Having worked at and with small companies that rely on advertising to overcome obscurity and get business to their products - modern analytics tools have pretty much got this problem solved, the marketing people knew and actively tracked their ROI on a daily basis and were shuffling limited funding around between various forms of advertising to optimise sales.

Every time I see somebody claiming that advertising doesn't work, I just think that there's another person who has never tried to run a small online business.

That paper talks about how advertising experiments require larger-than-expected sample sizes to calculate ROI, because the signals are more diluted.

It also doesn't specify mediums of advertising, because of confidentiality agreements, but hints that it included mediums such as television and billboards which are traditionally harder to track. I don't think it's shocking that out of home media is hard to statistically track ROI on.