Measuring Success October 13, 2009
Measuring Success:Business Website
The means of measuring the success of a business Web site are decided using the ‘Business Analysis’ section of the development life-cycle of a business Web site. More traditional metrics include:
- An increase in market share
- An increase in sales
- The proportion of sales coming from the e-commerce section of the Web site
- An increase in brand awareness
- An decrease in calls to a helpline (if, for example, a business has put its support information online)
The above factors directly impact on the success of the business.
In addition to the above many businesses also examine the access logs on their Web server as a way of determining how ‘successful’ their Web site is. Unfortunately statistics from server logs are essentially meaningless and, in reality, can tell you nothing more than how busy the Web server is (also called server load). Some reasons why Web statistics can under and over-represent the popularity of your Web site are:
- Caching – A Web cache is a store of the HTML pages, CSS files, images, etc. for the Web pages you have viewed. Subsequent visits to the same page (within a certain time) will result in files being read from the cache rather than downloaded again from the remote Web server. The pages will be quicker to load but the Web browser will not request pages from your Web server everytime they visit your Web site. Consequently these accesses won’t appear in the server logs or statistics. Caching can occur at the Web browser or at the ISP level – users can control caching at a Web browser level (by changing the cache settings) but not at the ISP level.
- Users – The only way you can accurately determine the number of users your Web site has is to make them register and then log in each time they want to use it (assuming they don’t share their log in credentials with colleagues/friends). Programs designed to analyse sever logs often estimate the number of users based on the number of unique IP addresses appearing in the logs (1 unique IP address = 1 user). Unfortunately many Web users browse the Web from behind a proxy server (a server, usually run by the ISP, that sits between the user’s PC and the Internet) which makes requests for Web pages for them. Many people can sit behind the same proxy server so lots of users will appear to have the same IP address – programs analysing the server logs will count this as 1 user when in fact it is many. In addition, AOL (America Online), for example, use multiple proxy servers. When people using AOL as an ISP make a request for a Web page the requests for the HTML, CSS, images, etc. can come from any of the proxy servers. Therefore, Web server log analysis programs can interpret an AOL user as many different users. An additional problem is that when a user connects to some ISPs the ISP assigns them a dynamic IP address. Each time the user connects they may get a completely different IP address so again you cannot asume that unique IP address equals 1 user.
- Visits – Most server log analysis packages class a ‘visit’ as activity from a single IP address that is followed by a period of 30 minutes inactivity. As we have already discovered, it cannot be assumed that 1 IP address equals 1 user. Consequently this ‘definition’ of a ‘visit’ is completely flawed. In addition 30 minutes inactivity is a completely arbitrary length of time, a user may be answering some email, making a phonecall, etc. and so 30 minutes inactivity may not represent the end of a ‘visit’.
- Other data – Some server log analysis packages will also provide additional information on how they ‘think’ the site is being accessed. This information can include paths through the site (which can’t be determined due to users backtracking and using cached pages), time spent reading a page (a user may have been doing something else during that time) and a user’s entry point to a site (if the homepage is already cached then then entry point will appear deeper than it actually is).
A brief note about ‘hits’ – hits are the number of requests for individual files a Web server receives. Obviously caching will have a large effect on this value but it is also an often misused figure. Assuming there’s no caching, a single Web page containing 25 images and using 1 stylesheet will produce 27 ‘hits’ (1 for the HTML document, 1 for each image and 1 for the CSS). Therefore graphics intensive Web sites will have high numbers of ‘hits’ but may be considerably less well used than a Web site using far fewer graphics (which consequently has a lower number of ‘hits’). A much more useful figure is that of page accesses (or page impressions, page views), this figure represents the number of requests received for HTML pages. The page access figure is much lower than the number of hits and so is used less often since, to people who don’t understand how the figures are derived, it’s a less impressive number. However, all these figures are easily skewed by caching as we saw above.
Other ways of assessing the popularity/success of a Web site include:
- Site ranking within Web search results – “In a search for ‘business education’ on http://www.google.com our site is first in the list of results.”
- Link Popularity – comparing the number of other Web sites that link to you compared with your competitors can be a good indication of how popular your site is. This can be checked using an online