A partial archive of discourse.wicg.io as of Saturday February 24, 2024.

New referer headers

OpenLocationServices
2018-08-18

Browsers now refer to user using different ways. However for analytics purposes it becomes difficult to know where the user came from.

There need to be new referer header for:

  1. If the user inputted the url directly.
  2. If user came from the suggest sites tiles from the browser
  3. If the user came from a bookmark

So that becomes:

  1. “referer: direct”
  2. “referer: suggested”
  3. “referer: bookmark”

If there are any additional headers that I am missing, please let me know.

mkay581
2018-08-20

Yeah all of those are currently classified as direct referrers right now. What’s the use case for more granularity here?

OpenLocationServices
2018-08-20

Figuring out if people like your product? A bookmark hints that people like your product hence it is a motivation. It would help apps and companies make better content and figure out what people like? Suggested would hint that the user just came now or was suggested, so he has never been here. Gotta make the best impression? Direct would help know if users memorized your domain name and your brand is successful?

There are lots of ways to judge and help developers make better decisions.

mkay581
2018-08-20

You can find that out by not having this feature :stuck_out_tongue:

Not necessarily. That calls to a user’s state of mind which is impossible to determine accurately. The user could have just bookmarked it because they didn’t have time to review and want to later.

Again, they can do that already through analytics and direct referrers. What exactly would this new feature allow them to do that they wouldn’t otherwise be able to do? In other words, why would a site owner need to know whether a user came in from a bookmark opposed to typing in the url?

This also calls to state of mind and makes an assumption that isn’t necessarily true. They could’ve copied and pasted it from a direct message. Doesn’t mean that they “memorized your domain name” and certainly doesn’t mean that your brand is successful.

OpenLocationServices
2018-08-20

You’re missing out machine learning from the picture. More data = better picture of what people want. All of the information is some how interconnected. When you connect the dots, its easier to figure out what your core audience wants

What do you mean by the analytics. I dont want google analytics on my site. Only first party information. I dont like 3rd parties knowing about my customers.

Possibilities are there, but most of the time we can figure out what is happening. Why do we have referer headers? To know who is sending us the audience. Connect without apps and webapps sending audience so that we can collaborate in future. You are not looking at the bigger picture. Not necessary what I said about the conclusion is correct. But when you get different engagement metrics, you know something is wrong or right.

Garbee
2018-09-13

Businesses have done very well thus far without this granular information on making better content. Machine Learning is not some silver bullet that will magically be able to apply this data in a new way and make something meaningful out of it. Instead of assuming it will find a way, show how it finds a way. You can feed mock data into a ML program and see what it spits out. That will be perfectly fine to prove your point that ML can make use of this to improve a business instead of asserting it maybe could might be one day possible.

There are numerous analytics providers. Even some such as Piwik that are self hosted so your company/organization is always in control of user data.

The bigger picture they are seeing perfectly. The goal of any site is to produce better content, period. It doesn’t matter how users get to you in this level of detail. It only matters that they come and that the content is continually high quality. If you content isn’t good, it doesn’t matter if they bookmarked you or it was suggested somehow. People will just leave you then and not return.

Focus on content, not meaningless data-points.