TrueView video ad campaign causes overcount by Youtube Realtime Analytics
I am running a TrueView in-display ad campaign for one of my videos. When I first started the campaign, the view count in Youtube Realtime Analytics almost immediately shot up by several thousand. Unfortunately, the verified view count ended up being about 20% of that.
Since then, I have paused and restarted the campaign a few times, and each time I see the same surge in Youtube's realtime count, only to have the verified count amount to a fraction of it. Youtube's verified count and Adwords reported clicks coincide, so there's no issue of being overcharged. However, I am mystified as to what is causing Youtube's inflated realtime count.
Is Youtube somehow mistaking Adwords-inititiated impressions for views? Are my ads engendering an inordinate number of partial views? Is there a legitimate surge in views, the suddenness of which Youtube mistakes for spambots?
Given the linkage of Youtube and Adwords, I would think the former's algorithms would be able to recognize and adjust for the effect of latter's campaigns, without giving the customer a false sense of impending virality. Even though I am only paying for verified views, the towering bar graphs of Youtube's realtime page hold out the promise of thousands of views per minute and keep me funding campaigns which I might otherwise abandon. This seems a little unfair.
Re: TrueView video ad campaign causes overcount by Youtube Realtime Analytics
Hi @Big Bad B YouTube stats have been a problem for AdWords pretty much since the two were linked - I'm afraid I've no idea what the problem is but it's clearly something tricky to solve.
Regardless of the actual channel, it's never a good idea to make AdWords decisions based upon realtime, or even recent, data. AdWords reporting isn't realtime for any channel and even once the data comes in it can change. Personally I wouldn't make any decisions based upon any data that's less than a week old, and covers at least a period of a month, preferably longer. So in this case I'd advise ignoring the real-time performance graphs and concentrating instead on longer term performance data.