Some notes from UKSG Lib-Stats meeting 2007

40 people met to try and better understand the following:

** E-Books - Cliff asked if it was worth collecting/analysing e-book data as only 4 vendors are currently COUNTER Compliant. Most folks thought it was important to collect any kind of data for historical reasons, mainly to see how usage patterns change over time.
Peter Shepherd (COUNTER Director) and Jenny Walker from XREFER gave a useful explanation of what was meant by the term SECTION as defined in COUNTER BRA. A section is a chapter in a text book, or a single entry in a reference work such as a dictionary or encyclopaedia. Jenny quoted from experience at XREFER who are aggregators for many  publishers. Implementing COUNTER was not a big deal and was pretty easy to push through at management level, so no real excuse for others to become COUNTER compliant. 
Jenny thought we would not have the problem of merging stats from aggregators and publishers, which is so prevalent in e-journal publishing.
One very relevant point is how to collect meaningful stats from services like Knovel, which combines database, e-book chapters, interactive charts and spreadsheet data?
Important uses of stats for e-books might include looking at user sessions to determine who uses what, and how they are being used. So... are reading patterns linear - chapter by chapter for whole book, or just selected chapters maybe reflecting use of a reading list on a VLE site. For the first time it may be possible to accurately determine the learning habits of our undergraduate learners:
* What's the average session time
* what's the ave time taken to navigate through an e-book
* What time/day/month is the most popular for browsing (exams?)
* are reading list titles downloaded/printed most 

The ambitious JISC Observatory Project aims to do this:
 ..."To provide an in-depth understanding of how core reading materials are actually used in teaching and learning rather than just providing statistics on the number of downloads, pages printed, etc"
Are the current stats providing this data reliable enough...?
...those of us at the sharp end of collecting stats know how difficult it is just to get the basics right!

http://www.jisc-collections.ac.uk/projects_and_reports/coll_ebooksproject.aspx?keywords=observator

===

**Journal Archives - Cliff asked if we really needed separate stats for backfiles of current journals.
The majority thought we did, even for those archives which are free at the point of use, or included in the main subscription for current titles.
COUNTER has a separate (voluntary) code JR1A enabling publishers to implement this. Peter S from COUNTER suggested this would be reviewed when necessary.
At the moment even if JR1A is available, it is only a sub-set of the common JR1 report, so archive stats would have to be subtracted from JR1 to get a "true" total (for JR1).
The question of Open Access titles  was raised. Some titles in a backfile may be also be available as free content. If a title is available form several different "mirror" sites can the usage ever be known??
Cliff to distribute listing of current Archive availability.
It was agreed that if folks are buying archives to ask if/when separate stats are available.
====

**Databases - has the distorting impact of technology become so acute that the number of "Searches"  no longer function as a key metric?
Patricia Brennan from Thomson Scientific explained that the nature of federated searching within services such as CSA, WOK, and from library portals such as Metalib has confused the definition of a search when applied to usage statistics. 
Marion Tattersal (Sheffield) suggested that "sessions" were perhaps more instructive than searches when it came to a useful measure. The main problem is that an inflated count of searches occurs when federated metasearch engines are employed - one common search statement will aggregate across many databases, resulting in so many results that only a fraction are actually used - but many are recorded. Stats are consequently distorted, and are not very informative at all.
Cliff suggested that in the interests of getting any analysis right, the federated search services should not be compared with non cross-searchable services such as Scopus, Inspec etc.

=====

Finally thanks to Pat again, who gave a brief insight into how SUSHI will revolutionise the way we harvest our usage data, and thereby save many hours of work. SUSHI is a web service which can be set up to automatically download XML data from participating agents. Let's hope vendors get connected much quicker than for e-books COUNTER compliancy!

(cs 2007)