How long should AdSense test/experiments be?

Hi Everyone!

I set up my AdSense account yesterday and I received a little over 1$. Nothing fantastic, but I'm confident I can bring that figure up through increasing traffic and optimizing my ads. My CTR for a page is usually below 1%. Although my blog isn't product or particularly niche focused, I'm hoping I can get that into the 2-5% range.

The resounding cry (aside from focusing on content) seems to be on experimentation. Now for my question:

With a website that still has fairly low earnings, how long should I run a test with a consistent ad format in order to be able to get valuable results from the trial?

For example. If I decide to switch from a skyscraper to a horizontal banner (purely hypothetical) how long should I leave the test going before I try to draw conclusions about what is more effective. I'm worried if I try to do too many changes too fast I will incorrectly attribute the success of one format over another.

How many impressions or clicks do you usually feel is a large enough sample base to make an educated guess on the effectiveness of an ad positioning/color layout?

Thanks for reading and have a great day,
-Scott

 

I would say at least a week, so that the weekend/weekday trends get ironed out. Comparing day to day performance tells you nothing because of the huge variations. Unless you get a lot of impressions/clicks every day, you may need to run experiments for as long as a month.

I've found that if an experiment is going to work well, it'll be clearly seen to be working after a week or two. (Disclaimer: mileage will vary!)

Cheers, Cryo.

 

Thanks for the tips

 

yeah, atleast a week and few hundread page views. Stats can be totally misleading with just a few page views.

 

For small changes (placement, ad format) about 1000 pageviews.
For big changes (site design or major content changes) about 2 weeks.

At least that's what works for me. Plus: tests in weekends are NOT relevant, tests shoud be either in a weekday or in a complete week.

 

Thanks. I'm thinking of coming up with a weeklong schedule where I can list possible optimization strategies and then test them in the upcoming week. To be honest, I probably would have tried switching them every day if it weren't for you guys! Thanks for the tips.