I often get requests to provide statistics for such and such page or file. Here is some text to copy and paste when necessary as a disclaimer.

Do:

  • Look at and compare trends in visitor numbers, across time periods, regions.
  • Check the impact of particular marketing activities, for example, sending out a newsletter.
  • Look at your top pages overall, and top ‘landing’ pages (the first page someone visits on your site) as they can give you an indication of which pages you should put most effort into.
  • Check which keywords people are using to get to your site, and whether there are obvious keywords missing. If there are, then you might need to invest some time in search engine optimisation activity (a post on search engine optimisation to come soon).

Don’t:

  • Read too much into a specific number or percentage, as these can easily be wrong. There are many reasons for this, outlined below.
  • Think that the USA is more important in terms of visitors than it is – this is especially the case if you are using server log software (see below) that might over-report search engines, the majority of which are based in the USA.

Why numbers can’t necessarily be trusted

Different tools work in different ways. Server logs sit on the web server and note down each time a file is accessed. You can then use some log analysis software to produce reports that look nice. However, statistics can be skewed by:

  • Internet providers keeping a local copy of a page and serving this up to the user rather than requesting a new copy from the server. The server therefore doesn’t know the page was ever there.
  • Search engine visits included in the logs. Search engines visit pages hundreds of times a month, and if your log analysis software doesn’t know about a certain search engine, it can look like your figures are much higher than you thought.

Google Analytics and some other stats websites work by receiving messages from a internet user when a page is accessed – this message is generated by a piece of code included on every page of the website. This can lead to the following anomalies:

  • If the client machine isn’t up-to-date or certain functionality is turned off (like Javascript, which some organisations turn off in the browser for security reasons), then the message might never be sent.
  • Search engine visits are less likely to be an issue, as most of the stats websites have an up-to-date list of search engines and excluded them from reported statistics.
  • As PDF files and the like are not webpages, and cannot therefore include calls to the analytics website, they are not counted. There are ways around this, by altering the code on links on a site to include a call when they are clicked, but this still doesn’t cover every download (for example, if someone clicks a link to a PDF from another site which doesn’t therefore include a call to analytics).