Some examples of the lack of accuracy of Alexa

Hi,

After having spent some months checking the Alexa rankings of several sites, last week i began to worry about our traffic, because all of a sudden Alexa dropped us to the 100,000 mark. We had been consistently scoring about 30-40,000 (yes, without faking it).

After looking at our logs, we did not see any change, except some peaks in traffic, but the average was in line what with we expected, maybe a 10% higher.

Well, then i started to check other sites rankings.. when finally i decided to check in one personal site i have with my girlfriend. It's just a place where we upload pics and keep a very small blog. That site is totally private, no links in or out, and just for her and me. Following is the alexa rankings, together with the real stats of the sites for a given day.

Private site: 24 Mar 2004 Visits: 2 Pages: 7 Hits: 24 BW: 27 KB Alexa ranking: 449,000
Commercial site: 24 Mar 2004 Visits: 2780 Pages: 62864 Hits: 331559 BW: 2.02 GB Alexa Ranking: 204,000

Data is taken from AWStats package installed by defaul in CPanel. Of course i run later an offline detailed analysis of the logs with other programs, but that is enough to show my point about Alexa.

Regards, Eugyx

Well I look once in a while but don't rely on this. If a low user amount has the toolbar, then your rank goes down. If a day lots of them has it, then you go up. It's not really reliable.
Your logs show the hits and visits... more reliable... try to do better that your previous month.

I concur eugyx

My example is a common one, you'll see the metrics with many sites not within the typical Alexa demographic (ie UK/US).

I was wondering why one of my sites, that I knew to be taking 3K uniques per day more than another, wasn't actually ranking any better - perhaps by a minor margin, but still, nothing to brag about.

I suddenly realized that because the high traffic site was catering primarily to an Australian audience, the results were totally inaccurate.

I conducted a campaign to pull themed linkage from the US & UK, as this is consistent with future plans for the site. Lo and behold, I noticed a spike creeping up (Alexa graph) to around the 10K rank. This lasted for a couple of days before metrics returned to normal. I wondered what the big deal was, after all, uniques remained marginal for the site, in fact, over this 2-3 day period, they had dropped to the usual weekly low.

Upon closer inspection, I found that the spike was due to UK/US webmasters visiting the site to confirm reciprocation etc.

The crazy thing was, the traffic that drove Alexa was probably around 7% of the daily total for those days.

Alexa metrics are therefore only representative of specific demographics - notably those where uptake of the Alexa toolbar is common place.. hmm I wonder if Alexa discloses this info (I haven't looked) ?

I had wondered why so many sites showed monthly spikes then disappeared off the chart. I've spoken to a few webmasters, and they agree that their activity around those times did attract additional traffic from offshore.

My question - where can I find reliable Australia-wide metrics like those provided by Alexa ?

I haven't read the Alexa disclaimers recently, but when I did some months ago they did mention these shortcomings and specifically mentioned Korean traffic as skewing results.

My own site is 87,352 at the moment and I have been watching it carefully over the past several months. It was in the 59,000 range just a few months ago. I do not have the Alexa toolbar installed now because it seems to have a sludging effect on my (2.4 ghz, 512 ram) computer. I do have their stats script/graphic installed on my "about" page.

But for a time I did have the toolbar installed and casual visits to my own site did seem to have an affect on the rank number, which does not speak well for Alexa accuracy since I am in the five figure range... I never intentionally added visits, just my usual checking a few times a day.

Andi

I think I'm around 1,500 Alexa ranking, but ultimately it's not really useful unless you are under 10,000. And even then, it's really only useful for comparing to yourself historically.

Either way, logs are a infinitely better measurement.

- Shawn