Searching for Answers (and Sometimes the Algorithm)

So we’re in the middle of migrating from Primo VE to Primo NDE right now. Ex Libris is calling this the “Next Discovery Experience,” which is a fun name, though it does make me pause and think about how much of the previous discovery experience we actually got to explore before moving on to the next one.

For me, this migration isn’t really about swapping one platform for another. It’s more of an excuse to lean forward and ask what’s actually working, what we’ve learned, and what we’d do differently given the chance.

Those questions keep pulling me back to the same place: search. The interface is just what people see, the visible layer. Underneath it is this whole tangle of decisions about how queries get interpreted, how records get matched, and how results get ranked and put in front of people.

Lately I can’t stop thinking about what actually happens in the gap between a user hitting enter on a search and Primo spitting back results. Why does one search behave exactly like you’d expect, while another takes a totally different path? How is Primo actually reading the words people use in their queries, and what’s really driving what lands at the top of the list?

To being digging into this, I’ve been leaning on some great work by Can Li, who’s done research into Primo’s search behavior and ranking. Using that as a jumping off point, I’ve started poking around using the Primo Search API to get a better feel for how it interprets queries and surfaces results.

More to come...