Ok, let's talk about the business model involved in detail, starting with the decision process of a typical ad buyer. It starts with the value of a sale or conv

Ok, let's talk about the business model involved in detail, starting with the decision process of a typical ad buyer. It starts with the value of a sale or conversion - say, your company makes $100 with each mattress they sell. You can then extrapolate based off past data how many people you need to do specific upstream events in order to get that sale. Say, 100 people need to view your website. Something like that. So you turn around to some web advertising platform and buy an ad, and you spend something like $30 dollars to buy however many website visits it takes to drive one sale, and after adding a generous fudge factor to take into account all the uncertainty and mis-attribution you expect you wind up with a profit. There's both cost-per-click and cost-per-thousand-impressions pricing, but both end up feeding into "buy $X of advertising, which flows through our sale funnel resulting in $Y of profit, and $X is much less than $Y" at the end of the day. Why do these advertising platforms collect data? So that instead of a generic "show this ad to whoever visits the website", you can break viewers into different categories and sell those instead. If it takes 100 website visits from a generic user to drive a sale, maybe it takes 50 instead if they just performed a search for a mattress. Or if it takes 200 impressions to get a click-through, maybe it takes 100 if it's done from a specific geographic area. Or if you end up selling to a specific job title, the SaaS contract tends to be more expensive. Whatever it is, certain slices of traffic are more valuable to the ad buyer, so this ends up flowing through to the advertiser's return-on-ad-spend math, which flows into higher ad bids. And so the advertising platforms collect a lot of data, so that when a user can be shown an ad, the platform can figure out what targeting options allow them to make the most money. If you follow the math again back the other way through, it implies that there's a pretty powerful constraint on how relevant advertising can be to you. Specifically, the pure relevancy metric will get skewed by the producer surplus for converting a sale to you: doubling how likely you are to convert is just as relevant to the business math as doubling how much money is made off your conversion. A system that maximizes relevance of ads will not maximize advertising platform income, since it would fail to prefer a five-figure programming job placement fee over a $10 tee-shirt, and the difference in conversion value flows through to a difference in advertising bids. From a public policy perspective, I'm pretty pessimistic about preventing these privacy issues. There's simply too much of an incentive to section out users into specific cohorts and sell access to them to the highest bidder. In other words, privacy-violating targeting is such a big deal because targeting works - it reliably convinces people to take courses of action that result in more producer surplus in the hands of the producer of the good advertised. The only saving grace is that for certain products, it'd only change which company gets the producer surplus, rather than convincing you to push more producer surplus into corporate pockets.