You are wrong. The second paper is attempting to measure sales lift (with randomization, which many ad campaigns disallow because the marketer doesn’t want to pay for holdout inventory, so it’s not really the same situation that most companies are in when calculating lift), but sales lift does not mean there was positive ROI. You can increase sales all day, but if you’re spending more than you’re making to do so, it’s not effective and usually indicates some unsustainable effect that is more about short term attention than about a meaningful effect of information transfer about a product (and indeed in [1] the authors found a very similar effect - that there is usually positive lift on a customer activity like visits or site shopping cart checkouts, but that among users driven to the site by digital ads, their purchases were far less in total value, enough to more than offset the cost of advertising to drive them).

[1]: https://marketing.wharton.upenn.edu/wp-content/uploads/2017/...

I think you’re confused. I’m not disputing whether ads create positive lift in intermediate metrics like visits, clicks or sales.

I’m saying the research pretty conclusively shows that despite that created positive lift in intermediate metrics, it does not translate to positive ROI, or a weaker version is that it may create positive ROI but on average marketing campaigns even from the country’s biggest marketers would need to be 10x-1000x larger than they are to even be able to distinguish such wildly incompatible outcomes as +100% ROI vs -50% ROI, to the point that for most companies buying digital ads, they literally have no idea if they are making money or losing money by doing so.