Author
Message
GegaBit
Posts:3311
Senior member
Member since: 2006-11-30
:: Quote ::
Subject: AdSense A/B Experimenting, Now this we LIKE
Many moons back I requested this specific feature from them an behold, today AdSense announced A/B experiments

A/B experiments for ad unit settings - the easy way
[adsense.blogspot.com]

Quote:
This new feature is available directly from your AdSense account and lets you quickly create, run and analyze A/B experiments for specific ad units -- all without needing to make any changes to your ad or page code. It works by comparing your original ad unit settings to a variation you choose, and ultimately tells you which one performs better



ok ok y'all have me tho thank for this amazing feature that should give a face-lift to the monotonous cumbersome task of improving earnings via watching ice melt A/B testing smiling smiley
November 06, 2013 06:15PM
Pengi
Posts:3345
Senior member
Member since: 2006-12-17
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
"So there" to all those who say Google never listens!
November 06, 2013 07:09PM
Ian C. Purdie
Posts:2220
Senior member
Member since: 2008-12-12
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
Now if only they reversed Panda, Penguin, Hummingbird and the rest of the Google Zoo.
November 06, 2013 10:30PM
Pengi
Posts:3345
Senior member
Member since: 2006-12-17
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
Have you got any particular comparisons you're trying GB - it would be interesting to hear whether or not this works for you.
November 07, 2013 08:25AM
GegaBit
Posts:3311
Senior member
Member since: 2006-11-30
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
Yes I do Pengi,
I'm running now what should answer the age old question:
To Blend or Not To Blend

Testing now on 2 channels an alternative in-your-face red ad title text style vs. the usual black text blue link style.

What's interesting, and I wish I could communicate to Google, is that their method of testing this via rotating styles, has its own effect on regular users and should be isolated from the experiment (entropy)
November 07, 2013 12:32PM
DamonHD
Posts:6158
Moderator
Member since: 2006-11-30
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
Ah, yes, the blend vs red. I have blend+red too!

Rgds

Damon
November 07, 2013 06:00PM
GegaBit
Posts:3311
Senior member
Member since: 2006-11-30
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
2 days AdSense quick initial A/B testing findings

1- On busy pages: blended ads works best
busy = lots of content

2- On sparse thin content pages: In your face non-blended different color ads outperform by 67%

This is consistent with my historical test results years back, even though test needs to run closer to a month for certainty.
November 08, 2013 07:09PM
Pengi
Posts:3345
Senior member
Member since: 2006-12-17
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
What sort of sample sizes are we looking at here? How many visits? How many clicks?

Do you get any information about whether there is any difference in the click conversion rate?
November 09, 2013 11:05AM
GegaBit
Posts:3311
Senior member
Member since: 2006-11-30
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
For just 2 days, that would be only few thousand impressions, haven't checked on clicks on day by day level as I'm running the test on 2 small channels. The numbers will be much more significant tallied over a month, but interestingly accurate & consistent so far with what I already expect, even for such a small sample.

They do give you what they call the "confidence level" in the report.
November 09, 2013 11:50AM
DamonHD
Posts:6158
Moderator
Member since: 2006-11-30
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
Very interesting stuff.

Thanks for the analysis.

Rgds

Damon
November 09, 2013 11:55AM
GegaBit
Posts:3311
Senior member
Member since: 2006-11-30
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
Day 4
On Thin Content Pages:

Unblended is getting higher CTR
but EPC seems lower as a natural compensation
to the extent that it is earning less than blended

In other words, standing out on thin pages is working as a concept, but Google's built in "balancing" mechanisms are defeating the concept and even seem to be over-compensating to the extent of lowering revenue & making the test pointless!

This might explain why most pages running AdSense are white background, dark text, blue links like Google pages themselves smiling smiley

Still many weeks to go go before final conclusions..
November 10, 2013 12:16PM
GegaBit
Posts:3311
Senior member
Member since: 2006-11-30
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
One month later & 53,000 ad impressions later:

Nothing !!

Goggle is telling me that standout red ads are ahead by 7% but over the blended ones with a 53% "confidence" rating !!

This means Absolutely Nothing to me, my only conclusion is that my test has failed, and the answer to my question according to Google is that it's a matter of coin toss.

When I have time to waste in the future, I'll just turn on the TV.
December 09, 2013 12:03AM
Joshua
Posts:2831
Administrator
Member since: 2007-03-16
:: Quote ::
Subject: Re: AdSense A/B Experimenting, Now this we LIKE
I'm running an A/B test also. Banner/Text vs Text only. My test has over 100,000 ad impressions. Text only is 3% behind but has 55% confidence. What's on TV?
December 10, 2013 09:29PM

Sorry, you do not have permission to post/reply in this forum.