Incomplete Downloads -- a measure of good design?

[tedster]

I would like to use the log reports of incomplete downloads as a rough measure of whether a site's pages load fast enough to satisfy visitors. Does anyone know of any benchmark I can use?

Also, I want to be sure I understand the statistic properly. I'm assuming it would measure getting another hit from the same IP before a previous hit was sent, correct?

So as long as the server sent the complete file, no incomplete download would be registered, even if the client never fully loaded the page. Am I right on this?

[tedster]

FWIW, I've decided not to make much out of incomplete downloads as long as they stay minimal. Heck, a really good top of the page could have someone clicking into the site really fast, and stopping a download midstream.

[Brett_Tabke]

I never thought much about it until I had a server that was kinda slow once. A client was running this huge piece of portal software that was bringing the system to it's knees. I started noticing all these reloads and figured out what the problem was.

I don't think you can target incomplete loads, more just reloads. As long as you aren't doing dynamic content or anything that would nix browser caching, I think you can put some stock in the numbers. I normally run about 2-3% of the requests as reloads. It runs lower on the sites with larger pages, and runs lower on smaller paged sites. Sites with heavy external js run upwards of 5% to 7%, while sites with Java applets can run as high as 15%.