The topic of "accurate metric" is also challenging. When I worked on this, I would often find discrepancies of up to 5% depending on the technical implementation of the metric. There was never a clear explanation, just that the sequence of events from mouseover to click to redirect to more redirects to the final landing looked more like a very wide funnel with leaky correlation than the straight pipeline you might naively assume.
For instance Google has a blocker for "annoying ads" in Chrome. Needless to say it is never going to block an annoying ad that is served by Google. The whole thing is asking for a lawsuit.
If Google is aggressive against adblockers it similarly becomes a target in a way it would be if there were 4 major players that each got 20% of the market and then a bunch of small players.
The most interesting thing happening in online advertising today is the pitch made by the video or podcast creator that is embedded in the content. I've got an emotional connection with Nate Silver or Ana Psychology that makes me more responsive to their pitch than the average pre-roll.
Harder to block, so I manually skip it...
If ads were served by the publisher than the advertiser would always think that the publisher was cheating and claiming they showed more impressions and generated more clicks.
If ads were server by the advertiser then the publisher would always think that the advertiser was cheating and claiming that they showed fewer impressions and got fewer clicks.
Both could believe that a third party is honest. Both will feel even more assured if the site has 50 different trackers from 50 different third parties tracking the process.