The number 1 thing you need to remember is that the data is not useful as an absolute measure - it's only good to compare sites.
The number 2 thing to know is that if you compare similar sites, the comparisons are more accurate. So if you compare trailbehind.com, everytrail.com, and backpacker.com, the comparison will be a lot more accurate than if you compare cnn.com, stackoverflow.com, and myspace.com.
Number 3 is these sites are more useful if you have benchmark data. For example, if you want to know the traffic for a magazine type site, go download their media kit (usually available as a PDF on their site), and it will tell you their traffic. Then you can use this data for understanding Alexa better.
I'm honestly not trying to attack you but there are several posts below (mine included) that indicate what you are saying isn't the case. If you have a factual basis that says we're wrong I'd be interested in hearing it. But in this post you are simply stating absolutes without backing those absolutes up.
I guess what I'm asking is "how do you know what you are saying is true?"
Download the media kits, see how much traffic these similar sites get, and then see what the error is on Alexa. I think you'll find that the numbers are of the right order of magnitude.
You're going to get a pretty clear signal when looking at sites relative to one another, since most of the error will effect the sites the same. And if you have sites that have the same sorts of traffic patterns, then the relative signal should be even clearer.
It's true that people gaming the system effect the numbers, but that's not going to distort the numbers enough to make them worthless.
You can't use Alexa or similar to see whether one site has 3000 visitors a month and another has 3500. But if one ha 3000 and one has 30000 then you'll see it in the graph.
But a couple months back I got linked to by Mathew Ingram and, while Michael Arrington was covering Mr. Ingram's post on Techcrunch he also shot me a link. As you can imagine that generated a huge bump. But Compete actually had my numbers trending down for that month (cut in half in fact) and the month after as well (which was completely contrary to my Analytics readings).
So now I don't even have much faith in the site for trending data.
No it doesn't. Their graphs for Justin.TV traffic aren't even usually the same shape as the real thing.
(Which is surprising. Grader.com is nowhere near as useful).
Please tell us Joel, what do you think really mattered in order of preference or apprx credit (say)?
- a much cleaner, better product than anything that exists out there ?
- market need.
- Seed traffic (hence contributed content, satisfying experience and stronger word of mouth)
- High profile startup in relevant space.
- word of mouth and viral effect because of satisfying experience
You are right, though; the reason why I started listening to the podcast was because I enjoyed reading Joel on Software and Coding Horror, and I've continued to stay because the show is much more casual than any other programming podcast I've come across (other shows I've found tend to take themselves a little too seriously).
Anyways, that's just my opinion, but I do think that a podcast is a great way to build buzz for a site, and if you know of any shows similar to Stack Overflow I'd love to hear about them.