Snowplow and GA are reporting different visitor numbers

@yali Thanks for your valuable response.
We have enabled gaCookies context and will be monitoring com_google_analytics_cookies_1._ga column once data start coming.
Also we are in talk with management on talking help from Snowplow support, so it will be useful if you can share details as well.
Thank you

Great stuff Vivek - let us know what you find when comparing the cookie IDs!
All the best,

Yali

@yali
So, we did analysis on ga cookie IDs for data we got yesterday and what we found is:

  1. when domain_userid changes, _ga changes as well which is a good thing
  2. But, there are many cases where _ ga is null but we are getting domain_userid so that makes distinct domain_userid != distinct _ga.
    What are your thoughts about this ? We will also continue doing some more analysis on ga cookie id.

That’s really weird - can you do some digging into in what situation a Google cookie isn’t available…

i think it’s the result of blocked GA-Tracking (by optout, addons …).
Over a long period we can see that GA is reporting between 10-15% less user then we track with snowplow.

@yali I think what @ecoron thinks may be correct !

Great you were able to bottom it out! So looks like Snowplow’s reporting more accurately than GA :smile:

@yali True :slight_smile:
I have also pinged you my mail id. Our team is looking forward to discuss more about snowplow.

Maybe you check also, how both trackers are implemented in the page. I would recommend to implement both trackers in the head section of a page, close together, to get most accurate results.
Sometimes i see cases where the Tagmanager(i wouldn’t do that) is used or where one tracker is in the head and the other in the footer.

@yali @ecoron
one more thing, do android platform won’t have domain_sessionid / domain_sessionidsx ? Because we are not getting these values for android.

Hi @vivek291836,

The domain_sessionid and domain_sessionidx are indeed specific to the Javascript tracker. On iOS and Android, we store this data in the client session context: atomic.com_snowplowanalytics_snowplow_client_session_1 in Redshift.

Documentation: https://github.com/snowplow/snowplow/wiki/Android-Tracker#12-client-sessions

In a future release of the Javascript tracker, we’ll start doing sessionization in the same way as we do it on iOS and Android.

Christophe

@christophe atomic.com_snowplowanalytics_snowplow_client_session_1 only have iOS entries. Joined this table with events table but could only find iOS.

@vivek291836 What version of the Android tracker are you running? You’ll need to be on 0.5.0 or later. Also, do you have client-side sessionization enabled in the tracker? https://github.com/snowplow/snowplow/wiki/Android-Tracker#12-client-sessions

@christophe we are using tracker 0.5.2.
There might be possibility that android team has not enabled the tracker, will contact them and get back to you.
Thanks for this useful information !

@christophe
There is one more issue our team is facing i.e. how to distinguish between desktop javascript and mobile javascript. Any insights on this ?

Hi @vivek291836. It depends, if mobile users are served a different website, I’d recommend setting a different app_id. If it’s the same website, then dvce_ismobile should tell you whether a device is mobile or not.

Christophe

@christophe we tried using dvce_ismobile but it was giving us huge variations. For example, we tried this for pageviews:
combined desktop+mobile gives variation of 2-3% compared to GA
Now, if we use dvce_ismobile as filter:
Desktop - (+25%) variation
Mobile - (-22 %) variation
We are using same app_id as serving a same website.

@christophe @yali @ecoron @alex @Gal @13scoobie
Is there a way we can distinguish between desktop traffic and mobile traffic from name_tracker column in events table. We were able to do the same for non-javascript traffic ( in name_tracker we receive information about different non javascript platforms) but we could not implement the same for javascript traffic.