Analytics
4.2K members online now
Understand information in your reports and troubleshoot reporting issues such as self-referrals, (not set) data, and inaccurate information
 
Guide Me
star_border
Reply

Why is segmenting by Days Since Last Session inaccurate?

Visitor ✭ ✭ ✭
# 1
Visitor ✭ ✭ ✭

Hi all! 

 

I was recently playing around with the segmenting tool.  

 

I'm trying to segment my Events by my retention metrics — Date of first session, Days Since Last Session (Recency), and Total Session (Frequency).

 

I'm doing so in a User ID view. 

 

The other day, I looked at the Recency & Frequency report and, in the Recency histogram, I saw that 20 users had 0 Days Since Last Session.

 

I then created a segment of Days Since Last Session = 0.  

 

I expected this segment to display data for those 20 users.

Instead, this segment displayed data for about 340 users (about 97% of my site's total user base). 

 

Why isn't the Recency segment returning data for the group of users I assume I selected by setting Days Since Last Session to 0 in the segmenting tool?

 

Thanks!

Zack 

1 Expert replyverified_user

Re: Why is segmenting by Days Since Last Session inaccurate?

Top Contributor
# 2
Top Contributor
Segment, Sessions, Users = confusion.

When you build a segment, it selects sessions that match. If you select a segment with a Users scope, then it selects all sessions for users that had at least one session that matched. That means that 97% of your users had at least 1 session with a duration of 0. The segment also includes all the other sessions those users had.
Mike Sullivan, Google Analytics Top Contributor
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.

Re: Why is segmenting by Days Since Last Session inaccurate?

Visitor ✭ ✭ ✭
# 3
Visitor ✭ ✭ ✭
Hey Mike,

Thanks so much for getting to back to me (and sorry I'm so late here).

That makes sense to me, but then how could the percentage of users with at least 1 session with a duration of 0 ever be less than 97%?
I'm thinking that to be counted as a user in GA, a person has to have visited the site at least once, so by definition, a user is a person with at least one session, which, I assume, would be characterized as 0 days since last session.

Is my thinking at all on point?

If there's any way I can further clarify, let me know!

Thanks,
Zack

Re: Why is segmenting by Days Since Last Session inaccurate?

Top Contributor
# 4
Top Contributor
You are mixing terminology...this might help:

Length of an individual session: Session Duration
Time between sessions of the same user: Days Since Last Session

A 'bounce' session, with one pageview has a Session Duration of 0.
A user visiting twice on the same day has a Days Since Last Session of 0.
New users to your site have 0 days since their last session, however if a user first visits your site in Aug, 2015 and returns in Jan, 2016, when you look at the Jan 2016 report, you will NOT see their first session because it is outside the date range of the report.

The Frequency & Recency report shows Sessions, not Users.
Mike Sullivan, Google Analytics Top Contributor
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.

Re: Why is segmenting by Days Since Last Session inaccurate?

Visitor ✭ ✭ ✭
# 5
Visitor ✭ ✭ ✭
Hi Mike,

Thanks for checking in here again.

And thanks for pointing out my terminology slip and explaining duration vs. days since last session.

I think I'm starting to clarify what i'm confused about. Let me try again. Sorry I'm being a little slow here Smiley Happy

So, in your latest post, you say "The Frequency & Recency report shows Sessions, not Users."
Then, in your previous post, you said: "When you build a segment, it selects sessions that match. If you select a segment with a Users scope, then it selects all sessions for users that had at least one session that matched. That means that 97% of your users had at least 1 session with a duration of 0"

This assertion confused me because I expected that 100% of users would have at least 1 session with a duration of 0 because every session is by definition associated with a user (whether identified by cookie or user ID). That is, when I create a segment for "Days Since Last Session = 0", I assume that segment would return 100% unless, as you point out, their first visit was outside the current time range

More concretely, if I look at my All Website Data view, I see 84.89% of sessions are categorized as Days Since Last Session = 0 in the "Days Since Last Session histogram in the Frequency & Recency report.
I expected that creating a segment of "Days Since Last Session = 0" would return 84.89% of all sessions, not 98% of all sessions.

So, why does a segment of "Days Since Last Session = 0" return close to 100% of all sessions (on all of my views, both User ID and non e.g. All Website Data)?
In other words, how does GA keep track of Days Since Last Session?

