The New York Times seems to be stirring up the pot on the issue of accurate website statistics.

You can read the whole thing here...
http://www.nytimes.com/....FYjFK7n0HeOL0o4osG95w&oref=slogin

I do have to dispute one of their paragraphs...

But the Internet has given publishers a new form of ammunition: raw server data with precise numbers of site visits and page views. This data does not correlate directly to the number of visitors, but it does give them ballpark figures that they say are far more accurate than the extrapolations drawn by ratings companies based on panel samplings.

Raw server data is grossly unreliable for tracking number of visitors. This is because we've reached the day and age where "hotlinking" has become socially acceptable. This is the practice of websites linking to images and video on other sites, instead of hosting them on their own servers. This often happens on web forums where less-than-savvy users search Google Images for a cool "avatar" or "sigpic", to hotlink to.

I remember trying to analyze my traffic data on Junk Food Blog and Political Dogs using raw server data, and couldn't do it. The data showed there were more unique visitors than there were page views. Read that again. The visitors outnumbered the page views, because so many other websites hotlinked to my images.

When an image is hotlinked to, and someone visits the webpage that is doing the hotlinking, the server hosting that image counts a unique IP visit. So imagine if you have a very popular webpage, getting hundreds of views per day, and is hotlinking to an image hosted elsewhere? That server is recording is hundreds of IP visits, making that webmaster believe he's getting hundreds of extra visitors.

Another problem with raw server data is that they cannot differentiate visitors surfing the Internet from a proxy server. This where thousands of people are tied to the same IP address. An ISP like AOL, for example, has thousands of servers providing internet access to its users. One server could have tens of thousands of users logged on, and they all have the same IP address.

That's why browser-based stats trackers are much better. This involves a piece of javascript embedded into a webpage. Anytime someone visits that webpage, the stats tracking server associated with it counts an IP visit. And just in case that IP address is associated with a proxy server, the stats tracking server can differentiate these people with a unique cookie.

Examples of browser-based stats trackers include StatCounter and Google Analytics.

QuantCast is also a browser-based stats tracker that mixes in panel samplings, but for the benefit of marketers who might want to advertise on your website. The fact that it's browser-based makes it more accurate than Alexa, CommScore, Neilsen NetRatings, etc. Try them out. You can view my network statistics on QuantCast here...
http://www.quantcast.com/p-c5vtEkWO5_vKI

Note, I don't have all my sites on QuantCast yet!