The NY Times solution, for example, is to target ads based on relevance to content on the page being delivered. No tracking required, but its relevance to the page can sometimes be more effective than tracking ads.
Conversely tracking ads are not necessarily smart.
Go shopping for a fridge, then buy one. You'll have fridge ads following you around even though you're not in the market for a fridge any more.
> Go shopping for a fridge, then buy one. You'll have fridge ads following you around even though you're not in the market for a fridge any more.
This persistent meme commits the fallacy of believing that every person follows the precise same fact pattern as yourself: Buying a fridge on some predictable, perfectly recurring and invariant schedule.
In fact, it is probably far more likely that a person who just bought a fridge will be favorably economically incentivized by an ad for a fridge than a randomly selected person.
For one thing, we know that this is a person who will ever influence a decision to buy a fridge (because they in fact did buy a fridge). Many people will never select a fridge in their life (for reasons that include renting a home, moving into a home which already has a fridge, replacing a non-working fridge with whatever a repairperson selects and is thus not influenced by an ad, is not the person in their family who selects a fridge, etc.)
A person who just bought a fridge is more likely to return that fridge and buy another than a randomly selected person is likely to buy a fridge in the same time period.
A person who just bought a fridge is more likely to be a decision-maker in selecting another fridge in the near-term than a randomly selected person (for reasons that include: buying another fridge for their garage, buying one for their business, recommending one to their contacts).
Buying a fridge is a rare event that probably absolutely correlates positively to buying another fridge.
I don't know anyone who has ever returned a fridge. I know plenty of people who have bought one though. Can't say I find your argument all that convincing as a result.
A more effective ad targeting model would advertise fridges to the friends of someone who recently bought a fridge though, due to a) keeping up with the Jones' effect, and b) friends are likely at similar life stages, ie. getting married, buying a home, etc.
While this is true, it's more likely that the fridge buyer is also interested in an oven and some other kitchen appliances, or even a complete kitchen.
The fridge ads are mostly interesting for someone who has searched for a fridge (either on Google or on the fridge seller's website) but hasn't bought one yet.
Google’s publisher network used to be, as best as I could tell, entirely contextual to page content. I used to see CPMs anywhere from $5 - $1000 depending on the content.
This happens to me all the time. A few weeks ago I registered for a half marathon and I’m still seeing targeted ads for it. As if I would somehow be able to run it twice lol..
The NY Times solution, for example, is to target ads based on relevance to content on the page being delivered. No tracking required, but its relevance to the page can sometimes be more effective than tracking ads.
Conversely tracking ads are not necessarily smart.
Go shopping for a fridge, then buy one. You'll have fridge ads following you around even though you're not in the market for a fridge any more.