Valleywag's Nicholas Carlson recently took a shot at Mike Arrington's claim that TechCrunch received about 2.7 million unique visitors per month. Carlson's "evidence": a chart from Compete.com. A quick visit shows their estimate: 870,992 people (shown above). Wow! Off by a factor of 3.

Was Arrington exaggerating? Not so fast. Compete gathers data from multiple sources, but it's not clear that their sample can accurately measure a site such as TechCrunch. Did Carlson bother to think about the chart he posted? If so, does he believe that TechCrunch "uniques" grew more than 8x in 2007? That's unlikely. I'm pretty sure the blog's (impressive) growth came earlier.

To get a quick "sanity check" on the data, I used Compete for the above chart: TechCrunch (in blue) against Slashdot (in red). There's no way the Slashdot data is correct: relatively flat from Jan to June (which I snipped), then a leap from less than 300K visitors in June to over 700K in July. More likely: Compete's methodology and/or data sources changed, doing a better job of capturing the Slashdot (and TechCrunch) visitors who were there all along.

The first commenter ("Truc") on the Valleywag post points out another factor that Carlson should have mentioned:

Compete only measures/estimates U.S. traffic.

Perhaps half of TechCrunch's visitors are from outside the US, so Compete's 871K for the US translates into an estimated 1.74M uniques worldwide. Granted, we need another factor of 1.55 to get to Arrington's figure -- but that's well within the range of accuracy for Compete or any other metric that I've looked at.

To settle the issue, Arrington posted a screenshot from Google Analytics showing 2,647,027 unique visitors. Of course that number isn't perfect either: some people delete cookies, others read from more than one computer, and some "visitors" are undetected bots. Still, until I see compelling evidence to the contrary, I think that's a much better metric than third-party sampling.