In a recent thread over at the webcomic list, I asserted that PW stats tend to be higher (sometimes much higher) than other metric site as sitemeter or google analytics. This has been my experience. It was pointed out to me, though, that PW doesn't count people who have ad blockers and so the count could be lower.
What's your experience been? Are your PW stats inflated, deflated, or spot on?
Interestingly enough, I find them to be lower XD
They're slightly lower than my other stats sometimes, but mostly similar.
Google says I got 750~ unique visits yesterday...
PW says 2k.
So... huge diff. I dunno why.
The pageviews seem to be closer on tho. Google says
near 4k and pw says 3.7k~
I find PW's "uniques" to be skewed very, very high, but their overall page views are about normal with my other metrics.
Almost invariably, my stats on Project Wonderful are grossly higher, sometimes more than double what's on Google Analytics and the internal hit counter on Smack Jeeves, which are both about the same. It's quite shocking to see that some people's stats read lower, that seems to indicate some level of randomness. The ad click stats on PW, though, seem to more or less jibe with the referring traffic stats on GA, which is really what matters to me, and I suppose there is logic to the idea that I shouldn't complain about my stats seeming higher to advertisers.
Welllllllllll.... if people read my article from this week (one comment... really?) and understood it or perhaps asked questions until they did understand it you would all know how/why PW stats can be higher or lower than Google Analytics.
http://forum.webcomicscommunity.com/index.php/topic,312.0.html
That does explain why they're different, but not why they're inconsistently different, why one site's stats will be wildly higher and another's will be wildly lower. Unless computer caching operates under some law of physics to which I am completely unfamiliar (and I am open to that being the case), logic as I understand it would dictate that the way PW was able to count one site would be more or less similar...not the same, but similar...to the way it counted another. The same should be said for GA. What I mean is, if PW's stats are only about 60% accurate on one site, they should be about 60% accurate on another site, and GA being 50% on one site, it should 50% on the other, and so the rough differences between PW and GA should sit somewhere in that 10% range for any given site. The only other variable would be the configurations of the computer accessing the sites, the general statistical difference in the audiences between the two sites would need to be considerably different, which would only really occur if the comics had wildly different demographics. Webcomics appeal to a pretty thin spectrum, as we know, so that's probably not the case.
Like I said, I thought I understood before that PW just had wonky tracking, but now I'm wondering if it all isn't just a big conspiracy to make us buy toilet paper.
Quote from: Gibson on February 19, 2010, 06:02:31 PM
Like I said, I thought I understood before that PW just had wonky tracking, but now I'm wondering if it all isn't just a big conspiracy to make us buy toilet paper.
Well, I don't remember it being specified what metrics they were being compared to. If people were comparing PW to the different counters they used and coming up with different results, that would make sense.
Concerning PW, the numbers I get out of that are very weird. The number of visitors I get from my ads seems to make sense, but it cranks up my uniques crazy high. However, if it also cranks up *other* sites uniques crazy high... well, I get click-through rates of 1-6%, if the uniques from PW were 2-5 times higher than they really were(and what my other counters say) then the click through rate on my ads would be, like... 4-30%. No way.
In conclusion, I think the earlier article about voodoo is right.
In general, web stats are unusable for any sort accuracy. They're sort of like stock prices. Day to day they mean nothing, but trend-lines are very useful.
Aside from the trending indicators, the only thing I even use them for anymore is to gauge relative sizes of sizes, so long as you use the same tool for the numbers in the comparison. Even then, the strata are basically "small, medium, large, zomg hueg".