Forum Moderators: martinibuster
...and the AdSense algorithm automatically sends an invalid clicks email if a publishers stats are, say, outside 99% "norm" band ..
...then every 100 days on average every publisher will get an invalid clicks email...
Having thought about the problem, I cannot see a flaw in the logic. If the algo sends the email on outside 99.5% of norm then it's once every 200 days, or whatever. But if it is not set fine enough then blatently bogus clicks will get through.
So every publisher must get such an email sooner or later, it just depends how fine the hair trigger is.! One assumes some genius at Google has thought about this.
my stats are up and down each day but the average over a ten-day period is just about stable.
if you look at the ultimate "statistically stable" test, throwing a coin...we know that as time progresses, it will settle down at 50-50 for heads and tails. But we cannot assume that every hundred throws. If we throw the coin three hundred times, we would get something like:
1. heads 56 tails 44
2. heads 51 tails 49
3. heads 47 tails 53
nothing statistically significant in those figures but nowhere near a 99% accuracy that google is after, allegedly.
It all depends on how long a time span google looks at for their 99%. Back to the coins again, if we look only at batches of ten throws, either heads or tails will get seven or eight out of ten occassionally. If we look at a hundred thousand, it would be strange to see either of them with more than 52,000. If Google uses a week as its "standard", I see less problems, but I still think 95% would be a more realistic figure to aim for.
week as its "standard"
I would think it is dependent on the temporal (like time-of-day, day-of-week, etc) and spatial (like ip address, connection, cookie) variation across several weeks. And just a single instance of deviation would not mean anything except inherent transient noises in traffic patterns. Though i do think they have statistically reliable algos that would trigger invalid clicks email, termination of the account would involve human judgement.
I just completed a site overhaul moving from mostly banner ads to sky scrapers to the right of the content. At the same time Google reindexed my site and doubled its cached content. My CTR for Monday was almost double the previous Monday with about 10% extra impressions leading to a massive increase in clicks. I haven't the slightest bead of sweat forming on my brow as I am confident Google will know why my stats increased. And probably to a greater degree than I do!
Admitidly, I can not back up this confidence in Google with hard facts but the alternative just doesn't sit right. And some times you just have to go with your gut.
I do not believe that Google sends out these messages at random.
Like crxchaos, I'm confident that Google uses all necessary means to detect fraud.
If I ever get such a warning email, I'll remove Adsense ads from my site and that's it.
Life goes on with or without Adsense :-)
[webmasterworld.com...]
No, not worried, just curious about the statistical background of "invalid clicks" emails.
The old test used to be if you had an infinite number of monkeys and an infinite number of typewriters, then sooner or later one of them would write the works of William Shakespeare..
.. it seems to me, that the Google algo must trigger the "invalid" clicks" email whenever clicks are outside a certain parameter. That parameter has a finite (rather than infinite) boundary.
Therefore sooner or later every publisher will get such an email. I would be extremely surprised if any publisher does not get the email at least once for every 100,000 clicks on their sites (or whatever figure you want to put on it)
it could be that there are an infinite number of monkeys back at Google sending the emails out ;)
...then every 100 days on average every publisher will get an invalid clicks email...
Olias,
Thats my point. Going back to my first post
>>So every publisher must get such an email sooner or later, it just depends how fine the hair trigger is.! One assumes some genius at Google has thought about this.
I really don't know whether the Ph D's at Google have thought it through
