I put together the first search statistics as part of an internal discussion on how often people change the default search - about 65% of the time - and what they change the search to.
I've added overall search statistics to <http://lib.tamu.edu/directory/bponsfor/opac-usage-stats/> and annotated the files with some notes about what those searches are supposed to be doing (as well as where the stats have problems). These statistics, btw, are just a 10 day (more or less) sample from last fall.
As to the question of how often people click on a link in a bib record, the answer is less than 7% overall. And, interestingly, the public more than library people.
Bennett
>>> On 2/6/2008 at 4:15 AM, Weinheimer Jim <j.weinheimer_at_AUR.EDU> wrote:
> If I may be allowed to do a little armchair analysis, these statistics seem
> to me to be based only on the initial search, i.e. when people first come to
> the catalog and type their search into the search box. If we want to measure
> the actual use of a catalog, it should be continued to include the hyperlinks
> based on the headings within the records. Therefore, if somebody thinks "I
> want Roman history in the time of Cicero," they would probably type in "roman
> history cicero" and get something like this: (this link searches princeton's
> catalog as a general keyword search, I think it will work)
> http://catalog.princeton.edu/cgi-bin/Pwebrecon.cgi?SAB1=roman+history+cicero&B
> OOL1=all+of+these&FLD1=Keyword+Anywhere+%28GKEY%29&GRP1=AND+with+next+set&SAB2=&B
> OOL2=all+of+these&FLD2=Keyword+Anywhere+%28GKEY%29&GRP2=AND+with+next+set&SAB3=&B
> OOL3=all+of+these&FLD3=Keyword+Anywhere+%28GKEY%29&PID=UKW7ypc9EtTUuUe2aXbiQ0_M
> kZ&SEQ=20080206045027&CNT=50&HIST=1
>
> and the user should be able to find the heading(s) that they really want; in
> this case: Rome --History --Republic, 265-30 B.C. (This is how the catalog is
> supposed to deal with natural language questions)
> In this example, the correct heading doesn't come up until record 15 (an
> earlier record has a "Fiction" example), so many users would probably have
> given up before finding the correct heading. Consequently, I would consider
> this an example of a failed search.
>
> In any case, I would think that a valid way of looking at the use of the
> catalog is how these headings are used: do people search first on keyword
> "mark twain" and then click on "Twain, Mark, 1835-1910" when they find a
> correct record? If so, something like this should be considered mostly a
> successful search. Very few people, including experienced catalogers, know
> the precise form of a name, title or subject heading, and will mostly opt for
> keyword as an initial search.
>
> If I am wrong in my assumption and the statistics are not based only on the
> initial search, but actually includes all the hypertext heading links, then
> this would be a serious problem for the catalog, since it would mean that
> people are not clicking on the headings once they are in a record. In other
> words, if somebody searches "leo tolstoi" and they see a relevant record with
> "Tolstoy, Leo, graf, 1828-1910" and they do not click on that link, then this
> would indicate that there is something that needs to be fixed.
>
> Jim Weinheimer
>
>
>> Actual stats are at
>> <http:/
>> /lib.tamu.
>> edu/direct
>> ory/bponsf
>> or/opac-us
>> age-stats/
>> Count%20of
>> %20first%2
>> 0web%20searches%20-%20staff%20and%20public.xls/view>. Duh!
>>
>> Bennett
>>
>> >>> On 2/5/2008 at 12:54 PM, Bennett Ponsford
>> <BPonsfor_at_LIB-GW.TAMU.EDU> wrote:
>> > Additional rough stats on Which search users started with (our default is
>> > "keyword sorted by relevance"). Library people use
>> "keyword" and "title
>> > starts with" almost equally, however the public uses their most
>> popular title
>> > search (title keyword) less than 10% of the time. I was also
>> surprised by
>> > the number of the public who start with "expert keyword",
>> particularly in
>> > comparison to the usage by library types.
>> >
>> > In terms of talking to our users, we did a survey this fall and asked
>> people
>> > whether they had last searched the catalog for something where they already
>> > knew the author or title or something on a subject. 78% of
>> undergraduates
>> > searched by subject, whereas 64% of grad. students and 68% of faculty were
>> > looking for a known item. And the big theme of the follow-up
>> focus groups
>> > was: simplify, simplify, simplify. What we need is a simple
>> keyword search
>> > that guides undergraduates to what they need while not getting in the way
>> of
>> > people who know what they're looking for.
>> >
>> >
>> >
>> >
>> >
>> > Bennett
>> >
>> >
>> >
>> >
>> > Bennett Claire Ponsford
>> > Digital Services Librarian
>> > University Libraries
>> > Texas A&M University
>> > bennett.ponsford_at_tamu.edu
>> >
>> > TAMU 5000 | College Station, TX 77843
>> >
>> > Tel. 979.845.0877 | Fax. 979.845.6238
>> >
>> >
>> > http://library.tamu.edu
>> >
>> >
>> >
>> >>>> On 2/5/2008 at 10:55 AM, Karen Coyle <kcoyle_at_KCOYLE.NET>
>> wrote:
>> >> Janet Hill wrote:
>> >>> Repeat after me: There is no typical
>> user. There is no typical user.
>> >>> There is no typical user.
>> >>> Persuade us with actual data, carefully collected, clearly
>> defined, and
>> >>> thoughtfully considered.
>> >>>
>> >>
>> >> And in that spirit, the first question that came to my mind when I
>> >> looked at Selden Deemer's stats was: I'd really like to know WHO did
>> >> those searches. Not individually, but in some grouping. In particular,
>> >> I'd like to see the stats for staff use separated from non-staff. This,
>> >> of course, is not going to be 100% accurate because most of the time we
>> >> don't ask people to identify themselves at public access points. But at
>> >> least in the back rooms and on any machine with a staff sign-in. It
>> >> could turn out that there are some important staff needs that should be
>> >> addressed, even if those features are rarely used by the public.
>> >>
>> >> Then overall statistics can help us make design decisions: what
>> >> functions have to be very efficient because they will be used
>> >> frequently? What functions can we provide in a less efficient manner
>> >> because they are rarely used and don't impact overall response time?
>> >>
>> >> --
>> >> -----------------------------------
>> >> Karen Coyle / Digital Library Consultant
>> >> kcoyle@kcoyle.net http://www.kcoyle.net
>> >> ph.: 510-540-7596 skype: kcoylenet
>> >> fx.: 510-848-3913
>> >> mo.: 510-435-8234
>> >> ------------------------------------
Received on Wed Feb 06 2008 - 13:17:38 EST