Forum Moderators: martinibuster
Mathematics How many clicks do you get each day? Statistics also apply to AdSense. The standard deviation for random events like clicks of ads can be calculated as the square root of the number of clicks. So, if you have an average of 100 clicks per day the standard deviation is 10 which gives the following fluctuations pure because of statistics:
32% of the days less than 90 or more than 110
5% of the days less than 80 or more than 120
1% of the days less than 70 or more than 130
Day of the week. Every site has its own day of the week characteristics. Some effects are:
Season Many activities like travel are seasonal. Advertizers will adjust their bids according to the season. Also search behaviour changes with season.
Ad targetting Google tries to find the best matching ads for your page. The only way do to this is to sometimes try new ads. It is just trial and error. There are people posting here to tell their ads are less targetted, but I never see posts that people thank Google for better targetting their ads and increasing their income, but this happens also :)
Football matches, elections etc. Large events trigger priorities of people. If you see a large decline on just one day, it might have been because of the SuperBowl. I'm from Europe and the first time I heard of this event was actually at WW (no flames please)
Change of competition for keywords New advertisers, advertisers switching the content network on or off in their AdWords account and promotion campaigns for a specific product or service can all highly influence the EPC.
Smart pricing The AdWords bid price is not always paid by the advertisers. Google applies a smart pricing algorithm which reduces the price if the estimated conversion rate is lower than expected. Changes in this algorithm might cause sudden changes in the EPC.
Click fraud Impressions are about the same, but CTR is much higher than expected from the reasons mentioned above? Someone might "help" you by clicking your ads. It is also possible that someone is clicking on a competitors ad to use the daily budget. If this happens to you it won't hurt to inform AdSense.
Don't know if I covered all possible reasons for fluctuations, but as you can see AdSense can hardly be called a constant source of income. I am happy with every dollar they send to me. Until now it is the only source of income which works while I sleep or when I am on vacation.
Changes in the URL filter The URL filter is a nice feature to block ads from competitors on your site, but some webmasters also use it to block ads with they think are not related to their site. Your visitors could have another opinion of it and you might disable your money maker without knowing it. Always watch statistics carefully after adding or deleting entries in the URL filter.
In my experience, many owners of small companies have from Monday to Friday no time to surf around.
They use the weekend to search for interesting offers for their companies.
Sorry but you can't apply a lot of the maths above to Adsense. Why? Because your data+response both change in real time. In other words, you don't get the same ads appearing, and the advertisers don't pay the same amount to appear. Also, you don't have the correct data from Google. You get a kind of average for a 24 hour period in US time.
Google is chaotic, you need to model it. This needs detailed data and we don't have it. Google does, we don't.
Besides standard deviation is not the square root of the number of the clicks. Plus SD is only an indicator of data scatter, not an absolute measure. Here is SD.
x = one value in your set of data
avg (x) = the mean (average) of all values x in your set of data
n = the number of values x in your set of data
For each value x, subtract the overall avg (x) from x, then multiply that result by itself (otherwise known as determining the square of that value). Sum up all those squared values. Then divide that result by (n-1). Then, there's one more step... find the square root of that last number. That's the standard deviation of your set of data. (from RobertNiles.com).
Thank you for taking the time to give some more information about standard deviation calculation. We were however talking about two different types of standard deviations.
The standard deviation you are mentioning is the standard deviation of the actual data after all external effects have been accounted for. You need a model of the process to calculate it. This is the standard deviation after all effects like bid prices, weather, etc. have been taken account in the result value.
The standard deviation I mentioned is more correctly called the "binomial standard deviation". It is the standard deviation when all parameters in the model are constant and there is only a choice between two values: clicking on an ad or not. So the standard deviation I mention is about the different CTR values you see when you have a fixed amount of impressions per day and no other variables that influence the outcome. It is statistical noise because of a limited amount of samples (impressions) per day.
Without making this thread a mathematical one which might not be interesting for the other readers I will give an example with a fixed amount of 10,000 impressions per day and an average CTR of 3%. Look at the CTR as tossing a coin and in 3% of the cases you see one face of the coin (a click) and in 97% of the cases you see the other side (no click).
The average amount of clicks per day: Avg = 10,000 * 0.03 = 300.
The binomial standard deviation is calculated as: S = sqrt( N*p(1-p) ) where N is the amount of samples (impressions) and p is the probability that we see a click (3%). Then:
S = sqrt( 10,000 * 0,03 * 0,97 ) = sqrt( 291 ) =~ sqrt( 300 ) = sqrt( Avg )
Because for most AdSense users the percentage of impressions that lead to a click is low, the factor (1-p) almost equals to 1, so we can say that the binomial standard deviation is close to the square root of the average number of clicks.
Interesting that even in a well defined science like mathematics you can talk about one parameter in different ways.
Luckily for us (and I've been told this directly by someone that knows) Google supposedly statistically samples the effectiveness of ads on Google's search results first before introducing those ads into the AdSense network. Only the best performing ads actually make it to your web site and some real stinkers supposedly get disabled as they are statistically insignificant to generate revenue.
So what happens if there are limits set by adsense algo to control effective CPM for sites
I think that's just a little paranoid thinking as google wants to make as much money as possible and seems willing to spread it around as much as possible to maximize their bottom line. I think the publishers web site itself (traffic, target, etc.) and the quality of (and price) of the currently running ads are the basic limiting CPM factors.
Based on my direct ads sales to customers prior to having AdSense the CTR and CPM are very similar to what I was charging for banners, links, etc. and what the actual performance of those ads. Therefore, based on the known pre-AdSense performance history of my web site I've concluded that what I see in Google tracking is very reasonable and less than likely to be manipulated.
But I've been known to be wrong on occassion and have one ex-wife to prove it.
Therefore, based on the known pre-AdSense performance history of my web site I've concluded that what I see in Google tracking is very reasonable and less than likely to be manipulated.
I'm inclined to agree. I don't know much about statistics, but--as a layman--I'm always impressed by how much consistency I see in search referrals, traffic for specific pages, affiliate bookings (when adjusted for season), etc. For example, I can say with almost complete certainty what 4 of my top 5 pages will be on any given day, and I can confidently predict (for example) that my article on [X] will get 200 to 250 page views on any given weekday. Similarly, I know within a fairly narrow range of euros what I'll earn from hotel affiliate bookings on today, tomorrow, or the next day.
Also, since AdSense "smart pricing" was introduced last April, I've probably seen a 25% climb in my average CPM. That's a fairly significant improvement, and it doesn't suggest a "flattening out" of CPM by Google's compensation algorithm--at least not from my admittedly limited perspective.
something that is hard to measure which is the human factor that a) creates the ad copy and b) clicks on the ad copy. Yes, statistically you can create trend analysis to map what you see but a few days of ads simply not matching the expectations of your visitors and the results of all those squiggly little equations are out the window
Exacly why I started this thread. There are so many factors that influence the earnings that there is no need to panic if one or more days are out of bounds. My goal was to create a FAQ about possible reasons for strange statistics rather than modelling the whole process in some magic formula. The mathematical error was only ment to be one cause of fluctuation but accidentaly it became one of the main points in this thread. If AdSense could be described with statistical formulas only it would be easy for everyone to become a millionair.
When I started with AdSense I saw huge fluctuations--mainly because of the algo searching for the right ads--and I started analyzing what could influence the earnings. After my analysis I realized that AdSense is a very nice and easy way to earn money but I don't want my whole life to rely on it. There are just to many factors you can't control. And I didn't even mention algorithm changes in the search engines which can stop free traffic overnight.
possible reasons for strange statistics
Soccer match!
