From this, we can safely assume that most of you influence between a half million and one and a half million dollars in spending. On the surface, that's pretty amazing. But when you multiply the number of readers with the amount you all influence, the total dollar amount is nothing short of staggering. Even at the most conservative, if we assume you each influence about $500,000, multiply that by 70,000 readers, then you, as a readership control or influence something on the order of $35 billion! Instead, if we were to assume that you have an impact of $1.5 million each, which is also not unreasonable, you folks, in total, represent a purchasing budget of $105 billion. That's well into "Holy Cow" territory!

We've often thought of PalmPower, with over 2 million readers, as our flagship publication. But as we've seen above, PalmPower's Enterprise Edition has a very targeted group of enterprise readers with a powerful degree of influence. In fact, even though this is only our eighth issue of PalmPower's Enterprise Edition, as you can see in Figure A, we've already got more than seven times the readership of a typical business trade publication according to data from the Audit Bureau of Circulations and the Standard Rate and Data Service.

FIGURE A


PalmPower's Enterprise Edition beats out most business trade publications in terms of readership. Roll over picture for a larger image.

Behind the scenes gathering data
This project was massive. First, we had to sift through over 27 gigabytes of log files. To put this in perspective, imagine someone gave you a hundred-page report of raw data and asked you to find trends in that data. Now, imagine every man, woman, and child in the United States gave you such a report and you had to analyze all the reports together. That's the amount of data that had to be sifted.

Once we were able to extract PalmPower Enterprise Edition transactions, we had to get access to the only verifiable information Web sites are able to store: the raw IP (Internet Protocol) address of the person reading a page. So, from the mass of raw data, we extracted millions of raw IP addresses of the form "204.142.109.6." We then brought those addresses into a database, sorted them, and extracted almost 70,000 individual addresses. And that's where the real fun started.

Before we go on, it's important to understand that the 70,000 addresses we found represent only the minimum number of readers. That's because not every person using a Web browser gets a unique IP address. Many companies use what are called "proxy servers." As Figure B shows, the proxy server stores a local copy of our magazine, which means we can't see how many readers are really out there. The bottom line: the number 70,000 is, by necessity, a minimum count.

FIGURE B


Many readers access the magazine via a proxy server. Roll over picture for a larger image.

To find out where you work, we couldn't just go from an IP address to a domain name. We had to go from an IP address to a company. So, we custom wrote a program that, for every single IP address, queried a domain name server. It then did a Web page query of Network Solutions to access the Internet's Whois directory in order to discover what company had purchased a domain associated with the IP address in question. For each query, the program got back a complete Web page (with all the normal, human-readable HTML gobbledegook). The program then had to scan through the raw HTML page to extract the registering organization's name. Of course, many IP addresses couldn't be traced to a registrant organization, and the raw Web page query process had to be re-run many times, whenever the Internet hiccupped or Network Solutions' servers were too busy.