With All That Traffic, Does Facebook Test Its Site?
Most of the bits spilled over Facebook this week were spent beating up Beacon, but at UIE’s website, Jared Spool’s article on The Five Usability Challenges of Web-Based Applications leads with a recap of an earlier Facebook misadventure — the introduction of its MiniFeed feature.
In case you missed it: Back when MiniFeed was first launched, within 24 hours of its release, nearly 10% of Facebook’s users rejected Mini Feed (and petitioned against it). Jared points out that this type of user-behavior is unique to web-based applications- we don’t typically organize around our dissatisfaction with desktop apps.
Give Jared’s post a close read for some great insights, but here are the 5 challenges he cites, along with some quotes pulled to highlight a nagging question of my own:
Five Usability Challenges of Web-Based Applications
- Scalability: “A contributor to Facebook’s mini-feed debacle was the scale of their design. Facebook, making any change to their site, instantly affects eight million people. If even one percent has issues with the change, that’s 80,000 affected users.”
- Visual Design
- Change Management: “The designers at Facebook learned the hard way that quick changes to the application, even if the team thinks it’s an improvement, can have serious negative results if done incorrectly…. We’re now seeing teams start to design the change process along with designing the changes themselves.”
So here’s what’s got me scratching my head:
Why doesn’t Facebook use their incredible scale as an asset when they introduce changes to their site? Why not test these changes with a statistically significant fraction of their traffic, and then optimize based on this sample’s behavior? (Think about the way the Amazon redesign trickled out, and the insights likely gathered along the way.)
With our own site testing work, we’ve been fortunate to see some strong results for our clients. We encourage folks to test big differences – shouts not whispers. Whenever you’re testing, some experiments produce the anticipated result, some don’t. But each time you test, you can introduce a big change, in a well-defined area of your site – to a deliberately limited portion of your visitors.
That means that while increased conversion is almost always the ultimate goal, every test lets you methodically introduce site changes in a way that respects your users and your brand.