone, beneath the “purchase step 1” step within the Fallout Visualization.
We can now see that 18.2% of All Visits made it to the “purchase step 2” page and 18.2% of the Visitors that made it to the purchase step 1 page did not make it to the “purchase step 2” page.
OK. Let’s add the last page of the checkout process. Drag the page named “purchase: thank you” and drop it into the Add Touchpoint drop-zone, beneath the “purchase step 2” step within the Fallout Visualization.
We can see here that 14.1% of All Visits made it to the “purchase: thank you” page, and that 22.5% of the Visitors that made it to the purchase step 2 page did not make it to the purchase: thank you page.
Now let’s assume that we’ve built this Fallout Visualization and management asks us for a combined view into the checkout process for both the website and the mobile app. We can easily add additional “pages” to each step as an OR statement to get that view.
The app pages have “app:” in the front of the page names. Let’s start with the first app page in the checkout process. Drag the page named “app: purchase step 1” and hover over the “purchase step 1” page in the Fallout until that step gets a blue box around it.
Drop the “app: purchase step 1” page.
This will create an OR statement for that step. It will show how Visitors are navigating the process where they hit the “purchase step 1” page OR the “app: purchase step 1” page.
Let’s do the same for the “app: purchase step 2” and “app: purchase confirmation” pages, dropping them in the appropriate steps in the Fallout Visualization. This will give us a view into how visitors navigate the process, regardless of whether they’re doing it on the website or within the mobile app.
Hovering over any step in the Fallout Visualization will bring up a pop-up that shows a number of stats:
There will be times when you may want to change the steps in the Fallout Visualization. This can be done easily by deleting any step(s) and dragging any new step(s) that need to be added to the visualization. Let’s say we want to replace the final step in the Fallout Visualization with a different step. The first step is to delete the existing step from the Fallout Visualization.
Mouse-over that step in the Fallout Visualization and click on the X to the right of the description of that step to delete that group of pages.
This will delete that step in the Fallout Visualization.
The Fallout Visualization is not stuck using just one dimension or data point. As an analyst, you can mix and match what you use in the visualization. For example, imagine that the Checkout Process has a number of pages that a Visitor needs to navigate through, but the final conversion event isn’t measured by reaching a particular page or group of pages. Instead, there is a custom event that represents the final conversion. We can drag a metric that represents that final conversion event as the final step in the Fallout Visualization, even though all the other steps in the Fallout Visualization are represented by touching specific pages.
Let’s use a metric to represent the final conversion event.
Drag the Online Orders metric from the Metrics section of Components in the left-rail and drop into the Add Touchpoint drop-zone underneath the purchase step 2 step in the Fallout Visualization.
We now have a Fallout Visualization with a combination of dimensions and metrics to represent each step of the process.
There are times when you want to understand how Visitors follow a specific path with sequential events with nothing in between, and other times when as long as a Visitor hits key events at some point, it still counts as success, even if there are other events that occur between each step of the process. The Funnel Visualization supports the ability to configure each step to be Eventual Path or Next Hit. Eventual Path will allow other events to occur in between a step and the next step. Next Hit will only count instances where a step is hit with the next Hit being the next step in the Funnel Visualization.
Let’s imagine that we want to measure the number of Visitors that hit step 1 of the purchase process and their very next Hit was step 2 of the purchase process. This can be configured in the Fallout Visualization by clicking on Eventual Path under step 1 of the purchase process and selecting Next Hit.
As you can see, the Fallout Visualization is now configured with step 2 of the purchase process as the Next Hit after step 1. The rest of the steps are configured as Eventual Path
The next step after viewing the success rates for each step of the process is to dig into what people do next after they hit a step of interest in the process. When we look at the Fallout Visualization for this checkout process, we can see that we’re losing the most Visitors between step 2 of the process and the actual Online Order event. We can drill into this step and see what happens next for Visitors who have an Online Order event as well as for Visitors who don’t.
Let’s take a look at what Visitors do after step 2 when they do have an Online Order event. Right click on the green bar to the right of step 2 in the Fallout Visualization to bring up the menu of options.
Click on “Breakdown fallthrough at this touchpoint” to get a view into what visitors who do make it to the Online Order event do right after hitting step 2 of the purchase process. This helps us understand what is driving success. The result is a table showing what Visitors do immediately after step 2 in the process.
