Googles Page Rank is an algorithm that attempts to inform you where people are coming from when they land on your website and which sources are the most frequent. Note that, as its name suggests, a page rank is an index estimated page by page; its not an overall website assessment.

Nevertheless, say youre trying to determine the rank of your home page. You figure thats a good start, and customers can navigate more precisely once theyre in your domain. The ranking model begins by checking all the incoming links to the home page over some given duration (say, the last 24 hours or the last week, depending on the site traffic and how current the information must be). Customers can land on the home page starting from many links, and the links generating traffic to you differ in their importance. In particular, the influence of the incoming pages varies, as weighted by two factors:

a. The page rank of the source link (higher is better)

b. How many outbound links that source page contains (A lower number is better, in that the link to your home page is therefore more selective.)

Thus, if page A contains a link to your home page, and it has a high page rank of its own and relatively few outreaching links, it carries more weight than page B with its lower page rank and more outreach links. This algorithm is obviously iterative because we need to es- timate the ranks of pages A and B before we can bring them into the estimation of the rank for your home page. In theory, the iterations could continue ad infinitium. In the actual algorithm, there are starting values, and about 100 iterations bring most estimates to the convergent approximation. Finally, Google then exercises the universal modeling prerogative of including a term for wiggle room or a fudge factor.

1. How would you describe this algorithm in network terms? Would you use the same network principles if you were to design a competing algorithm?

2. Critics say this method doesnt account for the fact that many websites are not managed well; they might not be updated, links might not work, etc. How would you improve on this algorithm to address these concerns?