0

Comparison is between the Google Analytics script and a "chat widget" loaded in a iframe on the onload event. The iframe is not loaded if the useragentlooks like a crawler. The code in the iframe creates a websocket connection and then a user is upserted in our backend.

I'm looking at one website where we currently get around 2/3 of new users reported by Google Analytics as users in our backend.

With the new Google Analytics UI, it's not clear to me how much crawlers are actually filtered or not (e.g. is Google Analytics always filtering Google Crawler?). That could explain a lot.

Our "tracking id" is saved in localstorage which is a bit more persistent than a cookie, but I can't see it making a big difference.

The extra loading time and the websocket connection can probably explain a bit (old browsers that don't support TLS 1.2, baldy configured proxies, etc) can explain some of it, but 1/3 seems big.

Any insights in Google Analytics metrics or other ideas are welcome.

Rhangaun
  • 1,430
  • 2
  • 15
  • 34

1 Answers1

1

Bot traffic in Google Analytics is not filtered by default. You can enable it in View Settings.

Also it's possible to pass User Agent strings into Google Analytics as a custom dimension and then create filter to exclude sessions based on User Agents that you know to be bots. https://www.lunametrics.com/blog/2015/04/01/eliminating-bot-traffic-from-google-analytics-once-and-for-all/

The second reason of difference may be that GA tracker fires when page starts to load, but "chat widget" when window is loaded. Some bounced sessions (and users) are counted by GA, but not counted by "chat widget".

zborovskaya
  • 1,113
  • 1
  • 7
  • 14