The same analysis can be performed but for Visitors that fallout of the process at that step. Meaning they don’t make it to the next step in the Fallout visualization. In this case it allows us to start understanding what Visitors who don’t make it to the Online Order event do immediately after step 2 of the purchase process.
Right click on the green bar to the right of step 2 in the Fallout Visualization to bring up the menu of options again.
This time we’ll select “Breakdown fallout at this touchpoint” to analyze where Visitors fall out of the purchase process and what they do immediately after step 2 of the process.
The Fallout Visualization can do more than just enable an analyst to understand how Visitors navigate through a process of interest. It can also allow an analyst to quickly build segments of Visitors who have followed a particular path, or get an analyst started with a segment of Visitors that started a process but didn’t make it to the end goal.
Let’s say that we want a segment of Visitors that started the checkout process and made it to step 2 of the process.
We can simply right click on the green bar to the right of step 2 in the Fallout Visualization to bring up the menu of options again.
Click on “Create segment from selection” to open the Segment Builder.
From here we can give this Segment a name and save it for future use. We could also change this to be a Visit based segment versus a Visitor based segment which would constrain the segment to visits that meet that criteria. In addition, the segment could be tweaked to look for Visitors that made it to any step in the process but didn’t make it to the end goal.
The Fallout Visualization supports side by side comparison of how multiple segments of Visitors navigate through the process in the Fallout Visualization. This provides an easy way to see if one segment of Visitors may be navigating through the process more efficiently than another segment of Visitors.
Imagine a situation where there has been a change in the checkout process and the organization wants to verify that Visitors on each of the major browser types can still make it through the process without any trouble. The real question here is whether a certain browser type has significantly lower conversion rates in the checkout process compared with the other browser types.
Click on the arrow to the right of the Browser Type dimension to show the browser types collected in that dimension.
Drag the Apple browser type over to the very top of the Funnel Visualization, until you see a blue Add rectangle at the top of the visualization. Drop the Apple browser type.
Applying the segment to the Funnel Visualization will give two bars to the right of each step. The top bar represents All Visits and the bar underneath represents browser type of Apple. As we analyze this, we can see that the Apple browser type has a significantly higher conversion rate in the purchase process when compared with All Visits.
We can continue to add segments to the Funnel Visualization by dragging them and dropping them at the top of the visualization.
Let’s add browser types of Google and Microsoft to the Funnel Visualization by dragging them over.
We can now see the performance of All Visits as well as browser types of Apple, Google, and Microsoft and we can compare them against each other. This is a very powerful way to understand how different types of Visitors navigate through the purchase process and if there are certain types of users that may be struggling.
The Venn Diagram visually shows the relative size as well as the overlap or intersect of up to 3 segments against a metric of interest. Any segment(s) can be dropped into the Venn visualization. In addition, dragging a dimensional element (such as Paid Search Marketing Channel) can be dropped into the visualization as a segment.
In this exercise we’ll add a Venn Diagram to a Panel and add 3 segments to it.
Click on the + below the bottom of the last panel in the Project we’ve been working on to create a new Panel.
Pro Tip – minimize Panels that you’re not using to improve performance of the project.
This will give you options to add a visualization to the Panel. Click the Venn icon to create a new Panel with a Venn Diagram.
The Venn Diagram configuration screen allows users to add up to 3 segments as well as a metric to apply to the visualization.
The segments that are added to the Venn Diagram can be actual segments that are defined within Adobe Analytics, or any dimension can be dragged on to create a segment based on that dimension having a value, or a dimensional element (Paid Search in the Marketing Channel dimension for example) can be dragged on as a segment as well.
Let’s say we want to build a Venn Diagram that shows the relative size as well as the overlap of Visitors that interact with the Display, Email and Paid Search Marketing Channels. We can drag those Marketing Channels over to the Venn Diagram.
Drag the Marketing Channel of Paid Search over and drop it into the Add Segment drop-zone.
Drag the Marketing Channel of Email over and drop it into the Add Segment drop-zone.
We have 3 segments that will be used for the Venn Diagram. Now we need to add the metric that we want to use. In this case, let’s say that we want to analyze how Visitors interact with these 3 channels.
