Websites that were optimized for both feature phones and smartphones saw the best performance in March, according to Keynote Systems’ index for Mobile News and Portals, released exclusively to Wireless Week.

Most sites employ device/browser detection mechanisms to determine the browser type that is requesting the page and will then return a page that is optimized for that particular device/browser type. Different strategies exist; some sites have a generic mobile site optimized for all mobile devices, while others build two versions, one for the smartphone (touch screen device) and a simple version for the feature phone (non-touch screen keypad device).

The feature phone class usually lacks certain features, CPU power and browser capability compared to the smartphone. Therefore, it is important to optimize web pages not only for mobile in general, but also specifically for different browser types so that any mobile device type can still get a good mobile site experience.

Google and Facebook continued to have the fastest page load times out of the 14 sites tested, with 3.85 seconds (1st out of 14) and 3.97 seconds (2nd) respectively for the month of March.

“This is largely because they both rely on few page objects and a lighter relative page weight,” said Herman Ng, mobile performance evangelist for Keynote Systems.

Ng says the number of bytes downloaded on the Google home page is approximately five times higher than Facebook’s, therefore Google has a better throughput transfer per bytes than Facebook. He also aid it’s not surprising to see that Google also has the highest average page load success rate of 99.42 percent (1st).

Out of the 14 sites in the study, 13 sites return a mobile-optimized page to mobile users. Although LinkedIn does have a mobile site, it is not redirecting mobile users to that from all four devices used in the study. The New York Times returned a mobile site for HTC EVO, Droid X and BlackBerry Curve, but it returned a full web page for iPhone users.

The iPhone and many recently launched devices are capable of processing and rendering a full-size web page, but because the page and objects need to be downloaded through higher latency carrier networks, the user experience is affected by longer page load times.

The average page load time for The New York Times is 14.09 seconds (8th) and LinkedIn is 15.75 seconds (11th).

Keynote’s monitoring also picked up some repercussions related to the 9.0 magnitude earthquake on March 11. While many people watched the devastation from their televisions and PCs, many also used mobile devices to access the latest updates from news networks such as CNN.

“This led to a spike in CNN page load time on March 11, followed by a steady slower page load time for iPhone 4 on AT&T for four days,” Ng said, adding that CNN did a good job of holding its ground in performance with an average page load time of 10.45 seconds (5th) and a success rate of 98.28 percent(9th).

Also during the month of March, The Weather Channel temporarily redirected BlackBerry Curve users to a full web page instead of to its mobile page for approximately eight days. The change decreased The Weather Channel’s overall average page load time slightly to 8.35 seconds (4th). On the other hand, out of the six of the 14 sites that optimized their mobile site for devices like the Blackberry Curve and feature phones, four of them had the top five fastest average page load times.

“Bing, with 7.1 seconds, is a good example of a site that provides feature phone users with an optimized page so the user experience is not affected by slower performance,” Ng said.

Keynote generates its monthly numbers by repeatedly testing the sites in the index hourly and around the clock from four locations over the leading four U.S. wireless networks, emulating the browsers of four different devices: the iPhone 4 on AT&T, the HTC EVO (Android) on Sprint, the Motorola Droid X (Android) on Verizon Wireless and the BlackBerry Curve on T-Mobile USA. Data is collected from San Francisco, New York, Dallas and Chicago and then aggregated to provide an overall monthly average in terms of both speed and reliability.