At a high level, marketers measure the effectiveness of ads using metrics like ROI, ROAS, etc. The way those metrics are measured is by choosing a certain attribution model (eg: last click, first click, etc) and attribution window (eg: 1 day view, 28 day click). As you can already tell, it's a bit like accounting - you can get different results depending on which rules you follow. So clearly, there's a lot to criticize.
But, a lot of that criticism goes away as soon as you introduce incrementality and start adjusting for it (randomly split the population, only show ads to one group, then observe the results between the two groups). If the two groups exhibit similar behavior, your product is already getting a good amount of WOM and other sources of distribution, so no need to push that paid channel quite as hard (or at all). Most young companies are not in that position, and the amount of adjustment needed is in the sub 20% range. Whatever the range, my point is that the concept of needing to make this adjustment is well known and understood, and the fact that this was never mentioned in the original article is really strange - it almost sounds like if the involved people didn't know about it.
Here's more info on incrementality: https://www.adroll.com/blog/marketing-analytics/beginners-gu...