Click on Build to build the visualization.
The result is a Venn Diagram that shows the relative size of each segment and the overlap of each segment. This may not look as you would have expected. The relative size looks good, but there is no overlap between any of the segments. This is because by default, the Venn Diagram is configured to analyze at the Hit level. So, this is showing the Visitors that interacted with each of these Marketing Channels at a Hit level. It’s impossible to have a Hit where a Visitor interacted with multiple Marketing Channels, so that’s why there is no overlap. We’ll cover changing the level of the visualization in the next exercise.
In the last exercise we built a Venn Diagram visualization that analyzed Visitors that interact with various Marketing Channels at a Hit level. In this exercise we’ll walk through changing the level of that visualization from Hit level to Visitor level. The Venn Diagram creates a hidden table with temporary segments in the background based on what is dragged into the Venn Diagram configuration screen. By default, these segments will be configured at the Hit level. Let’s change the level now.
Click on the Manage Data Sources dot in the top left corner of the Venn Diagram that we created in exercise 4.1.
This will bring up the Data Source Settings window.
Select the Show Data Source checkbox. This will expose the underlying hidden table that the Venn Diagram visualization is built from. By default it is a hidden table.
If you click on the “i” on the right side of the first column in the table that is exposed, it will show you how this segment has been defined. You can easily see that it is configured as a Hit level segment.
Click on the Pen in the top right corner to edit this segment in the Segment Builder.
Click on Hit under the segment definition and change it to Visitor. This will change the level of the segment from Hit to Visitor. This will change this segment to look at Visitors that have touched this Marketing Channel in the Reporting Window.
Click on Save to save this change and you will see the Venn Diagram rebuild and you’ll also see an immediate change on what the Venn Diagram looks like. There is now some overlap.
In order for the Venn Diagram to show the overlap at a Visitor level across all 3 Marketing Channels we need to configure all 3 of the temporary segments at the Visitor level. So far, we’ve only done this for 1 channel. The 2nd and 3rd columns in the table still need to be configured to Visitor level instead of Hit level.
Go ahead and perform those same actions to change the level of the temporary segments from Hit to Visitor in the 2nd and 3rd columns in the table.
Once they are saved those segments, the Venn Diagram should look similar to the following. As you can see there is quite a bit of overlap now between the channels.
Clicking the Manage Data Sources dot in the top left corner of the Venn Diagram visualization will allow us to hide the table that the Venn Diagram is built off of again by unchecking the Show Data Source checkbox.
Many times when analyzing behaviors of Visitors there will be a need to build a segment or segments based on those Visitor behaviors. Many of the visualizations within Analysis Workspace support that requirement, and the Venn Diagram is no exception. In this exercise we’ll learn how to build segments based on Visitors that fall into any cross-section of the Venn Diagram.
Hover over the center cross section in the Venn Diagram that represents Visitors that interacted with all 3 Marketing Channels in the current Reporting Period.
Right click in that cross section. This will bring up the “Create Segment from Selection” menu option. This option will build a segment based on Visitors that interacted with all 3 Marketing Channels in the current Reporting Period.
Click “Create Segment from Selection” to open up the Segment Builder. With the criteria of Visitors that touched all 3 Marketing Channels already populated.
From here we could modify this segment criteria and save it if we want.
Let’s save it so we can use it for our next Lesson.
Give it a name similar to “Paid Search – Display - Email Visitors” and click Save to save this segment.
Learn how to build a Segment Comparison panel
Learn how to use the Segment Comparison panel
Segment Comparison uses the power of Sensei to find differences in segments. It will algorithmically find the top differences in Metrics, Dimension Elements and Segments as it applies to the 2 segments that it’s comparing. This is very powerful in that it has the ability to uncover relationships in the data and in these segments that perhaps was not known before. This can result in the creation of new segments based on the output of this analysis.
In this exercise we’ll configure the Segment Comparison Panel to compare the Paid Search – Display – Email Visitors segment that we created in Exercise 4.3 against Everyone Else (anyone that doesn’t fall into that segment).
Drag the Segment Comparison panel over from the Panels section in the left-rail menu and drop it under the bottom Panel in the current Project. This will present the Segment Comparison configuration screen within a new panel.
