Forum Moderators: martinibuster
How can eCPM be slashed by a THIRD from one day to the next!? Today is looking like a normal day and I pretty sure yesterday wasn't smart pricing. If that were the case then each day this week should be the same as yesterdays shambles. Will keep you updated (if you're interested ).
My Saturday and Sunday eCPM's are always much higher than the rest of the week because my actual page impressions are considerably lower.
Ok, so I'm maybe having less traffic, that's correct up to a point, however I'm also having even fewer spiders and rogue bots on a weekend and this I find is making my eCPM fluctuate more.
Does this mean that these bots etc only work a 5 day week?
Unfortunately this doesn't resolve your Sunday scenario however it probably explains many of the wild fluctuations when the overall confirmed click levels are reasonably constant.
It's the middle of the summer and consumers are hording their cash for their activities and advertisers know it.
Lost the regular ads on my business page and half my 24 themed pages went generic for my niche/homepage for the weekend.
CPC went from 33 cents to 22 cennts, just lost the juicy ads for the weekend. I think it has taken this long for most to get around to creating duplicate timed/bid tiered campaigns.
[edited by: Khensu at 12:56 pm (utc) on July 24, 2006]
I think you are starting to see the effects of daytiming.
Would this not only affect sites that have an end product to sell with their click thrus?
I don't know however I would have thought that more likely.
My realtively expensive construction products have seen no difference in advertisers since most buyers, whether trade or retail, are seeking information about availability, the location of suppliers and possibly pricing with only a very small minority positively wanting to purchase on-line.
I know my products are different however each sector adds to the whole therefore we each have to keep reminding ourselves what's happening with our own sites may be completely the opposite for others.
Saturday, on the other hand, was weaker than average (and weaker than other Saturdays in July) across the board.
Fortunately, long-term averages are what count, and I've found that bad days are usually offset by good days (as they were this weekend).
I know that doesn't explain everything in all circumstances but it does contribute in various degrees.
That is what I would be doing if I was still selling my product and not giving it away as downloads with the ads on the page.
Europe,
Your pushing a good niche for the summer so you should be not as vulnerable.
[edited by: Khensu at 2:08 pm (utc) on July 24, 2006]
How can a bot, that doesn't interprete JavaScript have an influence on adsense?
I didn't say about AdSense Javascript:
my actual page impressions are considerably lower.
Adsense seems to recognise some bots etc and not show them in the Page Impressions however there are many more that it allows actual page impressions to be counted.
In fact it ties in well with my logs.
How can eCPM be slashed by a THIRD from one day to the next!?
Open that CSV file in Excel (or your favorite open source alternative) pick a blank cell and use it to calculate the standard deviation of your daily totals. You calculate the standard deviation just like you were doing a total, except instead of SUM, use STDEV
As you select the list of daily totals note that the first line is a heading and the last line is the grand total so don't include them in the selection. If you have been doing Adsense for 200 days, for example, it might look like this: =STDEV(F2:F202)
In another blank cell, calculate the average for the same range, like this =AVERAGE(F2:F202)
Now comes the useful part. If everything stays the same from now on - your web pages, Google's algorithms, ad prices, etc - you can say you are having "normal results" any time your daily earnings are within the range from the "highest expected", equal to the average plus 3 times the standard deviation to the "lowest expected", which is equal to the average minus 3 times the standard deviation. (In my case the low point is less than 0 so it could be 'normal' for me to have a zero income day)
Any time your results fall outside of that range you could start looking for a special cause - added pages, algo change, ad market soars/drops etc. but if it is within that range it is just business as usual.
Now, on a forum of this size there are bound to be people who did much better in practical statistics than I did so please chime in and correct me where I am wrong or where this sort of analysis could be improved.