1 Jul 2010
While at Velocity last week, I took in so many sessions and workshops, by the end it felt like data was shooting out my eyeballs. Fortunately, my brain was saved from short-circuiting by the fact that many of the slide decks have been published online for a more leisurely, synapse-friendly review.
In the past week, I’ve spent a fair bit of time going through the presentations and started collecting my favorite performance graphs and charts. It occurred to me that, taken all together, these provide a pretty nifty snapshot of the current state of web performance. Let’s take a quick tour together:
Site Speed and User Behavior
With both of these graphs, you can see the dramatic relationship between landing page speed, bounce rate and pages viewed per visit. But the drama really comes into play in the performance bottom line, conversion:
Hooman and I weren’t the only ones providing solid data on user behavior. Many of our colleagues came with compelling data of their own.
In his presentation Performance Testing: Putting Cloud Customers Back in the Driver’s Seat, Imad Mouline presented this great graph showing that speeding up a page by just 4 seconds decreases abandonment by a whopping 25%:
These next two graphs demonstrate two important things to remember about web users. First, as Yahoo’s Stoyan Stefanov demonstrated in his Psychology of Performance presentation, for most of us there’s a distinct difference between perceived speed and actual speed. When it comes to web performance, this difference works against us:
But as Lenny Rachitsky showed in his The Upside of Downtime session, even when users’ perception of speed and responsiveness works against you, you can still earn their trust and respect in a crisis, so long as you’re willing to be honest and communicative:
So How Does User Experience Affect Business Metrics?
There were a handful of messages that were threaded throughout many of the presentations and discussions at Velocity. The relationship between performance and the bottom line was one of them. In his excellent Metrics 101 workshop, Sean Power presented some visuals that hammered these points home.
In the short term, Sean used these next two graphs to show the prohibitive costs of downtime, even if your site is unavailable for just one hour:
Over the slightly longer term – a one-week period – Sean shows that as latency increases, conversion rates drop:
In this next graph, Sean shows the long-term effects of poor performance, which can take years. We can see that a poorly performing website suffers not one but two waves of abandonment, as users spread the word of their poor experience and drive other users away from the site.
What Are the Causes of Slow Performance?
There’s no single answer to this question, of course. I’ve addressed this elsewhere (along with some pretty decent graphs, if I do say so). Here are two more to add to the mix.
In the same Metrics 101 workshop, Sean showed this awesome diagram, which pretty much sums up the dialog between a browser and a data center during a typical page request:
Further complicating this dialog are the calls to to third-party content providers, such as ads and widgets. Google did a survey of a number of sites, testing their page load time with and without popular third-party widgets. They documented the results on this chart in their session Don’t Let Third Parties Slow You Down:
What About the Mobile Web?
In my brief keynote at Velocity, I talked about mobile web use and the fact that, at this point, pretty much all we know is that we don’t know much, but we’d better figure things out quickly because the mobile web is a tidal wave that has not yet even begun to crest. Maximiliano Firtman, in his Mobile Web High Performance session, provided a couple of graphs that illustrate this:
(If anyone out there has some solid data on performance and the mobile web, I’m keeping notes for a future post. Drop me a line in the comments or email me at email@example.com.)
And finally, the last pair of graphs…
Practical Planning for Performance
In the last part of his Metrics 101 workshop, Sean offered a realistic step-by-step plan for dev/ops folks to measure performance and outcomes in their organization. He said that if you create only one graph for your site, it should be one that looks like this, which shows the direct impact of page load time on conversions:
And because even the most optimized site still goes down from time to time, Lenny shared this excellent, common-sense plan for communicating with your users:
Have I missed any of your favorite presentations and visuals? Let me know in the comments.