From here we can choose up to 2 segments that we want to compare. They can be actual segments. Dimensional elements can be dragged over as segments. Dimensions can be dragged over, which just like the Venn Diagram, would represent that a value has been collected in that Dimension. Metrics can be dragged over as well which would represent Visitors who had the event that the metric represents.
Drag the Paid Search – Display – Email Visitors segment and drop it into the Add Segment drop-zone.
By default the Segments Comparison configuration screen will offer to compare the segment that has been dragged over to Everyone Else that doesn’t fall into that segment. If that is the goal of the comparison you can simply click Build to compare that segment to Everyone Else. If the goal is to compare 2 distinct segments against each other, the other segment can be dropped into the Compare Segment drop-zone and it will replace the Everyone Else segment.
There are also Advanced Settings that can be used to restrict specific dimensions, metrics, and segments from being used in the comparison. Click on Show Advanced Settings.
From this view we can see the number of dimensions, metrics and segments that are being excluded from the analysis. Clicking on Dimensions, Metrics or Segments in the Excluded Components will show what is being excluded. Click on Dimensions to see what dimensions are being excluded from being used by Segment Comparison.
Additional dimensions, metrics or segments can be excluded from use by Segment Comparison by simply dragging them from the left-rail Components menu and dropping them in to the Excluded Components bar.
Let’s build this Segment Comparison panel based on the segment of Paid Search – Display – Email Visitors compared against Everyone Else. Click Build.
The result will be Sensei processing through every metric, dimensional element as well as segment to determine where the biggest differences are in the data. The goal is to expose these differences in an easy to understand panel that allows an analyst to quickly view and drill into these differences. In many cases this will expose relationships in the data that the analyst was not previously aware of and may result in new segments being created.
Let’s go over the completed Segment Comparison panel.
The Top Dimension Items Against Segments table is where the dimension item differences for each segment will be exposed. This is a very powerful view into the data. It will show each dimension item in the table and the percent of visitors that touched that item, from each segment that is being compared. It then applies a statistical test against the percentage of visitors for these 2 segments to see if there is a large difference. The result is a Difference Score which is the 3rd column in the table. The closer to 1 this Difference Score is, the more significant the difference. The table is sorted highest to lowest by the Dimension Items that have the highest Difference Score down to the lowest. Sensei is not aware of what each metric, segment, or even variable represents in the data. This could lead to some relationships being very obvious. At any point a user can click on the X to the right of any dimension item in the table to remove it from view.
The Dimension Items by Segment bar chart visually represents the percentage of visitors that fall into each segment being compared for the selected dimension item. It will default to the top dimension item in the table, but any dimension item in the table can be selected to visually represent that dimension item in the bar graph.
The Top Segments Against Segments table is where the segment differences for each segment will be exposed. This is a very powerful view into the data. It will show each segment in the table and the percent of visitors that fall into that segment, from each segment that is being compared. It then applies a statistical test against the percentage of visitors for these 2 segments to see if there is a large difference. The result is a Difference Score which is the 3rd column in the table. The closer to 1 this Difference Score is, the more significant the difference. The table is sorted highest to lowest by the segments that have the highest Difference Score down to the lowest. Sensei is not aware of what each metric, segment, or even variable represents in the data. This could lead to some relationships being very obvious. At any point a user can click on the X to the right of any segment in the table to remove it from view.
The Segment Overlap venn diagram visually represents the relative size as well as overlap of visitors that fall into each segment being compared for the selected segment. It will default to the top segment in the table, but any segment can be selected to visually represent that segment in the venn diagram.
The Segment Comparison panel defaults to trending metrics over time in a line graph, comparing dimension item segment size in a bar chart, and using a venn diagram to compare segments but this can be changed easily as analysis needs change.
Click on the gear in the right corner of the Dimension Items by Segment bar chart. From here you can change the visualization type being used.
Change the Visualization Type from a Bar to a Line.
This changes the visualization options available. You can change the Granularity of the line graph from Day to Week for example. Change Granularity from Day to Week and click anywhere outside of the configuration screen to view the changes.
The goal of Segment Comparison is to uncover relationships in the data that may not be obvious. It then allows an analyst to drill into the data to explore these relationships.