That high percentage makes me think that the segment includes every session recorded within the set date range. So, I guess I'm thinking, if a user visited in the date range, they'd be tallied as having a session with a Days Since Last Session value of 0. Then, any and all of their subsequent visits during the time range would be included in the Days Since Last Session = 0 segment. That is, if a user has at least one session with Days Since Last Session = 0 — If a user visits during the set time range and that visit is either their first ever, a revisit within a single day, or their first visit in the given time range — then they'll be included in the "Days Since Last Session = 0" segment.

Sorry, again, for being slow here. Thank you so much for all the advice you've given so far. I do appreciate all your help. Let me know however I can clarify my writing.

Re: Why is segmenting by Days Since Last Session inaccurate?

Top Contributor
# 6
Top Contributor
Wow...deep dive indeed...

When a user visits your site, they get a cookie and a timestamp of their visit. Days since last session is calculated from the previous timestamp. To get days since last session=0, they need to visit with 2 sessions on the same day or be a New User. Time period of the report does not affect the calculation. Users that visited before the time period are NOT New Users on their first visit in the time period -- they are returning users with a days since last session>0.

Side note: Session Duration is the length of an individual session and has nothing to do with the time between sessions.

In any regard, the very high percentage indicates either a LOT of new users (easy to confirm with the New vs Returning report) or a LOT of double sessions in the same day (look at the Frequency & Recency report with a Returning Users system-defined segment applied). If the latter is true, look for self-referrals in your data - visits that are broken and then refer back to your own website (creating 2 sessions for a single visit).

I have an 86.9% Days since last session=0 on one of my sites, but I also have 72.5% new users. Over half of my returning users come back same day. The Behavior Flow report can give you a good idea of what is happening on your site.
Mike Sullivan, Google Analytics Top Contributor
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.

Re: Why is segmenting by Days Since Last Session inaccurate?

Visitor ✭ ✭ ✭
# 7
Visitor ✭ ✭ ✭
Ha. Yea, I'm still super new to analytics and am quickly realizing I have a tendency to get lost in the weeds just by diving into various curiosities like this one (regardless of whether or not the concept's important). Working on keeping perspective, but it's still a struggle Smiley Happy

Thanks so much for those suggestions. I'll try those out this week. I appreciate all your help, Mike!

Re: Why is segmenting by Days Since Last Session inaccurate?

Visitor ✭ ✭ ✭
# 8
Visitor ✭ ✭ ✭
Mike -- thanks for your time and expertise here. I'm realizing Days since Last Session isn't what I thought it was. I have been looking for a recency metric -- i.e., days since the user was most recently on my site, measured in "days ago" as in "this guy was last on my site (or purchased for an ecom site) 3 days ago, and this other guy 8 days ago". Not a latency metric which is what Days Since Last Session appears to be (i.e., days between recent visits). Any idea how to get the recency metric I am talking about?

Re: Why is segmenting by Days Since Last Session inaccurate?

Top Contributor
# 9
Top Contributor
This is one of those numbers that is easily confused, and the closer you look, the more confusing it gets.. The Days Since Last Session is the Recency "dimension" -- not a metric -- and therein lies the source of a lot of confusion. If a single user visits 5 days ago then twice today, the first session today is recorded with Days Since Last Session of 5, and the second is recorded with Days Since Last Session of 0.

It gets more confusing if you use a longer time period, since the session from 5 days ago would also appear as Days Since Last Session of 0 (assuming his first visits). This leaves you with a recency chart showing 2 sessions at 0 days ago and 1 session at 5 days ago, which a lot of people interpret as 2/3rd of my visitors are same day visitors...but all three are the same person.

The Frequency/Count of Sessions chart has a similar challenge - it double-counts returning visitors which causes the chart to not be what you thought it was, and the more you try to understand it, the more confused you get. I wrote a 'misunderstood metrics' article on that one if you really want to dig a little deeper:
http://help.analyticsedge.com/googleanalytics/misunderstood-metrics-frequency-count-of-sessions/
Mike Sullivan, Google Analytics Top Contributor
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.

Re: Why is segmenting by Days Since Last Session inaccurate?

Visitor ✭ ✭ ✭
# 10
Visitor ✭ ✭ ✭
Wow, very insightful Mike, thanks!

I guess my question then is how would one get the more standard Recency metric on a user level? i.e. the one for something like RFM analysis -- number of days since a user has either last been on site or last transacted?

Possible in GA?