GA4: Reporting on New vs. Returning Users in Looker Studio
Вставка
- Опубліковано 31 тра 2024
- Learn how to create a pie chart that shows new vs. returning users using ninja Looker Studio techniques including custom fields, blending data with a full outer join and the mysterious COALESCE function.
Find an introduction to blending data in Looker Studio here: • Mastering Data Blends:...
And view our blog for written instructions & code snippets: twooctobers.com/blog/new-vs-r...
00:00 Intro
01:48 The problem with GA4's returning users metric
05;15 A donut chart using Google's metrics
06:47 A different way of calculating returning users
08:07 Creating a new & returning users blend
09:40 A donut chart using our new metrics
11:33 The magic of a full-outer join + COALESCE()
Brilliant video, thank you :) I don't know why Google is making it so difficult for something that should be standard!
Great explanation, thanks!
This was really easy to follow and I appreciated you taking the time to create the graph and explain the Coalesce etc. I'm glad to have found your content.
Glad it was helpful!
I can't believe how quickly I found this via Google, how easily you explained it, and how quickly I was able to replicate what you did. Thank you!! This is much more representative of the behavior our clients are trying to understand.
You're welcome!
Great teaching! Thank you. This was so helpful!
You are so welcome!
Why do you use active users rather than total users?
Any Idea why (not set) comes in New/ Returning User dimension ?
Sort of. Here is how this dimension is defined in Looker Studio: "New users have 0 previous sessions, and returning users have 1 or more previous sessions. This dimension returns two values: 'new' or 'returning'."
This does not correspond to a dimension in GA4 reporting, which is curious, but I suspect that it relates to the 'ga_session_number' event parameter which is included in the GA4 BigQuery export. So I compared my Looker Studio report, which shows 9 'New / returning' (not set) users for Jan 1 - Jan 24 to user data in BigQuery. What I found in BigQuery was that there were 9 user sessions that did not include a 'first_visit' event, but had a value of 1 for the ga_session_number parameter. In other words, there were 9 users that did not count as new, but also did not have a previous session. I don't know why these user sessions did not include a 'first_visit' event, but if Google is listening, I think I've at least narrowed down where the bug is coming from ;)
@@Twooctobers There can be multiple reasons for this - it could be related to the users’ browser settings, such as blocking cookies or using incognito mode, which prevent Google Analytics from accurately tracking and categorizing the user.
This was very helpful in understanding the nuances to tracking by user types in Looker Studio. I was building a report detailing acquisition by user type and this is where I noticed the over inflation... I am definitely in agreeance that I rather look at new as unique first time and returning as repeat. How do you calculate something like sessions and engaged sessions on this unique requirement?
Thanks so much very helpfull
Have you tried setting an an extra Data Range Control to provide Date Range Comparison between two dates in GA4. Back to UA I know it was possible because I had a report doing that but I don't know how to do it in GA4 in Looker Studio
If I understand the question correctly, you can't do it with an extra date range control, but there is a 'Comparison date range' option with many chart types. It behaves differently depending on the type, so may not do exactly what you want. You *can* group charts and controls, so it would be possible to have two charts side-by-side with two different controls, but unfortunately, there isn't a way to make one control relative to another. E.g. if a user picks a date range in control A, you can't have control B update to the previous period in relation to control A.
Thanks for the video! I would like to point out that instead of 'active users' it would have been better suited to use 'total users' since a new or returning user isn't necessarily an active user.
You're welcome!
An active user in GA4 is defined as "any user who has an engaged session or when Analytics collects: the first_visit event or engagement_time_msec parameter from a website", and the New user metric is "measured by the number of new unique user IDs that logged the first_visit event". So all New users are Active users. Given the way I am calculating Returning users, it therefore makes sense to subtract New users from Active users. support.google.com/analytics/answer/11986666
Is there any tutorial for mobile cohort users? I want to make a Retention for cohort users
No, sorry, haven't done a tutorial on that.
In your third part of the video, A different way of calculating returning users, there is a difference between the number of new users in the table (9815) vs the number of new users on the scorecard (9847). How can this be explained?
There are various reasons why the same metric can be different in two different places. One is sampling, but twooctobers.com doesn't have enough event volume for sampling to kick in (10 million events in a given report). Another is thresholding, which applies if Google Signals is enabled and the reporting identity is Blended or Observed, which was the case when I recorded the video. The last is that User and Session metrics are ESTIMATES(!) in GA4. The specifics are pretty technical, but Google explains them here: developers.google.com/analytics/blog/2022/hll
This article does a good job of explaining the impact in less technical terms: www.quantable.com/analytics/how-accurate-is-hyperloglog-in-ga4/
This was super helpful! I have been racking my brain on something though. If google is counting the user twice, once as a new and once as a returning, why would the returning user number be inflated? From what I understand from your explanation, a returning user will have been already been counted as a new user. So for example user A comes to my site 1 time, user B comes to my site 2 times, and User C comes to my site 3 times. According to GA my New user count would be 3 and my Returning user count will be 2. Whereas, if we go by your definition in this scenario New user count would be 1 and my returning user count would be 2. However, in your formula for returning users (Active users - New users) the result is a deflated Returning user count and the new user count remains unaltered. I would have expected the New user count from GA to be the inflated value. Thoughts?
From my POV, the GA4 returning user count is inflated because I think of a user as either being new or returning in a given time period. In other words, if I am reporting on August, 2023, I think of a new user as a user who first visited the site in August, 2023. This matches how Google counts new users. But I think of a returning user as someone who visited the site before August, 2023, whereas Google will count a user who visited on August 1st and 2nd as a new user on August 1st and a returning user on August 2nd. IMO, my definition matches how marketers think of new vs. returning users better, but they are really just two different definitions.
Hopefully that helps, and I showed both approaches because I knew some people would prefer to stick with Google's definition.
Hello, thanks for this explanation. Can you please advise why in your example above why Google shouldn’t be counting the user that came back on Aug 2nd as “returning”. Sorry, I’m just a little confused by that
@@27131006456299 it's just a matter of opinion, really. I think of the word "user" as describing a person. If the same person visits my site three times, I don't say I had three users, I say I had one user and three sessions. If I ran a coffee shop and on a given day had one brand new customer that visited my coffee shop twice on that day and three customers that had been to my coffee shop before, I wouldn't say I had one new customer and four returning customers. But I demonstrate how to do the visualization both ways because I know some people prefer Google's method.
Thanks, this makes sense to me. At our company we also think of user as a person and that’s why we are struggling with how GA4 is doing it now. Would you say the method you show is the more accurate way of calculating new vs returning the if we consider users as a person?
@@27131006456299 yes, I would.