One of the greatest observation I made was while making WhatupNYU.com for the individual users. Through some research I was able to change some of the settings to check how many individuals would visit as the site gets slower. I was able to run some A/B testing on the site after releasing it.
First lets look at some data of the individuals:
During the first three hours of launch there were over 5,200 unique hits on the website from mostly NYU students, and there are about 50-120 individiauls who visit the website on a daily basis.
So the concept of the site is that everytime you load the page it takes content from either the cache or Facebook and displays it to the user. I've been able to test some of the results to see which one would have a greater bounce rate. I found that when it loads from cache there is a bounce rate of only 40% when in comparison to when it loads from Facebook it's over 70%. That's a very clear result - the faster a website is the lower the bounce rate would be.
Moreover, I've also done some testing on how many individuals stay on the website if I change the amount of posts are loaded from Facebook or from cache. When the data is loaded from the cache, and you load more content - there's a lower bounce rate and also an individual stays longer. So even though the data is loading slower (as there's more of it loading from the cache),individuals want to stay on the site longer as there is more posts to look at.
This is an interesting idea - if the data, in this case posts, is relevent to the individual having more of it wouldn't make the individual close the website right away as now they have more to consume. In this case the bounce rate was just above 25%, and it was loading over 500 posts.
However, in the case of loading the data directly from the Facebook API having more data did increase the bounce rate, and it was over 80% when loading just a mere 200 posts.
Lastly, our solution was not the most efficient solution, but the quickest solution we could use to build a tool that took data from the Facebook RSS feed and displayed it on the site. The A/B testing was also performed after the first 3 hours of the website.