eValid™ -- Automated Web Quality Solution Browser-Based, Client-Side, Functional Testing & Validation, Load & Performance Tuning, Page Timing, Website Analysis, and Rich Internet Application Monitoring.

Summary
How do the websites of the what Keynote claims to be the Top 10 Fast Sites actually compare in terms of Best Online Experience? Does having the speed record for the top page on the site really imply quality?

Goals
For this comparative analysis of websites we define two important dimensions that allow us to rank order each website in terms of Best Online Experience:

• Ease of Navigation. This factor is a combination of the average rate of download of pages and the number of successful page retrievals. A high score for ease of navigation means that users will have quick responses as they work on the website, and that they will find very few broken or unavailable pages.
• Quality of Maintenance. This factor is based on the relative age of pages found on the website. A well-maintained website has very-current information, and indicates that great care has been taken to assure the quality of the information.

We used the eValid website analysis engine, programmed (see below) to collect specific metrics that contribute to these factors. To assure accuracy the measurements were done over a short time frame and using identical access capabilities.

Results
The measurements we made are the top 10 sites of the Keynote Business 40 sites produced the relative Quality of Online Experience comparison chart shown. For simplicity the data shown are all normalized relative values.

The best-performing websites are in the upper right of the chart, and the worst-performing websites are in the lower left of the chart.

Individual Summary Results
The table below shows all of the results used to construct the comparative chart. Click to see the eValid Summary Report with the raw data. Click to see the 3D-SiteMap for each website. Here are Animations and Usage Instructions for the 3D-SiteMaps. Right-click anywhere on any 3D-SiteMap for documentation. Here are some Usage Hints if you want to dig deeper in the 3D-SiteMaps.

#	Company	Click For Summary Report	Comments	Click For SiteMap	Comments
1	AltaVista www.altavista.com	eValid Report	B-: A very large, but relatively slow, yet very current, site even though the average page size is small. There a loads of broken links.		B-: Good fanout, but a lot of cross-layer links, and some accumulation nodes.
2	Apple www.apple.com	eValid Report	A: A very large, very fast site with ~10KB pages that are relatively current and with a very low unavailability ratio.		B-: Good structure for a large site, but excessive fanout and too many layers accessible from the top.
3	Dell www.dell.com	eValid Report	A: A moderately large (1700 pages to depth 3) site, very fast, relatively small pages, very clean, reasonably current.		A: Very clean organization except for backward links, good interconnection, and not overly complex.
4	FedEx www.fedex.com	eValid Report	A: A moderately sized site (~1000 pages to depth 3), with very current data and a very low page unavailability percentage.		B+: Classic structure, good dispersals, a few backward links.
5	HP www.hp.com	eValid Report	B: A very large, relatively fast site with annoyingly high unavailable page counts, but very well maintained otherwise.		B+: Quite a few backward links from the root, but very good access to multiple layers, great flexibility.
6	Oracle www.oracle.com	eValid Report	B: A moderately sized site (~1000 pages to depth 3), but fairly dirty. Very fast and very current.		B-: Classic structure, but covers too many layers, has too many backward links, and has a few too many accumulation nodes.
7	Sun www.sun.com	eValid Report	C: This is a very large site, moderately fast, with somewhat larger pages than average (50% over 10KB), and with good link quality.		B: Good organization from the base page, good spread out and good dispersals, a few backward (reverse) links.
8	Symantec www.symantec.com	eValid Report	B: A big, moderately fast site with very good link quality, relatively smaller average page size, and reasonably well maintained.		C: Deeper than most in terms of dependence, some upward dependence, some isolated collection nodes.
9	Ups www.ups.com	eValid Report	A+: A very compact site -- only ~375 pages to depth three -- which is very fast, very clean, very current.		A: Clean structure, a few cross-layer links, but no backward links and good dispersal.
10	Yahoo www.yahoo.com	eValid Report	A: A large and fast website, with relatively small pages, but with annoying rates of page unavailability.		A: Classical structure, good expansion, no cross-overs, good dispersals.

Methodology
To collect the data used in the chart we used the eValid website analysis engine with parameter settings as follows:

• Search Mode. We used the HTTP Normal mode which scans pages in the background without using the cache (all pages are freshly downloaded).
• Depth of Search. All of these tests were done with a maximum tree depth of 3. Most of these websites are many layers deeper.
• Large Page Threshold. We chose a relatively small 10 KByte limit for this factor because user response is usually best when the total page size is small. 10 KBytes @ 5 KBytes/second = 2.0 secs.
• Slow Page Threshold. We chose a 2.0 sec threshold to decide if a page was slow loading.
• Old Page Threshold. Here we chose a threshold of 1 day (24 hrs) to emphasize the stability of the site.
• Broken/Unavailable Link Counts. This factor is expressed as a percentage of the total number of URLs visited.
• Excluded URLs. To make sure that the report was not biased by the existence of long and very-long files -- PDF format files, for example -- these suffixes and a similar set were blocked from consideration.
• Mapped Links Limit. Because these corporate websites tend to be very large we imposed a 5000-link limit in each run. This limit means that eValid maps only the URLs that give rise to the first 5000 links. After the limit is reached eValid continues to search out properties of the already-enumerated URLs that gave rise to those links.
• 3D-SiteMap Limit. For simplicity the 3D-SiteMaps were made with depth 2, 1000-link limits, and don't include graphics URLs.

Web Access Details
All tests were done on a 384 Kbps (peak performance value) ADSL connection. The data was collected with successive eValid runs during normal business hours.