User experience measurement for free

The question of how fast a Web site is in the real world is tricky to answer because there are so many factors involved. A free service that worth checking that does just this is the Speed Survey (see links below).

In fact this service is not only free, there is also no registration, no contract and no commitment. The developer doesn’t even ask you to display the Speed Survey logo.

The service generates reports (see example) that break down the results by time of day, geographic location, content displayed and browser type.

The data collection method depends upon the client browsers being able to execute JavaScript (this applies to around 86% of the browsers in use – see the Web Applications newsletter regarding thecounter.com (https://www.nwfusion.com/newsletters/web/2003/0915web1.html).

Each page to be instrumented requires the inclusion of the following code immediately after the

tag and immediately before the tag. The first script sets up various global variables including your estimate of the number of hits that page will get (it is important to not over inflate this value as it determines the sampling frequency – the larger the number, the less samples) and the beginning load time of the page hit.

The last script determines the page load time and then requests a URL from the Speed Survey site with a tail that defines the various performance metrics gathered. This script also sets the data in a cookie to track the time between the loading of different pages from the same site.

The information gathered includes the time from click to page display start; the time from page display start to page display end without images; and the time from page display end to loading of the last image. From these values other metrics are calculated.

Most server page types can be instrumented this way including HTTP and FTP delivery, pages that use Server Side Includes, pages based on Perl/PHP/ASP/JSP, and password protected pages.

The value of this service lies in comparative data comparing month on month and site versions. You can see if site changes make a difference and if you divide your server farm to deliver two different versions simultaneously you could generate side-by-side comparisons over the same sample user population.

This is a terrific service particularly as it is free. I must admit I have no idea why the owner makes it available but when I hear from him, I will let you know.