Page load speed and pages per visitors

My empirical data shows that page load speed helps

[lammert]

One of my websites is almost static in content and has a fairly constant stream of visitors, mainly from referring sites and the search engines. In the last months I have played around with some options to increase page load speed. I constantly monitored the page load speed mentioned in Google Webmaster Tools, and at the same time looked at the number of pages per visitor. The sample size of visitors we talk about is about 100,000 unique visitors per month, over a period of five months.

The optimizations I have done had no direct visible effect on the page layout. What I did was combining multiple images to one to reduce HTTP requests, push JavaScript to the bottom so scripts are not postponing rendering etc. I have reverted the changes on different occasions to be sure that the pages/visitor values were related to the page load speed, and not any other reasons like seasonable differences etc.

Over these five months I have seen a fairly consistent relation between the two. The faster the page loads, the more pages a visitor wants to view. My figures:

4.7 secs: 1.51 pages per visitor
3.0 secs: 1.56 pages per visitor
2.7 secs: 1.58 pages per visitor

Anyone else with statistics to share?

[LynxSI]

I would agree with this theory although I don't have the sample size you do.

People have incredibly low patience and tolerance for slow sites these days. If they can just click-click-click through pages they will keep searching for what they want. If every click is costing them 2+ extra seconds they will get bored if the page they are on is not directly relevant to their search.

I've definately found that faster page load speeds like the improvements you have done will contribute to a better page rank in google! (People happy = Google happy)

[castor_t]

@lammert - You made it 2.7 seconds from 4.7 seconds without any difference in the visible output? Would you mind sharing the changes you made on the pages.

[lammert]

The main reduction of load time was caused by combining many images to one. This site has a menu bar which originally consisted of one image per selection. All these images have now been combined in one large image with a client-side map to link areas of that large image to different URLs.

JavaScript files were also combined to one large JavaScript file. Many browsers stop downloading and rendering other content while downloading and parsing JavaScript files. Combining JavaScript files in one large file causes only one such rendering stop which on average gives a faster overall rendering speed than loading smaller JavaScript files sequentially.

JavaScript has been moved to the last part of the HTML code. This doesn't change the overall load time much, but it allows the first part of the page to be rendered and displayed in the browser before the JavaScript part is starting to delay the processing. The visitor can start reading while the browser parses the last part of the page.