Compete.com an Alexa competitor

[KLB]

I was over at http://whois.domaintools.com running some whois queries on IP addresses and noticed a banner ad for http://www.compete.com/, which was billing itself as a competitor for Alexa. Out of curiosity, I went to the site and queried my own site and was extremely impressed by the accuracy by which it reported my U.S. traffic. Its page views are in line with what I know them to be for my U.S. traffic as are the number of visits and number of page views per month. In a way, it was pretty freaky.

It appears that they may be buying click stream data from ISPs, which would explain why their data is so accurate. For more on how your ISP is involved see:
http://www.techdirt.com/articles/20070313/213014.shtml
http://internet.seekingalpha.com/article/29449

On one hand I'm sure a lot of advertisers will like the ability to get really accurate U.S. traffic data about websites, but on the other hand as a U.S. Internet subscriber, it really pisses me off to know that my ISP is selling my web surfing habits to others.

[Chris]

It is nice and accurate, and there isn't a webmaster skew.

http://snapshot.compete.com/online-l...italpoint.com+

[rpanella]

Theres also another site called quantcast.com. Both of these places show similar levels of traffic for my site but both are about half what Google Analytics reports for absolute unique US visitors over a month.
________
Og kush marijuana

[MaxS]

I'm a bit wary to believe those rankings -- I compared Sitepoint/DP with a site of mine that receives around 15,000 uniques a day; my site ranks higher.

Do you really think SP/DP receive less than 15,000 uniques a day?

[ses5909]

That's a really interesting site!

[Chris]

I'm a bit wary to believe those rankings -- I compared Sitepoint/DP with a site of mine that receives around 15,000 uniques a day; my site ranks higher.

Do you really think SP/DP receive less than 15,000 uniques a day?

They aren't as active as you may think. They serve a specific niche and have a high number of repeat visitors but neither is a general audience site.

[MaxS]

They aren't as active as you may think. They serve a specific niche and have a high number of repeat visitors but neither is a general audience site.

Interesting. In that case, I'm happy .

I'll have to start using Compete instead of Alexa.

[KLB]

Also people need to keep in mind that Compete.com only looks at U.S. traffic. Thus it will look lower than Google Analytic's overall stats. It should, however, look similar to GA's U.S. traffic numbers.

At least in my case its numbers for U.S. traffic seem to be very close to my real numbers. The reason for this higher accuracy is the way it gets its data, instead of relying on a toolbar install like Alexa, they buy click stream data from U.S. ISPs. This means they get data on all pages millions of users across all walks of life, not just data from those who have opted to install a toolbar. From a statistics standpoint, this is a much more reliable way to conduct a survey.

[rpanella]

Compete also has a toolbar along with their ISP data, and ISP data can be just as skewed as toolbar users. Here's a study that was done to determine the co-relation between true traffic and the various free external metrics: http://www.seomoz.org/article/search-blog-stats
________
Volcano Review

[KLB]

The only real skew with ISP data is if it doesn't cover a broad enough cross section of ISPs including dialup, broadband and wireless. Remember this isn't a function of some tracking software on the computer, rather a matter of page requests that were captured and logged as the requests hopped across routers owned by ISPs.

Quite literally it is akin to the phone company logging every phone call you make. Scientifically speaking this methodology would be very sound, but from a social aspect there are some tremendous privacy concerns about ISPs selling the Internet equivalent of phone logs.

[Chris]

Compete also has a toolbar along with their ISP data, and ISP data can be just as skewed as toolbar users. Here's a study that was done to determine the co-relation between true traffic and the various free external metrics: http://www.seomoz.org/article/search-blog-stats

What bull**** that was. I see them yet again touting their arbitrary "PageStength" thing. Its a neat tool, but they assign way too much importance to something they made up.

[rpanella]

To remove the skew though they would need a very large sample from lots of different ISPs. They currently claim to track 2 millions users, many which are through the toolbar, and even this is less that 1% of the users. and they likely buy it from very few ISPs.

More akin to tracking all calls made from MetroPCS phones in New York.
________
HERBAL SHOP

[Chris]

The only real skew with ISP data is if it doesn't cover a broad enough cross section of ISPs including dialup, broadband and wireless. Remember this isn't a function of some tracking software on the computer, rather a matter of page requests that were captured and logged as the requests hopped across routers owned by ISPs.

Quite literally it is akin to the phone company logging every phone call you make. Scientifically speaking this methodology would be very sound, but from a social aspect there are some tremendous privacy concerns about ISPs selling the Internet equivalent of phone logs.

The issue really is about sample size, you get a huge sample size with ISP data, this makes it statistically accurate.

But popularity matters for the sites being measured too. They took a bunch of niche blogs, none of which is popular in the grand scheme of things. They see it as comparing a 100, to a 75, to a 10, to a 4. In truth it was like comparing a 0.75 to a 0.10 to a 0.04.

This is just another example of people with uncertain educational backgrounds trying to do experiments and drawing improper conclusions.

No kidding they couldn't see any accuracy, they have a sample size of probably hundreds of millions and they're comparing sites that, top to bottom, only span a few hundred thousand uniques. When graphed as a percentage of total Internet traffic, or something like Compete's sample size, the traffic of those sites would form an almost perfectly horizontal sliver of a line right at the bottom.

A real comparison would need to make sure all participating sites used the same tracking system (say analytics) and then you'd want sites that get 10 million + uniques a month, 1-10 million uniques a month, and under 1 million uniques a month.

Heh... maybe I should do one.

[rpanella]

What bull**** that was. I see them yet again touting their arbitrary "PageStength" thing. Its a neat tool, but they assign way too much importance to something they made up.

Yeah, well basically the point of it was to show that sites with such little traffic can't be that accurately represented with Alexa etc, due to the small sample size they have. Obviously the larger the traffic a site gets, the more accurate these services with small sample sizes will be, no ones arguing that. My point was just that at KLB's traffic level, the fact that Compete is very accurate is more of a fluke than the norm.

Heh... maybe I should do one.

That would be an interesting study to see how accurate all the different metrics are at different traffic levels.
________
Homemade vaporizers

[KLB]

The issue really is about sample size, you get a huge sample size with ISP data, this makes it statistically accurate.

This is what I was thinking.

But popularity matters for the sites being measured too. They took a bunch of niche blogs, none of which is popular in the grand scheme of things. They see it as comparing a 100, to a 75, to a 10, to a 4. In truth it was like comparing a 0.75 to a 0.10 to a 0.04.

Okay I'm really confused.

What is this comment referring to? Is it referring to the seomoz article or to compete.com???

This is just another example of people with uncertain educational backgrounds trying to do experiments and drawing improper conclusions.

This is what I thought of the seomoz article.

When graphed as a percentage of total Internet traffic, or something like Compete's sample size, the traffic of those sites would form an almost perfectly horizontal sliver of a line right at the bottom.

Yep. Although I hate the "big brother" aspects of how compete gets their data, what I like about it is that it eliminates the "webmaster" skew that Alexa has. This could actually help dispel the popularity myth that some web publisher related sites have. For instance, according to Compete.com, my chemistry site is more popular than Sitepoint.

Quite simply, because webmasters seem to be the primary ones who install the Alexa toolbar Sitepoint's popularity is being heavily overstated by Alexa. At the same time because the groups that tend to visit my site are prohibited from using the Alexa toolbar (corporate and academic IT departments) my site's popularity was being under stated.

What I've observed about Compete.com's data for my site is that at least at my traffic levels its methodology is producing some very accurate results. Compete.com could really shake up the web advertising marketplace.