Seems to be entirely subjective.
I bought a new blade for a bread cutting machine on amazon, the last one lasted ~25 years. My current recommendations: blades for bread cutting machines.
I also checked some Christmas decorations on Amazon, but didn't buy any. So next to the blades I get Christmas decorations recommended. January is a bit late for that, isn't it?
I also bought a single server rack on eBay. Now I regularly get spam emails from eBay reminding me of new offers for server racks. Even the company I work for doesn't have that many, so why?
In short I am not really convinced that these platforms are good at predicting what I want. Rather it looks as if they are good at showing me what I already have.
For instance, I recently bought a new iPhone directly from Apple, and I bought a case for it on Amazon. Now two months later, Amazon is recommending that I buy an iPhone (...which I bought two months ago from Apple).
Buying a case an then waiting two months to buy the actual phone would be a very strange thing to do, but Amazon seems to think I should do this.
And Amazon is definitely not alone: Google has been troubled with "AI" running rampant in search results for years, seemingly with no or very little QA.
But besides doing something like "purchases >$100 are not likely to be repeat purchases," I would think it would be difficult.
They show it because it "works" for them, not because it's wrong and they just can't somehow make sure they don't show something.
Because even if 99% of people don't buy anything twice, 1% probably does for some reason, and that signal just trumps all the noise. (And that's why most of ad spending is bullshit, but since Amazon's cost [even with opportunity cost] is basically 0 on their own site, they'll continue to do this.)
While that's definitely true (I get ads from Amazon for stuff I've already purchased all the time too), in that case Amazon is leveraging their own marketing channels to support their own business -- which has, effectively, zero marginal cost to them.
However, when advertisers buy space from Google or FB, that's not the case. There is a definite external cost to such ad buys.
As such, being able to identify the "value" of such expenditures is (or can/should be) important to the advertisers.
That said, much of advertising is based upon the idea of "top-of-mind awareness."[0]
The idea being that if you are considering a purchase, the brand that comes to mind without prompting when considering that purchase will be preferred over brands of which you aren't immediately aware.
Purchase decisions made on a whim or without any research are usually those which are low cost. Which is why companies like P&G, Coca Cola and like companies focus on "top-of-mind awareness."
If you're going to buy a washing machine or a riding lawn mower, you're much more likely to do research than if you're going to buy a soft drink or a bag of chips.
That said, the same "top-of-mind awareness" can be helpful even for big ticket items, as those brands may well be the first ones about which research may be done.
Identifying how such advertising may impact purchase behavior is a complex topic, and unless you can quantify specific ad-views to actual purchases, doing so is generally speculative and is the subject of a great deal of quantitative market research[1].
[0] https://en.wikipedia.org/wiki/Top-of-mind_awareness
[1] https://en.wikipedia.org/wiki/Quantitative_marketing_researc...
But if you can't see any change in business from 2/3 of ad spend at all it must be fraudulent.