webcentric - 3:07 pm on Aug 17, 2013 (gmt 0)
There's probably a calculation you need to make in regards to your 80% reduction figure before settling on that figure (maybe you've already made it but I'll point it out just in case).
As netmeg mentioned, G is taking clicks away much closer to real time now then they were a year ago. This means there's less of an adjustment at the end of the month and more taking place during the month. My question would be, what kind of a percentage of adjustment were you experiencing last year when final earnings came out versus say, last month's end of month adjustment. The difference between these two figures should be subtracted from you're 80% figure (assuming you're not already doing so). It may be only 1% or it could be 20%. In other words, if you were loosing 5% on average a year ago at the end of each month, you should consider that normal and subtract it from the 80% you are now experiencing because it's probably still happening, just at a different time. At least in this way, you're eliminating one known constant from the equation and can focus more closely on the true variance.
Of course, if you're beef is entirely about the difference between statcounter stats and G's stats, then this is less applicable. My point is about what G says, now vs then.