The purpose of this study is to design a methodology that estimates the quality of websites. It is scientific in that it calculates ranks from objective metrics.
PageRank has a patent by Google. It was supposed to represent the likelihood of a websurfer to stumble on a given web page.
In practice it is now correlated with the level of trust Google puts in the website domain. It is biased to older websites and big corporations.
Google also manipulates the public PageRanks for their corporate agenda. That makes at least two reasons why we should not limit ourselves to this metric. You can find details at a rant on Google PageRank.
In a certain way the webmasters ‘vote’ for pages by linking to them. So the links that point to a website can count as votes.
Update: I stopped using Yahoo! Site Explorer in April 2008. I found that this measure is unreliable.
For assessing a website quality we take into account the number of links that point to this website. The data comes from Yahoo! Site Explorer like so.
In practice a website shows incoming links of mainly one of three forms:
- the links demonstrate the website authority;
- the links show the website popularity among webmasters;
- the links are created by the same webmaster.
Therefore this metric is essentially useful to counter PageRank.
An interesting measure would be the number of people who read the blog feed. This metric is restricted to blogs. Furthermore the available statistics are severely limited, unreliable and biased.
This amounts to several reasons why feed statistics are not used to build the Cellarer.com directory of the best wine websites or the Cellarer.com directory of the top 100 food blogs.
This is the democratic metric. It assesses the number of people who are likely to visit the website next month. Update (on Andre’s suggestion in comment): these numbers are not taken from the webservers and therefore do not show the actual traffic.
This measure derives from the USA traffic levels measured by Compete.com. Update in April 2008: the metric is not the Compete value. The Compete figures over the past year are averaged, compensated and trended so that an aggregate is built.
The Compete figure is not available for sites that do not use their own domain name. Such sites are thus put at a disadvantage on this criterion. On the contrary the same sites are favoured by the two other measures since they receive links from well known platforms — which (usually) have nothing do to with the niche (food or wine). (Examples: the New York Times, Blogspot).
The valuation is a sum (in USD) that the website could sell for. This is hypothetical: it is a value that would serve as a reference if the owner wanted to sell and if she negotiated with a buyer.
There are a lot of things to consider for assessing how much a website is worth. Most of them are not public information: the revenue, the number of subscribers and registered users.
Update: I terminated this test in April 2008. I found that this aggregate is unreliable.
Many websurfers read top 100 lists to learn of a website they would read. They will probably not use most pages of the website. The popularity metric is an attempt to show how useful or entertaining each page in the website is. It is calculated by taking a measure of user interest (incoming links and traffic) and substracting a measure of the publisher size (PageRank). So popularity shows public support as opposed to institutional strength.
This methodology of PR + links + traffic is used to assess websites related to wine or food. Here is the directory of the best wine websites.
You can follow the directory evolution by subscribing to the feed of website articles or by subscribing to the mailing list of comments available below.
If you have a different view on website metrics, please share it by posting a comment.