Using Data Science to Compare Segments in Adobe Analytics - dummies

Using Data Science to Compare Segments in Adobe Analytics

By David Karlins

One of the most interesting data-science-influenced features in the Analysis Workspace of Adobe Analytics is called Segment Comparison. It’s one of Adobe Analytics most powerful features because it helps you easily differentiate two groups of visitors based on their behavior.

Have you reviewed a Venn diagram recently and wondered about the following?

  • What are the metrics that differentiate my segments? For example, you might discover a correlation between your logged in visitor segment and views of blog articles.
  • What are the dimensions that differentiate my segments? For example, you might discover that iOS device types make up a significant portion of visitors who have converted.
  • What other segments can be applied to my segments to differentiate them? For example, you might see that visitors referred from paid search have a higher likelihood of being in the custom segment you created of first-time visitors who land on a product page.

If so, you’ve come to the right place! Those questions are the ones that Segment Comparison aims to answer, using machine learning every step of the way!

First, you learn the steps to compare two segments. Then you dig into a few other examples of useful segment comparisons to try out.

Invoking Segment Comparison in Adobe Analytics

To compare segments in Adobe Analytics, follow these steps:

  1. Use the left rail selector to change the left rail’s view to panels.
  2. Drag the Segment Comparison panel into Workspace.
    Adobe Analytics Segment Comparison panel
    Dragging the Segment Comparison panel into workspace.
  3. Drag a segment into the Add a Segment area in the Segment Comparison panel.
    defining segments in Adobe Analytics
    Defining the first segment for Segment Comparison.

    In addition to segments, you can also drag dimensions, dimensional items, metrics, and time ranges into the box. In the example, a segment that was already created was used that focuses on iOS devices. If you’d like to create a segment in this workflow, hover your cursor over the Add a Segment box and click the plus sign that appears.

  4. Drag a second segment (or any other component) into the Compare Against box.

    By default, Adobe includes a segment focused on Everyone Else, in case you merely want to compare this segment against its inverse. However, this results in a Venn diagram without any overlap. This can be useful, but for the example, you are comparing against a second segment.

    In the image below, a Purchasers segment was added to the Compare Against box. Based on the assumption that some visitors on iOS devices have made purchases, you can expect to see some overlap between the two segments.

    defining a second segment in Adobe Analytics
    Defining a second segment for Segment Comparison.
  5. Click the Show Advanced link.

    The box that appears should look familiar if you remember the Excluded Dimensions box for Contribution Analysis. As shown above, you have the option to exclude dimensions, metrics, and segments from the analysis and results of your segment comparison. Just as in a Contribution Analysis, the reason to exclude dimensions is to avoid annoying results that may be accurate but aren’t actionable. Because Segment Comparison also analyzes metrics and segments, they are added as options for exclusion.

  6. Click the blue Build button and examine the results that appear, often in less than a minute.

    As you can see below, a wealth of visualizations are returned from Segment Comparison.

    Segment Comparison results in Adobe Analytics
    The top portion of results from Segment Comparison.

Let’s review them from top to bottom, left to right:

  • The Size and Overlap visualization is a Venn diagram that you could have easily built based on your two segments and a metric of unique visitors.
  • The next three summary visualizations show the count of unique visitors for each of your segments as well as the overlap between them. These summary visualizations can be a good reference as you perform your analysis.
  • The two visualizations in the second row of the results are live-linked, which means that when you click a value in the table on the left, the graph on the right is updated. The table on the left shows the top metrics that differentiate your two selected segments. A column for difference score sorts each of the metrics by their level of statistical significant, just like Contribution Analysis’s contribution score. You’ll be glad to see that Adobe runs Anomaly Detection on all metrics that are trended in the line chart on the right too! As you can see in the third row in the table above, purchasers end up watching a significantly higher amount of media (an average of 285.81 seconds) than iOS visitors (an average of only 66.03 seconds).
  • The third row shows two more live-linked visualizations: a table on the left of the dimensional items that differentiates your two segments and a bar chart on the right that shows the stark difference when each of your two segments are applied to the dimensional items. A difference score is again applied to these visualizations.
  • The fourth and final row of your Segment Comparison results uses your own data to help you differentiate the segments you’re comparing. Adobe analyzes all segments you have access to or have created so that it can provide a final list of differentiating segments. The freeform table live-links the differentiating segments to a Venn diagram on the right, which allows you to quickly find overlap between three different segments.

Adobe’s Segment Comparison tool is a fantastic and fast way to learn more about your visitors. You’ll be glad to hear that there is no limit to the number of times you or your company can use this tool, regardless of contract or SKU. If you don’t have access to it, be sure to work with your administrator to understand why.

Using Adobe Analytics to brainstorm Segment Comparison use cases

If your brain isn’t already working in high gear thinking creatively about how to use the Segment Comparison feature in Adobe Analytics, let’s give it a kickstart. Our first recommendation is to start with a segment of converters to your site.

As a reminder, someone who converts doesn’t necessarily mean he or she has purchased. For websites and apps that don’t sell anything, a conversion could mean a registration, a video view, or a threshold of unique views of content. Whatever your conversion, create a segment based on visitors who have accomplished it.

The most basic comparison to run is converters versus non-converters. At first, Segment Comparison will most likely tell you things you already know — your best marketing channels, micro-conversions that are leading indicators to success, or regions of the country that are more successful than others.

Segment clustering in Segment Comparison’s results may come in handy in use cases like these. Are there any unusual combinations of those dimensions that Adobe Analytics suggests reviewing? If Adobe provided you with any segments that are comprised of atypical dimension combinations, start your analysis there.

For those of you who do have a purchase funnel, consider creating segments in Adobe Analytics for each of the key steps of the funnel — visitors who get to a product page but don’t add to cart; visitors who add to cart but don’t get to the checkout page; visitors who get to the checkout page but don’t purchase; and visitors who purchase.

Play around with each segment in Segment Comparison to learn more about what differentiates one segment from the other. Because you have unlimited access to the Segment Comparison tool, you might as well try it out!

Last, and certainly not least, try a mixture of marketing channel and account purchase status. For example, segment visitors who access your site via the most successful marketing channel and compare it to visitors who purchase. Then compare it to visitors who don’t purchase.

Segment Comparison will often help you discover metrics, dimensions, and segments that differentiate segments that you hadn’t thought to consider comparing before. Ideally, that newfound information will help drive your data curiosity and improve how you use Adobe Analytics.