I don't think the article says advertising never works, but that people should be careful when measuring. Lets stick to the mentioned eBay example. I can totally imagine advertising for the keyword "ebay" not being worth a lot. Those people are likely to visit that site anyway. I was more surprised when "shoes" & co were mentioned. I would have expected bigger impact for those.

Hardly any of that applies to a small store with no brand awareness, since there is little/no of the "would have clicked anyway" mentioned in the article.

Another problem with measuring can be the competition. Building/Changes brand awareness (incl relative to competitors) takes time and seems unlikely to possibly be fully measured in a short experiment. Especially if the competitor currently doesn't run ads... but you not betting for them anymore might change the price and thus calculation for them.