The Gmail auto-complete bug that surfaced this week (and seems to be already fixed) is an interesting case. It’s a feature that was/is so well designed and implemented that, like Joni Mitchell’s lyrics, users didn’t realize how important it was until it was gone. In fact, some users only noticed problems after sending emails to the wrong contacts. They had grown to rely on the correct operation of auto-complete that they trusted it maybe too implicitly.
Beyond the response of users when it went awry, the recipient auto-complete is a more complex feature than it lets on. Of course I’m not privy to the actual algorithm, but it does a fine job at looking at different factors, probably including correspondence frequency and recency, to determine the importance of the contact and to decide what order to present auto-complete options as each letter is entered. Google tends to implement these kinds of data analysis driven features quite well (see also search.)
The takeaway for me from this bug is that users grew so reliant on having the “right” recipient that they trusted it completely, hardly noticing it. Auto-complete worked so fast and so often provided the desired address either on the first or second letter, that it was unnoticeable, but in a good way. I wonder, from the product perspective, how effective these almost invisible features are, seeing as they aren’t even noticed. Do they indicate how “good” a product is in its entirety? Do they cement the user’s loyalty to the product? Is it noticeable when it doesn’t exist in competing products? Maybe they are just part of the build up to that Steve Jobs product ideal: “it just works!”
Regardless, it’s an interesting product lesson and one that I’ll take with me next time I think of “delighting the user.”