compete.com stats
I’ve been reading quite a few articles about the recent economic crisis and kept seeing compete.com cited as the source for some of the graphs. The tag line in compete.com’s website is all you need to read to know what kind of a service they provide, “Track your rivals. Then eat their lunch.”. Types of statistics they provide is utterly fascinating.
I was curious to find out how they do it. Luckily it’s all documented. Quoting from how they estimate ranking,
Based on the daily web usage of more than two million members (and growing!) of the Compete community, Compete calculates and estimates total traffic and rank for nearly every site on the web. We use rigorous statistics to make sure our estimates balance demographic and connection factors that match the entire U.S. Internet population. Currently, we’re calculating the number of people in the U.S. that visit any given Web site each month (international usage calculations are in development).
Ok, so the statistic include only people in US. People in US who have access to the Internet. As it turns out this is around 72% of the total population. Or, about 220 million people. Out of this, as compete.com says their sample is 2 million people. A mere 0.9%. So the statistics will be generalized from this sample. Also, I’m assuming that these 2 million people are all those who have installed the compete toolbar.
Right. Now, say, you’re a social bookmarking service trying to eat all your competitors’ lunch. How accurate the information will be? Given that social bookmarking is a very specific niche, can you conclude that their sample is in fact a representative sample of all people that use social bookmarking services and/or who have a keen interest on it? If you’re operating in another niche market can you rely on these numbers?
I think the answer is up to you. In a world where 3 different website statistic gathering services give 3 different sets of numbers (Apache logs, Awstat and Google Analytics) for one site that you have total control over, it’s considerably more difficult to accurately predict numbers for a site that’s beyond your reach. Eat your competitors’ lunch you say?