How to get a 20% response rate with your banners (VII).
Cosmic Cat - A cosmic, free game
Free American Roulette
Free European Roulette
3 Card Poker Gold, Free
Sports info and betting
Independent funding for a free lifestyle
|Therefore, for each of your banners you should plan two kind of samples: a first one of about 1.000 exposures in order to eliminate your weak designs or tweak the variables that define their exposures (i.e. in targeted advertising, changing your target), and a second one, complementing the first one, of about 2.000.
In such a way, you will get a fairly precise estimation of the effectiveness of each one of your banners while keeping your investment as low as possible.
Another point to consider is how many different banners you should start with, and how many you should accept in the end.
In most cases, you would be better selecting two different banners as finalists, and from these two, use the best one as your main tool and the second one as a reserve, but in order to achieve such results, you should begin testing a bigger number.
Again, based on the data gathered by the search engine that I manage, five seems to be the minimum acceptable number. In other words, you should design five different banners before testing.
Once you have them, you should start testing each one of them and comparing the results obtained for each design.
Your tests should be eliminatory; that is, on each round you should eliminate the worst designs and keep testing the best ones. Just remember that you should not change the environment of the test while it is still running. Otherwise you would not get reliable data because the conditions under which the banners are found acceptable or not by the public would not be the same.
The Search Engine for Exploration, Survival and Adventure Lovers © - Andinia.com ©