Presentation on theme: "OPAC stats presented so quickly theres no time for snark on #code4lib Bill Dueber, University of Michigan But does anyone use it? Note – lots of stuff."— Presentation transcript:
OPAC stats presented so quickly theres no time for snark on #code4lib Bill Dueber, University of Michigan But does anyone use it? Note – lots of stuff in the presenters notes. Dont miss it. NEW!! NOW FEATURING ACCURATE DATA!
Why did the numbers change? HathiTrust and Mirlyn share a common code base and Solr backend I was incorrectly logging HT search events and (correctly) ignoring all other HT activity …so the apparent number of single-search sessions was grossly inflated NEW SLIDE!!
tl;dr Relevancy ranking is incredibly important. Everything else is ignored by almost everyone.
The argument for statistics (as opposed to just asking the Reference Librarians)
Our statistical universe Roughly 750K sessions with searches in 2010 Throw out sessions from known staff IP addresses Because, really, talk about self-selected freaks… Get 724K sessions, 1.67M searches WRONG
Our statistical universe Roughly 500K sessions with searches in 2010 Throw out sessions from known staff IP addresses Because, really, talk about self-selected freaks… Get 485K sessions, 1.5M searches (avg. 3.1 searches/session)
First wake-up call 45% of all sessions have exactly one action: a search WRONG
First wake-up call 17% of all sessions have exactly one action: a search
Corollary one Only 17% of all sessions involve someone seeing the Record View 12% of those (2 percentage points) are from See Holdings WRONG
Corollary one In only 28% of all sessions does the user see the Record View In only 37% of sessions does a user interact with a specific record (either by doing something to get to the Record View or clicking on an eLink for fulltext.) Are they writing down call numbers? Having failed searches?
Second wake-up call Anything not at the top of the screen is ignored
Place in result set of records touched ResultPercentCumulative 140% 212%52% 37%59% 44%63% 53%66% WRONG
Place in result set of records touched ResultPercentCumulative 144% 214%58% 37%65% 45%70% 53%73% 6376%
What does that mean? People do a lot of known-item searches and/or people really, really trust your relevancy ranking
Make sure your relevancy ranking is really, really good.
Percentage of sessions that… Use a facet: 4% Use a canned search (e.g. author or subject link): 2.6% Export records/search: 1.3% Prev/Next/Back: 0.8% WRONG
Percentage of sessions that… Use a facet: 7% Use a canned search (e.g. author or subject link): 4% Export records/search: 2% Prev/Next/Back: 1.4%
If youre interested Ill strip identifiers out of the data and provide an sqlite3 database after the conference …once authority to do so has been debated and eventually granted by the correct set of committees and subcommittees. Bill Dueber email@example.com