Summary
We performed detailed measurements of the top page download times, page sizes, and rates for the FT-150 WebSites. The data show very wide variations in these quantities that suggest that some of the FT-150 WebSites need improvement of adjustment.

Experiment Details
To collect the data we ran an eValid playback script that visited each of the 150+ WebSites. We put a 1.0 sec delay between each page download request to assure accuracy of the timings. The playback was done with the all cookies and entire contents of the cache deleted before the start of each download. We adjusted the script to assure that any URL redirections were not included in the timings.

The measurements were done on a dedicated high-speed DSL with ~1.6 Mbps download capability. The fastest-downloading page in the set used only ~40% of the available capacity. The machine used for the playback was fast enough so that the page rendering time was much less than 1% of the total time (this time is included in the measurements). The USA-based measurements were done 10 times and averaged. In addition, all of the measurements were done right after one another, so the total web latency factor would not have affected the data.

Results
The Complete FT-150 Top Page Timing Spreadsheet includes data from all four locations and for all 150+ top pages of the FT-150 companies. Below is a sample of the table that shows a portion of the data. In the fragment below and in the Complete FT-150 Top Page Timing Spreadsheet:

• Time. The time shown is the total page download and rendering time in milliseconds. The USA data is the average of ten successive measurements.
• Size. The page size shown is the total byte count of the page measured after the page download was completed and the page was fully rendered. This size includes all images and scripts and included files.
• Rate. The download rate is the speed at which that particular page was downloaded in KBytes/second.

Observations
Here are some observations about this data that we think are interesting.

• Download Time. The total download times span a 50:1 range -- from as little as 1.1 sec to over 66 sec. The average speed, about 5.9 sec, is nearly twice as long as the well-known "3-second click-away" threshold for web surfers.

  The observed range of full-page download times clearly is NOT a function of our connection to the web or due solely to congestion in the web. If it were then this would show up in variations over time [as web congestion ebbs up and down], but the data don't suggest this. Instead, these relative speed measurements are generally independent of the time of day and as a result the only variable is the candidate WebSite's ability to deliver data.

  Suggestion: If you site has a slow download time it is probably due to a combination of a slow server complex, an undersized server for the load you're experiencing, or a too-small interface (too little bandwidth) between the web and your server complex.

• Page Size. The page sizes also showed a very wide range: 32:1. Some top pages were as small as 17 KB and some were as large as 575 KB. The average across the 150 sites was about 121 KB.

  In some cases (as seen in the spreadsheet) the actual size of the top page downloaded varied quite a bit from the beginning of the test period to the end. This is normal because for certain WebSites the content varies over time and the size of the page would be expected to vary.

  Suggestion: Obviously, large top page size prevents quick download even if the connection is quite fast. But to assure the best reaction it is a good idea to keep the total download size of the top page to less than 100 KB.

• Download Speed. The aggregate download speed for these top pages varied by over a 16:1 range, from a high of about 80 KB/sec to a low of only 5 KB/sec.

  The average speed, ~29 KB/sec means that the average page size of 121 KB would download in about 3 seconds -- well within the "3-second click-away" threshold. Yet, in the data in our runs the actual average download speed achieved yielded only a 5.9 sec download total time. Apparently many fast-served WebSites also have very large page to page sizes.

  Because the page downloads never exceeded 40% of the available download channel capacity this variation is due entirely to web latency plus server speed issues. If you study the data you'll see that the relative rates keep the same pattern even when measured from the slowest served region (South Africa).

  Suggestion: If your site has low download rate this should be a serious concern for your WebSite managers. If your site is not within 10% of the average speed you probably should consider enhancing your bandwidth or decreasing the size of your top page, or both.

Comparison with Other Metrics
The total page download speed as measured by eValid is the client-side time it takes to download and fully render the page in the eValid browser. This time includes server latency, web latency, and "last mile" latency and thus is a true indicator of the end-user experience.

This value is not to be confused with any of the following commonly published measures:

• Ping Time. The ping time is the amount of time that it takes for a WebSite to respond to a simple HTTP request. On most systems this is measured with the ping www.domain.com command. This is the metric used by Keynote, for example, among very many others.

• HTTP Base Page Retrieval. This is the time it takes for the HTTP protocol to download the base page of the indicated URL. This time usually does not include the time need to download any LINK'd or included files and the time to fully render the page. This metric is used by Alertsite, for example, among many others.

• Augmented HTTP Base Page Retrieval. This is the time it takes for the HTTP protocol to download the base page of the indicated URL plus the time it takes to download any linked files. Because the base page parsing is not down with a browser, this time measurement leaves out all of the client-side parsing effort, leaves out page rendering time and does not take into account the ability of the browser to multi-thread download of page components. This metric is used by Gomez, for example, among several others.

Summary
It's very clear that the faster a top page is to download to the end user the better the perception of the company WebSite is going to be by that user. Some companies, apparently unaware of some basic rules of WebSite design, go overboard on top page complexity, and size at the expense of user satisfaction. Perhaps webmasters just don't actually know the effect at the receiving end of their top page designs? We believe that in most cases small changes in top page design combined with changes in web server infrastructure will make dramatic changes in perceived performance.