Website Optimizer is a free and self-service tool. Without extensive experience or resources, you can run multivariate experiments on landing page content, including headlines, promotional copy, and images.
[services.google.com...]
As far as I can tell this is a pretty basic testing system though. There are better methods and tools for optimizing landing pages (nothing can beat solid integration with your website's back end). I'm also pretty sure this system isn't using the Taguchi method, which many believe to be the best method for optimising landing pages. I haven't actually used Google's landing page optimiser yet, so my guess is based completely on what I could gather from the demo.
All in all I think this is a very welcome addition.
We also have a problem in that although our traffic is reasonably high volume, its not exactly Amazon, so it does take a while to reach statistical significance on many more subtle changes - I wonder how well this will work for smaller sites who have reasonably low numbers of page views. Or will it just let the user decide when the numbers "feel" good, which would be rather pointless.
Its an interesting direction for Google to go though... I guess its another data collection exercise on their part?
My approach, when I can do it, is not to "finish" a test, but continue to randomly and at low rates sample the "discarded" C/D/E/... cases to see if any of them start to perform again. Eventually I may discard the tail of the very poor performers to reduce clutter.
Practical cases in point:
1) I rarely delete my AdWords ads unless they have disasterous CTR and cost-per-conversion relative (say) to the top 25%. After all, variety helps combat ad-blindness and what works one day, or for one part of your target audience, may not work for another. I choose the AW "optimise" option to show the "poor" ads very little so it doesn't do me any significant harm at worst. (I do delete ads that become "wrong", eg with broken landing URLs.)
2) I continually rotate/test page/ad layouts, though clearly show the best-performing most of the time. I do sample all possible layout combinations for my site randomly. And some really astonishing (to me) successful combinations have emerged over time and helped increase my eCPM 5x. That test is not going to stop, on the grounds that conditions may change again; the continual testing should help locate a new optimum automatically.
In any case,having more tools to help test objectively is good.
Rgds
Damon
I know there are some companies that are worried this morning. This is a great tool. Let's see how well it works.
back end info:
target market : 14 to 30 years old
test goal : what colors work for that target market
banner file size : 3K to 5K
banner dimensions : 468 x 60
all banners text was the same font and size
time of sample 1 month
all I did was change the color background
and had a total of 16 different banners.
What I learned was the following :
a) there were 3 very active colors
b) basic prime type colors don't work
c) shading ( don't recall the right word ) worked.
My best banner had a click rate of over 7.5% the next was about 4% and the one after that had 2.5%.
so I would guess that a company could run the same test with a landing page and optimise it just on the color results, the go about checking the feel of the site to get the proper click distance.
Thanks in advance. ;)
AWA
What is 1MM pages views a day? 1 million main pages/month? 1 massive mega page/month? 1 mainframe melting page/month? Have not seen that acronym I must say. Interesting though that there would be a difference based on color alone however, that would be so pronounced at least.
It would be nice (and no I still have not signed up for the beta) if the software could adopt a survival of the fittest approach to under performers. Perhaps drop off the bottom 3 performers at a certain stage, it would be more/most helpful with the more variations you are running perhaps.
>>What is 1MM pages views a day?
good question, let me break it down.
1mm = 1,000,000
page views = full page loading + graphic ( banners and images ) across te entire web site.
>>1 million main pages/month?
don't recall the details
>>1 massive mega page/month? 1 mainframe melting page/month?
funny terms
>> Have not seen that acronym I must say.
MM acronym is what I learned in grammer school it a thousand thousand also as perr wiki = A confucing way for million
Interesting though that there would be a difference based on color alone however, that would be so pronounced at least.
It would be nice (and no I still have not signed up for the beta) if the software could adopt a survival of the fittest approach to under performers. Perhaps drop off the bottom 3 performers at a certain stage, it would be more/most helpful with the more variations you are running perhaps.
>>Interesting though that there would be a difference based on color alone however, that would be so pronounced at least.
If you can get your hands on the Las Vegas Color Report produced for casinos, then you can see the revenue of different colors ( besides a ton of other stuff )
>>It would be nice (and no I still have not signed up for the beta) if the software could adopt a survival of the fittest approach to under performers.
I did it by hand, if I was in the game again, I would take my log files, dump them into excel. run a simple AI program ( easynn is an excel add on that I have used ) and run some simulation.
problem is that it's a ton of work for most people.
Mojomike
side note : re-read ( or read ) Matt Cutler's pdf E-metrics, it my e-business bible, best 60 pages I have read in a long time.
This is the best tool in a long time that will help agencies add lots of value to clients re implementation & analysis.
Clients need this data but haven't a clue how to go about it. Companies like offermatica & vertster are not going to be happy.....
Well done
J
How much work hmmm....
my ideal working situation would dictate about 25000 page loads ( on landing page ) per day and the test would last about 200000 land page - page loads.
some math, if your click rate is 2.5 average per day, a jump to 3.0% click rate should yield about 20% increase in revenue. again this is just a theory, based on just having people click ad's. if you are talking signup's to list then you have to make extra effort into cookie tracking and have the entire site with the same color arrangement.
now it should not take more that 5 hours to make 10 different landing pages.
Now where I see a problem is :
I think that you might run into the Google duplicate content issue. that definitely needs to be looked at and confirmed.
