Page is a not externally linkable
Brett_Tabke - 9:27 pm on Mar 29, 2001 (gmt 0)
New Tool over at SEW: Webpage Size Checker [searchengineworld.com] WebPage Size and Speed I'm am a huge believer in keeping pages small. All the studies show, that users are very sensitive to page size and/or download time. Speed is Life For an example of how to do it: see Google and Yahoo. For an example of how not to do it, see CNN and ESPN. Server: Minimalist: Bleeding Edge: Design for who? Tips from a Speed Freak: Mod_Gzip 28.8k Take your logs and resolve the ip addresses. Throw out known cable users (home,rr) and any isp domain name with the word "cable" or "dsl" in them. Assume the users that are left are using 28.8k-56k modems. That is how many users are connecting at lower speeds to your site. Then compare the time from the first request of the html, to the last "object" request that is on that page (a graphic). That is a good indicator of how long it is taking to download the page for your users. Track Test
I got to thinking about this a great deal this week (thanks for the inspiration).
Page load speed. I am convinced it is everything. It is the difference between a successful site and a non successful site. It is not easy an easy task to reduce page size, but I try to keep all pages under 20k of html, and less than 30k total with graphics.
Try a test by putting a counter at the top of your page as the first thing and a counter at the bottom of the page. Use different styles of counters from time to time (javascript, ssi, graphic). Then compare the two counters. If there is a 4% or greater difference between those counters that points to a server problem or a page size problem. It will indicate people are not getting all the content.
It needs to work in 100% of the browsers on the net. That include browsers such as Scooter, IE, Netscape, Lynx, Opera, Slurp, and Googlebot. That doesn't mean it has to look the same, just that those agents can get to all the content with those browsers. Obviously differences will exist in things such as graphic support.
Leading edge or nonstandard technology. There is a reason they call it, the bleeding edge. Stay far away from anything nonstandard or requires your user to do extra work. (that includes, shockwave, java, and other attempts at embedded tech such as active-x or vbscript). You can't afford to lose (or slight) 10% of your audience. Granted, if you are running something such as a WAP site, WML would be appropriate.
Who is the typical user? Much of the common wisdom is that the average user is using around a P2 at 333mhz with 32 to 64meg of ram and 56k dialup. If that is the "typical" web user, that means there are a bunch of users running less than that. In order to be inclusive, you have to design for a whole lot less than that.
Because of modem compression, HTML will download twice as fast for most people as graphics. (eg: 20k of html will download as fast as a 10k jpg).
If your server supports it, you might try experimenting with Apache Mod_GZIP. It can reduce your html bandwidth and download times by 50%.
Try simulating walking through your site at 28.8k. Assume 20% of your users are on 28.8 or 28.8k performing systems.
The number of page reloads you are getting. Those can also point to a server problem. Make sure your logging or counter software doesn't automatically throw out duplicate requests - that is important data.
Try a big page and then try a ultra small page. Notice the difference in the page views per user. Between a 10-15k page and a 40k page, the difference will be dramatic (it may be good to have medical personnel handy).
External Related: