speed

According to the Site Performance feature in Google Webmaster Tools, your pages load reeeeealy slow but, other external tools or monitoring services tell a different story.

What should you believe?

First, it's important to understand the differences between these tools, the data they capture and how it's measured.

Page Speed evaluates the performance of a specific web page and individual elements in the browser. As a result, this type of testing may not accurately reflect latency experienced by users. Page Speed is for testing and improving speed for individual pages.

Tools like webpagetest.org and monitoring services often test latency for a specific URL at various times of day and locations around the world. As a result, these kinds of tests may not reflect latency as perceived by users in the region the site targets.

Google Webmaster Tools Site Performance data is collected from actual Google Toolbar users in the same geographic region as the target audience of the site. This data can be measured in several ways. One being, time between when the user clicks on a link "until just before that document’s body.onload() handler is called." If for example, if a user clicks on a link, is then redirected and then redirected again, that delay should be recorded and reflected in Google Webmaster Tools Site Performance data. These are the kinds of delays that impact users and Googlebot and that are totally missing from other tools including analytics.

Speed doesn't currently have a major impact on rankings but, slow pages deter users and hamper crawl efficiency. Crawl efficiency can be a major factor for pages with lower PageRank because " the number of pages Google crawls is roughly proportional to PageRank".

Experiment:
To determine if hosting impacts site speed according to the Site Performance feature in Google Webmaster Tools.

Hypothesis:
Hosting can impact site speed.

Background:
"80% of the end-user response time is spent on the front end." The preceding statement could be interpreted to mean that back end hosting has little impact on front end site performance.

Procedure:
To test this experiment, an existing site was replicated at a new and separately hosted IP address. DNS was changed from the original host IP to the new host IP. A few days later, DNS was changed back to the original host IP. A few days after that, DNS was again changed to the new host IP from the original host IP.

speed graph illustrating hosting impact

Result:
According to Google's Site Performance tool, pages at the new host IP (dashed line) loaded much faster than pages at the original host IP (solid line). There appears to be an obvious and immediate improvement of more than 50% when DNS was initially pointed from the original host to the new host. Similarly there appears to be a decrease in speed when DNS was pointed back at the original host IP from the new host IP and increase when pointed back at the new host IP again. Since DNS was pointed at the new host IP, site speed and performance have continued to improve according to Google Webmaster Tools.

Conclusion:
This experiment seems to indicate a strong correlation between changes in hosting and changes in site performance. This correlation is no real surprise given, the new host is highly rated as fast and reliable. Conventional wisdom is that "hosting" doesn't impact site performance but, I think it's worth testing just in case your site is one of those rare exceptions. :)