Do you mean a banner in the header section (at the very top of the page or do you mean just above the forum index? Something to consider, and I've never really thought about this before, is that ads at the very, very top of a page are subject to falling under the dropdown list embedded in the browser itself (in fact, much of all page content falls under this scenario as the page is scrolled).

The top banner is in the "header" section of the code. There are no drop-downs at all in my phpbb pages. Users can click on "view unread posts", etc. but they are well separated from the banner. And the banner is pretty well isolated. It's not impossible for a user to make an accidental click... and with a tablet or smartphone set to the default style it might be easy due to size of fingers. But I don't understand why it would have suddenly started around the first of the year.

The bottom banner is very much isolated from any page controls that I have control over. I think a visitor would have to be a real klutz to make an accidental click on it, quite frankly.

You can tell what units are losing clicks if you're watching performance reports close enough. If your top banner is loosing a lot of clicks and the bottom banner is less subject to this problem, this could be part of the problem. Also, the phpbb main menu (on the main product site) has major dropdowns, if your site is like that and your ads are under the menu, that could also be an issue. And yes I could interpret "top" to mean at the top of your forum index which is under the main menu. So, just trying to get a clearer picture since all we have to work with here are words.

Google scrapes clicks from all the ads (top banner, bottom banner and mobile top). There does not seem to be a pattern and in order to discern a pattern I would have to have a lot more time to hit the refresh button every minute or two.

The bottom banner tends to gather fewer clicks. Right now I have 4 clicks on the top banner, 3 on the bottom banner, and 2 from the mobile.

Remember that on the 27th I had 5 clicks for quite a while, then they were scraped off one at a time... then a couple hours later there were suddenly six clicks... and finally all of THOSE were taken back.

Really, this doesn't look like accidental clicks.

Well, I've never been a proponent of stat-watching but, in this case, it's almost the only way to get at least a partial handle on the problem. When clicks disappear, they're gone and then whatever channel or ad unit information you might have gleaned is gone with it.

If you see clicks in your performance report, you could simple record them by unit during the course of the day and then note them as they disappear. You don't need to catch every one. Just looking for general trends. Using reverse psychology and Murphy's law, the broken thing never breaks when you're showing it to the repairman so the problem might mysteriously stop if you do this (said with a big "yeah right" attached). Seriously though, it might be insightful to learn if the top unit is more impacted by this than the bottom unit. You say all units are impacted so I can't argue that it might not be something else. I don't use ads in my banner section and I don't have the degree of this problem that some do. I am wondering if that top location is part of the problem but there's also another thing that could be at play and that is the nature of your visitors. Forums attract a lot of repeat traffic and repeat traffic can be harder to monetize than one-time bouncing traffic sometimes. Also, visitors who return to your site over and over are a pattern unto themselves and if someone is clicking on an ad every time they visit your site, this could start to be taken as abuse by Google. Even a friend who just thinks they're helping you out could be causing this. Repetitive behaviors like that can be big red flags in an algorithm such as Google is using. More food for thought.

Yes... I think you're right about jotting down the data. Right now it's haphazard but I'm not sure I can make it more ordered given the fact that I have a day job. :P The only plus is that I work with computers all day long and since my job involves routing and DNS and such I have the pleasant ability to override the blockages inflicted upon mere users.

I will create a manual database and see what transpires.

The idea of repeat users "helping" has also occured to me; which is why I've been careful to never mention anything about clicking on ads. The ads, for the most part, are remarkably well linked to the content and/or interests and since we do get about 40% repeat visitors - and they tend to be obsessive about riding and upgrading their gear - it's not at all unlikely that they could be looking at the ads.

Gee... you'd think that this is exactly why the advertisers put ads on web pages. (And my site description is honest about it being a forum so that advertisers can deal with that as they think best.) But a Google minion toiling away in the cubicle-catacombs might not think that way.

I do have the sense that the top banner is being scraped more than the others; but the top banner is clicked more than the others, as well. Only keeping track for a week or so (or longer) will tell.

That, of course, will give rise to a whole 'nother set of questions.

Given the clear decline of the Adsense/Adwords revenue stream I am starting to wonder if Google didn't implement a disastrous algorithm and now can't figure out exactly which part of the code is responsible. Certainly, unnecessarily scraping clicks (and, presumably, refunding Adwords money back to the advertisers) would not be good for revenue.

FYI: yesterday and today have been remarkably successful. Yearly highs, in fact. Although clicks are rising, CPCs are all over the place and changing from refresh to refresh making "earnings" go up and down like a yo-yo. So I suspect that clicks are being scraped between refreshes and I have no way to track those.

March has started off remarkably well, well above average at the moment. I think we'll be having turkey again this Christmas if things carry on like this