[gergov]

Marketing ROI is very tricky to measure. On the other end, you would see that the more you spend the more customers you have. The question is: are those new customers due to your increased spend or not? There's all sorts of attribution magic that goes in the background to trace back adcampaigns to converting users (especially across devices). That's why the article mentions that most of the "big fraud" that was uncovered by simply turning off paid campaigns without conversions dropping. That's a pretty big giveaway that something is amiss.

There's an old saying in advertising that we know half of our budget is wasted, we just don't know which half. Ad fraud and bot traffic live in that area, kind of as expected waste.

Then there's the other issue - branding campaigns also cost an arm and a leg, and the most you get is vanity metrics like reach and likes and impressions and whatnot, which have dubious correlations with reality but they look nice on dashboards.

[ekianjo]

Its not tricky to measure at all. Thats why we have marketing mix models for like 20 years if not more.

[indymike]

I used to own a digital agency. I had a client who was the 80 year old CEO of a 120 year old military contractor. He sat through a meeting where his 27 year old marketing manager and I walked through advertising metrics for digital ads. At the end he said,

“This is all very interesting, and will be helpful for Kelly (the marketing manager). I’m going to decide if we renew next year with this: increase in sales divided by ad spend. I’m looking for a number greater than 3.”

[gergov]

The entire problem with ad fraud is that it's engineered to look good on those models.