Thanks for the information on alternatives. I too have been searching for an alternative to Alexa now and then when I have a few spare moments. My own sites which are blogs barely make the radar of Alexa but that’s not my biggest concern. But I have seen how stats can be manipulated to increase one’s ranking there.

One example of such manipulation is the use of a blog cycle in sites such as Alphainventions.com and Condron.us to generate mechanical views. One might easily argue that generation of phony statistics is the reason d’etre of these sites. Alphainventions (and not Condron yet anyway) also appears to make considerable income by allowing spammers easy access to his site and to yours. Bloggers go to these sites to publish their own blogs or to view blogs which have recently been updated and possibly to connect with other bloggers.

While doing some research on how to increase views I eventually came across the ranking sites. I don’t know enough about the source or quality of statistics to understand some of the arguments about why Alexa oughtn’t be so highly regarded. I can detect a phony stat now and then when it crosses my path though; and see some of the means of manipulation and generation of stats.

I’ve been looking at Alexa stats on a site (not my own) for several weeks and have noticed a startling face about their stat graphs: the week and the month graphs do not agree. I don’t mean that the lines don’t look the same; that could be explained by the magnification. No, the fact is that if you compare the “trailing one month” stat given by the graph for any particular day to the same day’s stat on the “trailing one week” graph you will find as often as not that they are not the same. How can they not be the same? They ought to be exactly the same. Yet you will typically find one to be 3.8 and the other to be 2.6 or some such large inexplicable difference. Is it due to a sloppy graph drawer? Poor mathematicians in the Alexa graphics department? Whatever the cause of these large discrepancies they certainly do NOT suggest the work one would expect from the standard of the industry.

That’s all. — James

Regarding your suggestion (above) to use a “bit of javascript in a widget” to increase the accuracy of stats. How? Yes, you might get a more accurate count of the statistic for a single site but there are plenty of free statistic counters (Statcounter, Ritecounter) which do that albeit with some inaccuracy. But these ranking sites are supposed to do more than count your stats. They must compare your site to other sites before ranking can be done. Until a large majority of these sites have that javascript in place then any attempt to compare them or rank them might be ludicrous. What is used when they don’t have the counters in a widget? Is it estimates based on some profile of the site?

Could simple trending be of value – not sure. If it is then we have kicked some wicked butt here *grins* In fact, maybe I’ll start believing Alexa if just for the ego boost *winks*

You are correct – stats for one site can vary. I have the proof in a recent audit. It’s so fraught with difficulties I can see why you blogged about it!

Please contact me by email if you want more details.

I don’t agree with the usual criticism that these services are “wildly inaccurate”. Tracking traffic to 10 million different websites is a much more difficult problem than determining the top TV shows last week.

Even numbers that stay within an order of magnitude of the actual traffic can have some value – especially if the demographic data is close.

It’s an interesting process to watch.

First and foremost, my understanding from speaking to someone who works for Hitwise is that it is an enterprise solution. This means if you are an SME, it is not as useful.

One site I work on is on the ragged edge of this demographic. Recently they were told gains had been made against their much larger rival who hash better market penetration and better brand awareness. I have trouble with this since I know their stats

But as a comparison, Alexa sais the smaller site had double the traffic of the larger one. Ouch – that seems wildly inaccurate.

I manage about 20 domains with traffic ranging from 500 to 1,000,000 unique US visitors per month. I’m pleasantly surprised at how accurate Quantcast is when compared to the numbers reported by Google Analytics. Even the demographic info is good. We have a few sites that we know draw high numbers from certain groups (Female or Africans-American audiences for eaxample), and their demo info is about right.

I’ve looked at the fee-based services like Hitwise, but for now I can’t justify spending the money over the free info from Quantcast.