QBike has excelled at search for years using a fast, proprietary algorithm based on - hold your hats - the grep command! Over the course of many months, I performed optimizations and tracked speed results using UNIX time, perl timers, etc., and found that grep was indeed a great basis to build a search engine. However, at some point, I did run into limitations.
At the suggestion of my colleague Joe Sill, I began building new search functionality based on SOLR. Using SOLR indexing, I was able to weave in product popularity, outbound click ranking, search result page placement indexing, and other valuable data I had been tracking for years. Doing so allowed me to order search results in ways never before possible using grep.
I'm at the stage now where I run both SOLR and grep searching on QBike. I call SOLR search "fuzzy search" since, indeed, SOLR uses it's own proprietary rank algorithm based on tokens and stems to return results. I call grep search on QBike "exact search" since, in fact, it does tend to really accurately hone in on exactly what the user is intending to find.