I talked with Christina Wodtke about Search Patterns. Christina moves fast and wears many hats including Principal Product Manager at LinkedIn, Founder of Cucina Media, Publisher of Boxes & Arrows, and author of Blueprints for the Web. Plus, she co-founded the IA Institute and led the ux design group at Yahoo! responsible for the revival of search and the reinvention of shopping.
Christina believes search is perfect for patterns because behavior is fairly stable and predictable. As she noted in Long Tails and Short Queries, the patterns observed by Amanda Spink in 1997 haven't changed much. For instance, most web queries are still short, 2 to 3 terms, and include little refinement.
In that context, query disambiguation is incredibly hard, and the challenge can be summed up as: "one more word!" How do we entice users to share just a little more about their intent? Clearly Search Assist is an attempt to do just that. Christina also noted that "size matters!" Approaches that work for web search may not work for site search, and vice versa. For instance, Best Bets won't scale for web search (though Wikipedia and Mahalo sorta aim to fill that gap).
Finally, Christina was generous enough to share some unpublished writing she's done on the topic of search. Here are a few of my favorite excerpts:
Fast and ugly is better than slow and pretty.
Scan time is as important as load time.
If you want to see how hummingbirds fly, you'll have a hard time. They move so fast they become a blur. But, if you ask them to slow down, they can't stay aloft, and you won't learn anything...search is so fast, that if the user is forced to think [e.g., in a usability test], they slow down too much and behave unnaturally.
With search, the physical actions are few: look, type, click. What you really want to know is hidden, even from the person searching.
Thanks Christina, for slowing down for moment, to share your search insights!