Advanced Segments: Should I filter by Sessions or Users?

Advanced Segments: Should I filter by Sessions or Users?

To filter by Sessions or by Users? That is the question… at least it is if you have ever configured an Advanced Segment in Google Analytics.

Advanced Segments - Should I filter by segments or users?Advanced Segments are a powerful tool for “slicing and dicing” your analytics data, allowing you to isolate and analyze subsets of sessions or users. The choice between sessions or users however, can have a significant impact on the data that gets captured by your segment. Thus, it can have a significant impact on the insights and conclusions that you may draw from your analysis.

Let’s take a look at the difference between session-based and user-based advanced segments, so that no matter what analysis question you are trying to answer, you will know exactly which type of segment to use.

Consider the following scenario where a single person (let’s call her Alice) visits your website three different times:

  1. Alice arrives at your site from a Google search. Makes a purchase. Signs up for your monthly newsletter. Leaves site.
  2. Alice arrives at your site from a link in your email newsletter. Makes a purchase. Leaves site.
  3. Alice arrives at your site from a link in your email newsletter. Watches a video. Leaves site.

Now let’s say you create an advanced segment to analyze “Email Visitors”, people who have arrived to your site from your email newsletter. Assuming that you have tagged the links in your newsletter with appropriate campaign parameters, you can use the source, medium, or campaign dimensions to define this segment (e.g. medium = “email”).

GA Advanced Segments - Filter by Sessions or Users

Consider the following two segments:

  • Segment U: Filtered by users, the segment will capture all of Alice’s sessions, since in at least one of her sessions she has arrived from the email. Thus, this segment includes 1 user and 3 sessions.
  • Segment S: Filtered by sessions, the segment will capture only Alice’s 2nd and 3rd sessions. At the session-level, only #2 and #3 are from the email. Thus, this segment includes 1 user and 2 sessions.

In this case, the difference between session-based and user-based segmentation is one-third of your data!

Let’s say you then wished to look at conversion rates for this segment of email visitors. Specifically, you want to know what percentage of people completed a purchase:

  • In Segment S (which includes Alice’s sessions #2 and #3), only in session #2 did she make a purchase. So the conversion rate would be 50% (1 out of 2).
  • In Segment U (which includes all 3 of Alice’s sessions), the conversion rate would be 66.6% (2 out of 3), since Alice made a purchase in sessions #1 and #2.

50% vs. 66% is a significant difference! Let’s just take the higher number, right?

WRONG!

If you are using Advanced Segments for any kind of analysis, you need to understand the implications of session vs. user-based segmentation.

We have used a single user as an illustration, but when extended to hundreds, thousands, or even millions of users and sessions, the scope of your advanced segment can have a significant impact on the results of your analysis.

So, which is the right choice?

Well, it depends on what you are aiming to measure and analyze.

What’s more important is that you are able to accurately describe exactly what your segment is capturing:

  • Segment S captures all sessions originating from an email link.
  • Segment U captures all sessions from users who arrived from email in at least one of their sessions.

If you are interested in assessing the ability of your email newsletter to directly generate conversions, session-based Segment S may be more relevant. But if you are more interested in the people who clickthrough from the newsletter, user-based Segment U may be more relevant.

Thus, the right choice is always relative to the question that you’re trying to answer.

What kinds of analysis do you use Advanced Segments for? Have you ever been confused by the data that GA shows you? Let us know!

By |2018-07-26T10:14:43-04:00March 26th, 2015|16 Comments

16 Comments

  1. Sasatorn 2015-11-17 at 03:55 - Reply

    Why is it that when choosing landing page condition and filtering by U vs S, for U, in some scenario the user are lesser than S?

    • Marc Soares 2015-11-17 at 11:00 - Reply

      Since a single user can have multiple sessions, in most cases you can expect the number of users to be less than the number of sessions. Let us know if that doesn’t answer your question or if you’d like some further clarification.

      • Sasatorn 2015-11-21 at 13:47 - Reply

        Hi thanks for your reply! I do require a lil more further clarification. I was expecting the number of users under both U and S filter to stay the same, let’s say if I filter by source/medium or url. However, what still puzzled me is that; by destination url, the number of users under U and S differs. Would really help if you help me on this point 🙂

        • Marc Soares 2015-11-24 at 16:16 - Reply

          You should indeed expect the number of users to stay the same, regardless of whether the segment is filtered by users or sessions.

          The first thing I would suggest is to make sure you have created and applied both segments to your report. When viewing only the “preview” that GA shows in the segment configuration panel, the number of users may differ due to sampling.

