Distinct Count in Segmentation

In last week’s Adobe Analytics release, a new feature was added within the segmentation area. This feature is called Distinct Count and allows you to build a segment based upon how many times an Adobe Analytics dimension value occurs. While the names are similar, this feature is very different from the Approximate Count Distinct function which allows you to add distinct counts to a Calculated Metric. In this post, I will describe the new Distinct Count segmentation feature and some ways that it can be used.

Segmenting on Counts – The Old Way

When doing analysis, there are often scenarios in which you want to build a segment of visitors or visits that have done X a certain number of times. For example, you may want to look at visitors who have viewed more than two products but never added anything to the shopping cart. Or you may want to identify visits in which visitors read more than three articles.

This has been somewhat possible in Adobe Analytics for some time, but building a segment to do this has always relied on using Metrics (Success Events). For example, if you want to build a segment to see how many visitors have viewed more than three blog posts, you might do this:

You could then use this segment as needed:

The key here is that you need to have a Success Event related to the thing that you want to count. This can be limiting because you might need to add extra Success Events to your implementation. But the larger issue with this approach is that a visitor can make it into the segment even if they viewed the same blog post three or more times because it is just a count of all blog post views. Therefore, the segment isn’t really telling you how many visitors viewed three or more distinct blog posts.

At this point, you might think, “well that is what I use the Approximate Count Distinct function for…” but, as I mentioned earlier, that function is only useful for creating Calculated Metrics. As shown below, using the Approximate Count Distinct function tells you how many unique blog posts titles were viewed each day or week and doesn’t help you answer the question at hand (how many visitors viewed three or more blog posts).

Segmenting on Counts – The New Way

So you want to accurately report on how many visitors viewed three or more different blog posts and have realized that segmenting on a metric (Success Event) isn’t super-accurate and that the Approximate Count Distinct function doesn’t help either! Lucky for you, Adobe has now released a new Distinct Count feature within the Segmentation area that allows you to build segments on counts of dimension (eVar/sProp) values. Before last week’s release, when you added a dimension to the segment canvas, you would only see the following operator options:

But now, Adobe has added the following Distinct Count operators that can be used with any dimension:

This means that you can now segment on counts of any eVar/sProp value. In this case, you want to identify visitors that have viewed three or more different blog post titles. This can be done with the following segment:

The Results

Once you have created your segment, you can add it to a freeform table to see how many unique visitors viewed three or more blog posts:

In this case, over the selected time period, there have been about 2,100 visitors that have viewed three or more blog posts on my site and I can see the totals by day or week as shown above.

As a side note, if you did try to answer the question of how many visitors viewed three or more blog posts using the old method of segmenting on the Success Event counts (Blog Post Views >=3), you would see the following results:

Here you can see that the number is 3,610 vs. the correct number of 2,092. The former is counting visitors who viewed more than three blog posts, but not necessarily three or more different blog posts. All of the visitors in the correct table would be included in the incorrect table, but the opposite wouldn’t be true.

Again, this functionality can be done with any dimension, so the possibilities are endless. Here are some potential use cases:

• View Visitors/Visits that viewed more than one product
• View Visitors/Visits that used more than three internal search terms
• Check potential fraud cases in which more than one login ID was used in a visit
• Identify customers who are having a bad experience by seeing who had multiple different error messages in a session
• Identify visitors who are coming from multiple marketing campaigns or campaign channels

To learn more about this new feature, check out Jen Lasser’s video release video.