Frank Smash is your average internet user; he casually browses the web most days. Today, he is considering purchasing a Thingamajig. Like many average internet users, Frank decides to visit the Thingamajig website to learn more about the product. Unfortunately for Frank, while Thingamajig spent a lot of time and money developing a beautiful website that has all of the information anyone would ever want, they didn’t spend any effort making it fast–each page takes what seems like an eternity to load. After browsing around a few pages, Frank gives up, leaving the website with a bad impression of the Thingamajig.
Does this story sound familiar to you? If you answered “yes,” read on.
One of the most important parts of user experience on the web is how fast the pages actually load. Research in 2006 showed that 4 seconds is the acceptable threshold for retail websites’ page load time. Even Google understands that: on April 9, 2010, Google announced that site speed would be used to help determine search engine rankings:
Speeding up websites is important — not just to site owners, but to all Internet users. Faster sites create happy users and we’ve seen in our internal studies that when a site responds slowly, visitors spend less time there. But faster sites don’t just improve user experience; recent data shows that improving site speed also reduces operating costs.
As you may have noticed, we have been optimizing the Gustavus website for speed and have achieved some promising results. According to the statistics that Google gives us about gustavus.edu, our pages take on average 1.4 seconds to load, which is faster than 81% of sites. This is down from about 2.5 seconds at the beginning of March.
We believe that a faster website has two solid benefits beyond making our users happier. A faster website saves people real time and prolongs the life of our hardware. According to our analytics, we receive about 1.3 million pageviews each month. If these optimizations reduced page load times by 0.5 seconds (which is a fairly conservative estimate), we just reduced the amount of time people spend waiting for a website by 7.5 days each month. That’s 90 days of waiting time each year. Additionally, we believe that these improvements will help us squeeze some extra time out of our current hardware–and since servers are expensive, this saves the College a fair amount of money.
The speed improvements are the result of a combination of strategies. We leveraged a variety of free tools to determine areas we could improve; Firebug, Page Speed, YSlow, and WebPagetest were all very helpful in this regard. While most of the changes we made were simple configuration tweaks (e.g. sending better HTTP headers for caching, serving static content from a cookieless domain, distributing multiple requests across multiple domain names, enabling Apache’s KeepAlive, etc.) the one component we installed and enabled seemed to have the single most-noticeable improvement: APC (Alternative PHP Cache).
Obviously, we are excited by these results and hope that you share in our enthusiasm.