Google Adsense - Opportunities Experimentation

Ryuzaki

お前はもう死んでいる
Moderator
BuSo Pro
Digital Strategist
Joined
Sep 3, 2014
Messages
6,230
Likes
13,100
Degree
9
Google is constantly shoving this card down my throat on the dashboard for new optimization opportunities:

6ITJ5JU.png

I've always ignored it because, hey, I know better! It's always goofy recommendations like these:

NzvKAmr.png

No, I don't want AMP. No, I don't want Matched Content. No, I especially don't want page-level ads for mobile. Yes, I know one of my ad blocks on a few sites goes to the bottom of the page on mobile, it's in the responsive sidebar. No, I don't want to kill my lower performing ad blocks.

*Screeeeeeeech*

Fonts and Borders and Backgrounds... I knew you could do split tests in Adsense now, but every time I saw these suggestions I thought they just wanted to change your settings without a test. I just never connected those dots. But if you click those suggestions you get taken to the Experiments page.

Some of you may remember that I said in some other thread that Adsense had a new User First beta program where they're trying to auto-optimize site layouts and the robots are taking it back to the earlier 2010-era with double blocks above the fold and cramming ads on top of each other, breaking their own terms of service. I speculated that they're trying to compete with Ezoic and worried about Media.net's fast rise too.

Well, that's obviously beta and not ready for market, but their Experiments seem good to go. You don't get to test layout but you can test everything else. Check this out:

yxEbPwA.png

I'm running 5 tests that's nothing more than switching the font around, away from one that looks like the one I use on my site and back to their default font. The number of days has to do with getting statistical significance based on number of impressions and enough time to rule out time-based variations in the marketplace.

If you click into each experiment you get more data, like the number of days it has been running, the impressions count, and the number of days to go. But you also get a table that shows rows of things that are being affected by the test, like RPM, CPC, etc.

RluAwJY.png

This is pretty cool though. You can see I'm switching from Verdana to "Default font" and getting an increase in RPM there. I checked them all and some are -50%, for example. It's only been 4 days for me so those numbers are meaningless right now.

I imagine if you have a huge site with boatloads of traffic, like a news or viral site, this could have a huge effect on your bottom line, and the tests probably happen pretty fast with a million pageviews a day.

Have you guys played with this? How did it shake out?
 
I have ran this in the past on some basic styling and coloring changes. In a few cases the system found some winners, hit apply to the change and off to the races.

I wasn’t doing enough revenue to really feel like a big winner, but a statistically signifigant winner in an A/B test is still a winner.
 
I've got under a month on some blocks and under 2 months to go on others, but a few of them are showing more data now. We're starting to get into the realm of having confidence in the outcomes. I may stop a few experiments early.

The reason I didn't do some of them in the first place was because I knew what was better. I should have trusted that.

But here's one that's getting a slight boost, and the information it's showing now:

KUnvlJm.png


It's only a 5% boost, amounting to an extra 4¢ per 1000 views, but what's neat is Google is pretty sure that's boosting the quality score drastically for advertisers. I'm skeptical since it's literally just a font change and a 4 cent difference. That number will likely decrease if I let the experiment keep running, which I will.
 
Back