I just noticed that the Velocity Online Conference is coming up next week. This conference is focused on best practices in performance and operations for web applications to improve the user experience as well as a company's bottom line.
I've written previously about the importance of perceived performance in the design of a SaaS application. For SaaS, where customers can easily switch to another provider, user satisfaction is critically important. Low perceived performance can lead to low satisfaction and high customer churn. High perceived performance can result in high customer satisfaction and stability.
But what is the empirical evidence and how much of a change in perceived performance is necessary to have a significant impact? Well I just came across some hard data - metrics for page views, amount of interaction/use and online revenue presented at Velocity 2009 from some data-heavy players.
The empirical data on the impact of web application performance
Microsoft reported that with Bing, a 2 second slowdown in response time reduced the number of searches by 1.8% and reduced revenue/user by 4.3%. That a lot of money left on the table.
Google reported that as little as a 400 millisecond delay resulted in 0.59% fewer searches per user. But perhaps more interesting, even after the delay was removed, these users still had 0.21% fewer searches, indicating that a slower user experience affected long term behavior. While Google did not report revenue directly, fewer searches likely means fewer AdWord clicks.
Shopzilla had the most complete data on the impact of performance on the bottom line. A year-long performance redesign reduced response times by 5 seconds (from ~7 seconds to ~2 seconds). This resulted in a 7-12% increase in revenue.
But it's really perceived performance that matters
This empirical "bottom line" data becomes even more interesting when it is reviewed in the context of the study 'The Truth About Download Times. This study found that users do not rate the download speeds of Web pages based on the actual stopwatch-clocked download speeds. Perceived speed is dependent on how well "users successfully complete their tasks on a site."
In other words, it's not just page load time that matters. It is the time it takes for a user to successfully complete a task that has the real impact. In the case of the Velocity 2009 data, users couldn't interact with a page until it loaded. Those few milliseconds of additional time, prevented users from accomplishing their search tasks. That signficantly impacted the bottom line.
Designing RIAs for perceived performance
So how do you design a SaaS to enable the user to perform real work more productively - to feel fast, responsive and streamlined. Our RIA designers focus on a number of application flow and design areas.
- Determine the high value scenarios and then reduce the number of actions required to accomplish them
- Minimize the need for complete page refreshes and reloads - the waiting between screens creates serious friction with the user
- Incrementally add information and functionality to a page, based on an analysis of how users process the information (e.g. load a modular component only when needed or in the background based on predictive analysis)
- Progressively download data locally to avoid round-tripping to the database (HTML 5 is a great development here)
- Validate designs (visual, information, interaction and architectural) to determine where users perceive adequate performance vs. where they get slowed down and wait for the web application to catch up
- Enable queries or actions to be canceled - it gives the user a feeling of control and lets them move on to something else they consider more important than waiting
- Add reliable indicators of progress on activities (i.e. don't let users get frustrated in front of a screen not knowing how long something will take)
RIAs are about more than features, graphics and nice visuals. Optimizing perceived and actual performance has a real and quantifiable impact on the bottom line. The more productive an RIA/SaaS design, the more users interact with software.
Categories: RIA, User Experience