Hacker News new | ask | show | jobs
by Nextgrid 1269 days ago
You can have ad models that are immune to fraud. A time-based model such as "your ad here for X days" is immune to fraud, just like ads in print or TV are immune to it. If the ad drives purchases, the advertiser renews; if it doesn't, you may have to lower your price until another one comes up.
1 comments
jefftk 1269 days ago
It's not immune to fraud: if I offer you ad space on jefftk.com how do you know how much to pay for it? Without some kind of fraud detection, how do you know whether to trust me when I tell you I have 25k unique visitors monthly?
link
Nextgrid 1269 days ago
Could always be an auction-based market - you start with a trial and only do a day/week/etc and gauge your results - if they're good you know you can bid more next time. With enough "liquidity" on either side the true price of said ad inventory will naturally come up without any tracking necessary.
link
jefftk 1269 days ago
That works for performance advertising, where you're able to tell whether your specific ad is getting results. But most display advertising (by dollars) is brand advertising, where Ford cares that a real person is seeing their ad but isn't expecting you're going to drop everything and click through to buy a truck.
link
Nextgrid 1267 days ago
How has that problem been resolved in offline (print/TV/display) advertising be solved, and could those solutions be reused for the web?

It seems like the free-for-all with regards to personal data and decriminalisation of spyware has led the online ads market into a near-optimal situation with regards to targeting & fraud detection, but 1) is it a trade-off people are willing to accept (the GDPR being enacted suggests not) and 2) can there be alternatives that give both sides what they want?

link
jefftk 1267 days ago
The traditional way to handle this in the offline world is with ratings agencies like Nielsen. This is the thing where they try to figure out how many people are watching each TV channel or who read a given paper.

The problem is, this requires the readings agency to build up a representative panel of people, and track their behavior. This is pretty coarse, but it works in a relatively centralized world like traditional TV where there are only a few dozen channels. On the web, however, people visit so many different sites that the panel would either have to be extremely large or you would only be able to generate reasonable ratings for the largest sites, probably both. This would be yet another force pushing hard toward internet consolidation.

Another option is that you could somehow build a new technology into browsers with some sort of privacy preserving API. I used to work in this area [1] but I am pessimistic about it: it's very hard (and may not be possible) to build something that gets all three of (a) minimal load on the users browser (b) actually useful fraud detection and (c) sufficiently private (with a separate question of whether usage would require consent under GDPR).

[1] https://privacysandbox.com/open-web/

link