Cookie vs Anonymous user?

Question

Software design

I describe the standard scheme.

The beast went to the site Flicked
through the list of products
We bake cookies We
issue recommendations and select, sort products to taste, improve the search

Then I thought of a more interesting scheme The user

went to the site
We create an anonymous user by IP without email and password, issue an anonymous authorization token The user
flips through the goods
everything in the database as viewed products of the user, albeit anonymous
We select recommendations, optimize the search

What is the advantage? If the cookie is lost, binding to IP
Since the user is anonymous, then everyone with this IP is selected according to the browsing history as soon as the user enters the site
More data, in theory their analysis will improve the selection of results. There will be more date for AI.

Tell me, what about the temporary key? After all, this is an anon user, the token is issued for the duration of the session, if an anon user with such an IP is already in the database, we generate a token for him and then record the user's actions.

But what if the user wants to clear the history? With cookies, just clear the cache and that's it, but here it's connected to IP.
Once a month, crowns to delete all anons and their history of views, or simply invalidate the data by moving, for example, to another table for analytics.

A kind of RFC, I'm still thinking about how this approach is better or worse than cookies.
I'm sure google does something similar to some extent, what sites I get on, how long, where the mouse is, etc.
At least in google recaptcha v3, this is exactly the approach of analysis, collecting all possible data about the animal and issuing the probability that it is a bot.

Answer 1

If the user cleared cookies, then he wanted his basket on the site to be cleared, do not interfere with the user to do what he wants.
ps Google the concept of a digital fingerprint or browser fingerprint
Google a library for the web that implements cookies based on it (of course, cookies will be stored on the server, but what is on the client is the implementation details of the library itself that you better not go into)
We use only these cookies, without regard to the browser, i.e. separate cutlets from flies.