Forum Moderators: martinibuster
That leaves me with some spare channels I'm not using, leading me to this idea:
I'm thinking about creating some channels, 10, 20, maybe 100. The more the better.
If I use PHP and without modifying the code, I can output AdSense code on my page that will either be normal without a channel id, or normal with a channel id (one of the 100 I created).
Using random sampling, I can output one random user with AdSense code with a channel id that has not yet been used for the day. I can use a database to verify if it has been used yet or not.
If that user ends up clicking on the ad, I will cease to use that channel id again for the day. If they end up not clicking on the ad, I could free it up an hour or two later in the day and reuse it again later.
By using a random sample, a unique channel id only shown to one person (per day), and information gathered from AdSense by hand or by script [webmasterworld.com], I can get some very detailed information, including the exact revenue amount for that advertiser in that one scenario.
If 100 random samples were done each day, and the information later extrapolated, that might lead to some very interesting and possibly valuable information about ad behavior and advertiser worth on our sites.
One concern I do have however is that maybe the mear act of using a channel might make the ad behave differently than the others, meaning the information would not be very representational.
If a channel ends up being used twice in that day, then that information would simple be discarded. There are lots of data points to use without it.
I'd love to hear your input and thoughts on this idea.
Thanks.
The concept doesn't scale once you get decent traffic :(
At the 95% confidence level, with 1,000,000 AD CLICKS PER DAY, the margin of error is ±9.8%. I know that's a large margin of error, but that's also an extremely large amount of ad clicks per day.
If your willing to have that high of a margin of error, then a random sample of 100 would be very scalable. If were able to create more than 100 channels, then that margin of error would drop.
What's really going for us is that were really able to use random and representational sampling. Its not like our data will be swayed because were only sampling people with internet access, since all our ad clickers have internet!
I admit, 9.8% error of margin at a 95% confidence level is high, but I think its a lot better than no data at all.
I can get some very detailed information, including the exact revenue amount for that advertiser in that one scenario.
Okay, I'm lost (as always). How do you come up with that data point?
1,000,000 AD CLICKS PER DAY
LOL that would be some serious scratch! If I had a million clicks a day I won't really care about tracking... hehe
Don't forget you can continue to draw data every day and the trend data should decrease the margin of error, no?
That is an excellent point! I didn't think about that. The amount of data would grow, but so too the number of samples. The only reason that margin of error was so high was due to the number of samples. The more samples taken, the smaller margin of error, almost regardless of the total number of ads clicked.
I would imagine though you would only want to analyze the data for a certain time period, with the site not changing. Say one week at a time (700 data points). Otherwise, with other factors like Smart Pricing or a modified website, a data point from two years ago would definitely not be very representational of a ad click today.
Okay, I'm lost (as always). How do you come up with that data point?
Sorry I didn't address that important part. You would be able to log in to your AdSense account, and see how much that channel earned on that day. It should have only one click for that day (and if it has more than one, then you would want to disregard that number). The only value of real concern would be the earnings, and not the CTR, which ideally would be 100%, but may be lower depending on how many times that channel was used by people who didn't click on an ad for the day.
The next day, you could reuse all the channels, because in the AdSense reports you would be able to isolate the different data points by day. So you would be able to isolate today's earning from yesterday's, or last week's.
LOL that would be some serious scratch! If I had a million clicks a day I won't really care about tracking... hehe
That's a great number of ad clicks per day, something I wish I had too, but I'm far from it. I used the extremely high number for my calculations to show that it could be scalable.
The idea is a mathematics solution that interpreters data from the AdSense channels we're allowed, making for a solution well within the AdSense Terms of Service.
Depending on the number of new channels this would result in, I'm thinking that a large majority of people would probably be able to get accurate click data about their accounts with room to grow without the use of statistical sampling. For those with an extremely large number of daily ad clicks, this could potentially provide an extremely large sample size that with statistical random sampling would result in a 99% confident representation of your ad click data with a very small error of margin.
If your interested or good at Algebra, I'm definitely looking forward to your opinions and feedback.
