[Seirdy]

> Think of your favorite site with the best experience possible. That is possible because people tested countless times what works, what didn't, what is the most efficient path to a rewarding UX, and so on.

Funny; those kinds of sites are my least favorite. All those colors and buttons are an information overload, and the animations make my laptop fans spin like crazy. Not everyone bought their computer under a decade ago.

Please, blue links and black text aren't evil. We need to make interfaces functional and stop rather than continuously A/B test them to maximize addictiveness ("engagement").

[schwinn140]

You ignored my statement. I asked you to think of your favorite site...not mine. I'm in no way saying that what I prefer is somehow supposed to be preferred by you.

"Think of your favorite site with the best experience possible."

Regardless of the site experience that you prefer, I can assure you that thought, testing, and iterations have occurred to deliver the experience that you personally prefer.

[Seirdy]

One of my least favorite aspects of websites is change and redesigns when the original design worked perfectly well. Given that change is bad when a site has already hit the "meh, good enough" threshold, I doubt that testing and iterations would do anything positive for existing users' experience.

Furthermore, I don't want hyper-optimized experiences. These experienced tend to be addictive, whether intended or not. Using an interface shouldn't feel "magical", it should work. I know that you weren't implying addictiveness or engagement, yet these values are (consciously or unconsciously) prevalent enough in the field that they've lowered my level of trust in analytics-driven iteration. Other responses in this thread should also show that I'm far from the only one who feels this way. Earning back broken trust in these situations typically requires going above and beyond past expectations.

User research is research, and should require informed consent held to the same standards of consent as actual research. In any human research, participation should be opt-in. Participants should be given complete information about analytics and how they will be used (with the option to see source code), own their data, be able to revoke their data, and see conclusions of the studies in a format they can understand. Any questions they have should be responded to before and after they opt-in. This shouldn't be buried in a confusing privacy policy but provided upfront, in a language they can speak. People frequently learn interfaces in languages that are foreign to them, but reading details of user research is a different story: you might need a translator. Otherwise, your sample will be even more heavily biased.

This is a lot of work, and might make analytics more trouble than they're worth.