          • Oz 2017-06-22 at 12:49 - Reply

            There appears to be a subtlety here that is important and may cause user counts to mismatch between User-based vs. Session-based segments:

            How filters are evaluated
            AND conditions in User filters require that all conditions are met in a single hit.

            AND conditions in Session filters require that all conditions are met by any combination of hits within the same session.

            https://support.google.com/analytics/answer/3124493

    • Oz 2017-06-22 at 12:50 - Reply

      here appears to be a subtlety here that is important and may cause user counts to mismatch between User-based vs. Session-based segments:

      How filters are evaluated
      AND conditions in User filters require that all conditions are met in a single hit.

      AND conditions in Session filters require that all conditions are met by any combination of hits within the same session.

      https://support.google.com/analytics/answer/3124493

      • Oz 2017-06-22 at 12:52 - Reply

        BTW, I believe using “exclusion” criteria can also cause a mis-match in user counts between the two types of segmentation.

  2. Niels 2016-01-05 at 08:30 - Reply

    Hi Mark

    Thanks for a good, clear article.
    However I am currently trying to wrap my head around an advanced segment that I am using in a custom report – and I’m not sure whether to filter “sessions” or “users”.

    The report is showing various goal completions taking place outside office hours – i.e. “Hour” is the only dimension and “goal completions goal XYZ” are the metrics.
    So I am able to see e.g. “How many users completed goal 2 at hour 17 (i.e. 5.00 – 5:59 pm) in the last week”.

    However, I only want to specifically see goal completions performed in sessions where the user(s) – at some point in their session – have visited a specific range of relevant pages.

    I have therefore created an advanced segment filtering:
    “Sessions | Include”
    “Page | Starts with | {pagepath}”

    So… will this segment yield the correct results if applied to the custom report?

    I know it’s a loong question – but maybe you can help.

    Have a great 2016 until then 🙂

    • Marc Soares 2016-01-05 at 09:36 - Reply

      Hi Neils,

      It sounds like you have the correct approach. The advanced segment that you describe will include sessions that have viewed at least one page matching your “starts with page path” condition. When applied to your custom report, it will show you the goal conversions completed in those sessions.

      Filtering by Sessions means that all conditions must be met within the same session. If you were to filter your segment by “Users” instead, you would also see visitors that may have viewed the target pages in one session and then completed the goal in a subsequent session.

      Thanks for the comment. Happy new year!

      Marc

      • Niels 2016-01-05 at 09:44 - Reply

        Hi again and thanks for the swift reply!

        That sounds great – that’s exactly what I needed clarified and a big help! 🙂

        Have a great week

        / Niels

  3. rakshit 2016-06-13 at 09:51 - Reply

    Hi I use advanced segment to differentiate between identified and unidentified users . I have created a cookie on my website and a custom variable with that cookie value to know the identified and unidentified users.
    Which filter should i use to measure Identified users and why ??

    • Heather Murphy 2016-06-23 at 15:48 - Reply

      Hi Rakshit,

      To confirm, are you using a custom variable (Classic GA) or custom dimension (Universal GA)? As well, what is the scope of the variable/dimension?

      If you are using custom dimensions (Universal GA) and the scope is set to users, your segment will automatically filter by users. If you are using Classic GA and custom variables, the behaviour of the segment might be completely different.

      Hope this helps!

      Heather

  4. Paul 2017-07-24 at 04:26 - Reply

    Hi, let’s say I want to create a segment of visitors who abandoned a shopping cart but made a transaction later on (in a different session/ same session). Should I create a user segment?

    • Jonathan Iudica 2017-07-25 at 12:19 - Reply

      Hi Paul,

      That is a great question, yes, you should create a user segment!

      Thanks,
      Jonathan

  5. Alex 2017-09-29 at 15:52 - Reply

    Hello,

    Thanks for your article, I’ve been reading about this aspect of GA and I think this is the most clear information I’ve found.

    If we want to understand what percentage of sessions and users complete a destination goal, say, I reckon I can use both, based in users and based in sessions.

    If I understand it right, a sessions based segment would give me the % of sessions and users that have completed a goal when looking at sessions individually. Whereas if I use a user-based segment (multiple sessions) it will include all sessions and users that at least have completed it once, right?

    What is the difference in percentages and absolute numbers (on sessions and users, from Audience) for these two segments telling us though?

    How can I use these segments to find out for example, out of three sessions, how many complete the goal and more specifically how many sessions prior to complete the goal are made and how many after? It would be interesting to find how our discovery and aftermath look like.

    I hope this makes sense.

    Thanks and best regards,

    • Max Bondarenko 2017-09-29 at 15:58 - Reply

      Great questions, Alex. To answer your last question, we created a blog post. Please let me know if this helps.

Leave A Comment