Hey @Sem,
It’s a really open ended question and doesn’t have one right answer. Understanding the behavior of new users is really critical in understanding how they retain over time. 

Two key aspects of this process would be gaining the understanding of your product’s critical events and the usage interval. Once you have identified that, you can try creating behavioral cohorts and using them in the retention charts and funnel analysis to see if you can find any behavioral insights. 

I found this playbook to be really helpful to get started.

Also, this following post might help you better understand the new user paradigm in Amplitude w.r.t. event segmentation just in case.

Hope this helps!

I’ll echo @Saish Redkar. This is a tricky question to answer without more details, since different apps and services have so different definitions for new and returning users, and specific use cases may call for different approaches as well.

So if you’d be able to elaborate on how you define your cohorts here, not in Amplitude, but in general? When is someone a new user, when they are returning? Are all your returning users a homogenous group for the purposes of this analysis (someone who comes back three days after signing up and someone who has been with you already a year might have very different usage patterns)? What makes you unhappy with the approaches you have tried so far? What kind of questions are you trying to answer?

@MikkoKarvonen  Thank you. You are absolutely right. I might think, that defining a new/returning user is pretty same as GA4 does.

So, I’m talking about anonymous users on a website. It’s pretty simple report. I’d like to compare how many users daily who are the first-time visitors vs those, who have previously been on a website (30 / 60 / 90 days to compare). 

The next, I’d like to compare behavioral pattern between new мы returning. (Conversion Rates, Onboarding behavior etc.)

Based on this assumptions I’m creating two different segments:

1. All users
   who performed New User = 1
   any time in each day,
2. All users
   who performed Any Active event > 1
   any time during last 30 days

Then I’m looking at “Any Active event” daily to see the linear graph. In my ideal world to get pretty close result to what GA4 does. 

The things that raise my concerns are:

• Huge data discrepancy b/w Amplitude and GA4 (Amplitude shows two times more the amount of new users daily)
• When I compare 1) all users (new user event) with 2) all new users (who performed new user=1 any time in each day) I see 124 vs 175 unique events. But I expect to see the same number because the only thing I did is just moved New User event from the left side to segment attribution. It’s better to show on a screen. Attached the report for a specific day.

So, these two thing make me feel that I cannot rely on the data. 

So essentially a new user is someone who visits your website for the first time (as far as the tracking is concerned, people changing browsers, clearing their cookies etc will always distort this data) and returning user is someone who comes back after their first visit.

One reason this is challenging is that Amplitude and Google Analytics approach things slightly differently. Strictly speaking Amplitude does not have a concept of a returning user, as it focuses on active users instead. Also the definition of a visit is bit different.

The users in your segment 1 are also included in your segment 2, since they would also have performed an active event during the last 30 days. This may be intentional, but I’m getting the impression that this is not what you are looking for.

First things that comes to mind: have you had Google Analytics and Amplitude tracking set up for the same period of time? If you’ve had GA before and are now adding Amplitude, Amplitude is naturally considering some users new that are actually returning, since it has not seen those users before.

Also, is the instrumentation the same on both tools?

This is indeed curious. I did some experiments on our data, and it looks like there is a difference between setting the first cohort up as New User = 1 time vs New User >= 1 time. If I use the >= version, the numbers match.

If I had to guess, this is probably related to the way time zones are handled, but perhaps some Amplitude representatives would have more insight?

Keep in mind that to some extent you are comparing apples to oranges, since GA and Amplitude define some things in different ways.

There is a chance that in the case of anonymous website visitors GA may be better in identifying the same user on different devices, if they use Chrome on both devices. That’s something Amplitude just can’t do.

On the other hand, Amplitude will give you much more flexibility to analyse your data than GA does.

@MikkoKarvonen  Man, who are you? God?

There are no words expressive enough to send to you my appreciation. I think, you are totally right, when talking about the different approach for new/returning users into GA4 and Amplitude. I would set up some basic reports for that and then, probably will move to GA4, to prepare some more detailed reports.

Basically, I’m using the same data source for both tools, sending data through Segment, so, any discrepancies appear on the toolset level.

 The users in your segment 1 are also included in your segment 2, since they would also have performed an active event during the last 30 days. This may be intentional, but I’m getting the impression that this is not what you are looking for.

Yes, you are probably right, those, who performed a New User event, can potentially perform more of events later. So, segments are pretty same. But how to differentiate them?

So, for example, a User perform a New User event, so he is tracked as a Unique at the exact date. On a next day, this Specific User can return, and can be tracked as Returning user. The issue can appear:

1. when a user performs more than 1 events on a day of a first visit. Would he be tracked twice at the same date?
2. it looks like these segments are pretty straight. So, If a user appeared in a Segment 1. He would be tracked later as a part of segment 1 (i.e. a New user). The same is true  for Returning users. Am I right?

Thanks a lot Mikko. You are greatly helping.

@MikkoKarvonen  UPD* Just made these two different segments and this seems to be more or less correct for my purposes.

1. All users
   who performed New User >= 1
   any time in each day,
2. All users
   who performed New User = 0
   any time in each day,

So, the segments don’t overlap.