Forum Moderators: martinibuster
Identical subject - say, family ski vacations in Elbonia - targeting same keyword(s): Elbonia hotels, Elbonia lodging, skiing the Elbonian mountains, etc.
12 landing pages on the "same subject". Verbiage sufficiently different to avoid issues of duplication.
12 landing pages varied to 4+/- different overall page layouts: Article left, article right, etc.
Amongst those 4 common page layouts you vary the placement and/or number of the AdSense boxes.
Results?
Anyone uniformly apply a systematic approach to the recurring issues of "better placement"? Of course.
I don't recall reading threads here that went into details about A-B-C-D testing, though I know that "testing" has been repeatedly discussed. Can't say I blame anyone for not sharing the detailed results of their testing efforts, especially since they went to the time and trouble, at their expense and difficulty, of testing.
Anyone care to discuss their version of testing methodology, if not their results? Just how sophisticated is your testing?
Google ever step up with instructional materials about either "what works for what how" or "here's the accumulated knowledge on the subject of A-B-C-D"?
Since placement is a recurring there is there a decent roadmap for "best testing practices"?
What are the "constants" and the "variables" of testing placement of ads in relation to content, and of headers in relation to ads and placement of? in relation to?
What are the "controls"? How far can you or have you gone using channels for tracking placement?
Has there been an ROI for the time you invested in testing?
Ever test the heck out of things, think you got it right, and then the bottom dropped out on clickthroughs? The explanation for that was what?
You need to keep track of what you test and the results (using channels) and continue tweaking until you hit the sweet spot, and repeat it for each site.
With 12 landing pages, it is hard to test unless you are generating traffic from Adwords or Overture, because traffic can be different. And you can even see a difference in CTR when you crunch the data down to a referrer level (ie. visitors who come from one search engine might have a higher CTR than another; visitors who come from a SE vs. an internal referrer might have different CTRs). And that data can be different when looking at a site by site basis.
>> Has there been an ROI for the time you invested in testing?
I don't look at it from an ROI POV, because theoretically, once you find the sweet spot, it will pay for for months or years to come.
Anyone care to discuss their version of testing methodology, if not their results? Just how sophisticated is your testing?
I think you will find many reluctant to give up specific stats and strategies based on their testing, at least not without any compensation for that knowledge and information. But I have discussed the basics of exactly how I go about testing when speaking at the WebmasterWorld conference and the Google AdSense Bay Area Forum.
Never had a CTR bottom out, so I can't answer that. CTR is generally pretty static unless there are other outside variables. But it isn't generally a case of "this positioning/colors/etc has a 20% CTR then one day drops down to 5% and stays there".
Good testing for AdSense is pretty complex, as you can probably tell!
On a section with 120 pages I've run the following for 7 days each, (20,000+ plus impressions for each layout)
The starting point was a 488x60 placed just below the primary image on the page. (Keep in mind the images are what draws the most traffic to these pages.) This position puts the ad block starting just above the fold @ 800x600, and the whole ad shows above the fold @ 1024x768. The page layout had a 2 column design at the start.
All layouts used a rotating color background color that contrasts mildy with the page, pastel colors on a white page background. All used a border.
468x60 = (right side column) "x" CTR
300x250 (same location) = 1.3x
728x90 (at the top of the page, spanning both columns) .5x
250x250 (in the left column, slightly higher up the page so it was adjacent to the image by about 50%+/-) = 1.1x
120x600 in a new 3rd column. (Started just under my main navigation.) = 1.1x
CTR is dismal on these pages in general. Probably because folks are looking at the pages as a part of a trip down memory lane or a hunt for a reference/id photo, and not with any kind of purchase driven purpose.
I may try one more layout change, but again it would require rebuilding the page layout completely to accomodate the change. So far, redesigning the pages hasn't seemed to be worth the effort.
If both A and B go down, if A goes down significantly more than B, B's the winner. Conversely, if both go up but A goes up significantly more than B, A's the winner. (Obviously, if A goes up while B goes down, A's the winner.) You should look at relative results between the two, not absolute results.
Better, apply the A style versue the B style randomly. That is, when generating a page, there should be a 50-50 chance of going either A or B.
Better still, regenerate the pages daily.
Attempt to randomize A versus B across time and space (pages) as much as possible.
With each page changing (or not) from A to B (and back again, or not) from day to day, obviously you can't track individual pages. Track aggregates.
Doublecheck everything. Be very careful with your channel IDs. If you mishandle channel IDs, your results are worthless, or worse, misleading.
