I think so because the entire history of advertising is based on hiding the ball in terms of ROI. How many sales does a Super Bowl ad generate? How many does a billboard generate?

For some extremely simplistic e-commerce conversions, such as for referrals to SaaS, it is easy to measure. For entire swathes of consumer products, the game is brand recognition and other soft metrics that drive huge ad spends. You can pretend it’s a science, but the incentives are all upside down for a true end-to-end ROI determination.

A large number of companies are advertising /only/ online, and therefore have a pretty pure measurement of ROI via click tracking from the ad to purchase. It's one of the promises of internet advertising: there's sufficient <strike>surveillance </strike> logging to be able to see what's happening with much greater clarity than TV or billboards.

The other half of the promise is that you can target exactly the 'right' people. And this seems sketchier to me, as the article kinda gets at: as targeting complexity increases, it's harder to tell whether problems are arising due to market fit or issues in the targeting algorithms. In the old world, you bought ads on content, and now you buy ads on eyeballs: my guess is that there's an unexplored happy medium between the two.