Thanks for the comment. TwigKit provides a single API not just for search suggestions, but for everything ranging from query to response and including things like faceted navigation and data visualisation. The TwigKit API works across many search engines, so the same syntax applies whether the platform is Solr, Fast, or any other.

Keep an eye on the blog, we’re going to be releasing our own search suggestion implementation under GPL in the coming weeks!

[1] http://www.sematext.com/products/autocomplete/index.html

We didn’t do any official studies around cognitive burden vs. utility, but intuitively I agree with Tyler’s assessments.

I guess that’s part of what I was trying to get across in the chart with the “cognitive burden” axis. My general feeling is that ‘completion’ helps the user more than it gets in their way, while ‘instant results’ runs much more of a risk of overwhelming people, with ‘query suggestion’ somewhere in the middle.

The disclaimer is that no user testing went into this post. But, if someone were to attempt to measure cognitive burden, my assumption is that they would measure the time that the user spends looking at the suggestions. That is, the number of milliseconds from when the search suggestions are shown, until the user makes their selection.

Utility could be slightly trickier to measure. Perhaps it’s the time it takes to find something (from the moment the user begins their query until they arrive at their destination) divided by the time it would have taken without any form of as-you-type suggestions.

I would be interested in other’s thoughts on how to measure cognitive burden and utility, as well as comments on whether or not you think my chart is very accurate.

Like you I believe that search suggestions add something to the user experience, and I always look for opportunities to incorporate it in my designs. Still, I feel insecure about the true usefulness of suggestions, and how I should go about finding it out. I mean, does it never happen that completion disrupts the user?

I wonder about the 2 axes in your chart; utility and cognitive burden. How specifically would you measure the cognitive burden of search suggestion features? And likewise, how specifically do you observe utility?

A likely answer to these questions is user testing, and I agree to that. But I would still like to hear your thoughts.

Andy, yes it’s ironic that Bing shows up as the first result for “search.” Speaking of which, there’s a whole collection of funny search suggestions over at http://searchenginesuggestions.com/.

BTW Have you ever searched for “Search” in Google? Bing.com is top! Weird (slide 13)