Ad Serving Algorithm/ probability/ mechanism.
Posted on 2010-08-15
Hello experts ,
Back again after a very long period, just had a solution few days back from experts exchange and m glad to inform that i have been sucessful in deploying my own ad server which is running quite well and on a high. I will straight forwardly put my current mechnaism and my problems that are arising which is urgently needed to be fixed.
1. Current scenairo
I have all those ads of our clients , we pick all the qualified ads , shuffle them, then , since we have targetting enabled , we merge the shuffled ads with all the contextual ads and reshuffle again and deliver one of the ads , this brings atleast a possiblity to balance our global and targetted ads.
As you mught have seen we have no algorithm working in it which really schedules an ad to run and maintain a balance of our total inventory so that one client doesnt complains of getting low clicks while the others get enough of them. In the general ad serving network i have seen that the ads which are set to run globally gets picked up atonce and it continues delivering impression until it replinishes the users balance, and then it comes to the next ads and continues accordingly, This however is obvious that the ad server may have choosen based on the user account balance criteria , say starting from low to high and vice versa. But imagine the situation in this type of algorithms when you keep on delivering impression and no one clicks them and nothing is changed , so you wont be keeping the ads delivered all day long and would pick up some other ad ..this will make other clients unhappy.
First of all i would like to request a simple , realtime , conventional applicable alogorithm that can be easily deployed and should satisy every user clicks based on their account balance. Secondly how to schedule ads to be run on the network , atleast delivering balanced impression/clicks from the total available inventory. I hope i am able to express it out. Pls do help me , anything which is just good will work.