USGS-VIZLAB / vizlab

Package with utilities for building vizlab pages
Creative Commons Zero v1.0 Universal
25 stars 14 forks source link

explore improving tracking of direct traffic on GA #56

Open eread-usgs opened 8 years ago

eread-usgs commented 8 years ago

e.g., as recommended here: https://megalytic.com/blog/understanding-direct-traffic-in-google-analytics

aappling-usgs commented 7 years ago

According to our analytics dashboard, our main traffic source has been direct for every viz to date, suggesting that the above link might be able to help us tease it apart further. I'm not sure which sources we're missing - our analytics already catch twitter, reddit, facebook, and r-bloggers, so...instagram and email? Or maybe facebook and twitter are only working on some vizzies, not all? This could be worth a little digging.

wdwatkins commented 7 years ago

If we continue using URL tags, this shouldn't be an issue in the future. That is what that link suggests ☝️. We started with Irma I think.

aappling-usgs commented 7 years ago

Sounds good! And yet, for Irma: image looks pretty direct-heavy to me...

Could you share a link to a place where URL tags are used?

wdwatkins commented 7 years ago

URL tags are just attached when you visit a site, they don't go in the code anywhere. i.e. https://owi.usgs.gov/vizlab/hurricane-irma/?utm_source=reddit-sci. So it is really just part of the social media outreach. I suspect a lot of this direct traffic is just us testing, since it is disproportionately from WI and that part is >90% repeat visitors: image Since all the traffic for recent vizzies was just the gif on reddit, it stands out more. We could eliminate this by either a) always using a tagged url when we test, e.g. https://owi.usgs.gov/vizlab/hurricane-irma/?utm_source=test, or b) Set up analytics so they only fire on prod

Any time we distribute a link to anyone though, we should be sure to tag it.

wdwatkins commented 7 years ago

Setting a start date to that data to when we actually release the viz would probably drop a good chunk of that I would bet.

aappling-usgs commented 6 years ago

We just had a team slack conversation about the suggestion to "b) Set up analytics so they only fire on prod":