maximizing library investments in digital collections through better data gathering and analysis...
TRANSCRIPT
Maximizing Library Investments in Digital Collections Through
Better Data Gathering and Analysis (MaxData)
Carol Tenopir ([email protected]) and Donald W. King
Description of MaxData
• Funded by IMLS• University of Tennessee, OhioLink and
Ohio Libraries, University College London CIBER (Centre for Information Behaviour and the Evaluation of Research)
• Comparing and building a cost model for data collection methods (transaction logs and surveys)
Study Objectives
• To compare what different methods of data collection can tell you
• To develop a model that compares costs and effort to the library of collecting and analyzing data from various methods with the benefits of the information you obtain
Study Methods
• Ciber deep log analysis– OhioLink– 5 Ohio Universities– Elsewhere
• User Surveys– 5 Ohio Universities– University of Tennessee– Elsewhere
• Other log data from vendors (COUNTER-compliant, proxy servers, link resolvers, etc.)
• Costs and effort of methods
Why do libraries gather usage data?
1. To make decisions and rethink old ones
2. To demonstrate the value of the library’s collections and services
3. To improve services and collections4. To find their place in comparative
rankings5. Because they are required to
Internally collected or added data can be used to show:
• Comparative amounts of use for databases
• Relative uses per size of user population within subject areas
• Cost per use
• Where users access the digital library
Ciber Highlights
• 50% of journals account for 93% of use, but 99% of titles were used at least once in a 7-month test period
• Health sciences titles are used more than any others• OhioLink users choose to download a full article 3 times
more often than view just an abstract• Half of all sessions viewed an article; average was 2
articles per session• Users who searched (rather than using the alpha or
subject lists) tended to view more articles and viewed more older articles
Usage logs give much useful data, but…
• Logs don’t show why or outcomes
• Requests or downloads may not equal use or satisfaction
• Log sessions may be difficult to differentiate or compare across systems
• For privacy or other reasons, logs do not show behavior by demographic groups
• Logs show only a fraction of total use
Our surveys:
• Have been used since 1970s• Include over 30,000 responses• Provide trends since 1977• University surveys include:
– 2 national surveys of scientists (1977, 1984)– Astronomers and pediatricians who belong to their
main professional societies (2003, 2004)– 3 University of Tenn. Surveys (1993, 2001, 2003)– Drexel University (2002)– University of Pittsburgh (2003)– 2 Australian universities (2004-2005)
Our Surveys are Designed to:• Provide a complete picture of information seeking
and reading patterns and how libraries contribute to overall information needs
• Distinguish:– Sources of articles read– How articles are identified/found– Time and depth of reading– Age of articles read– Format of articles read– Outcomes from reading– Value of reading from library and elsewhere
Our Surveys Establish Factors that Affect Reader Choices:
• Ease of use• Time required to use• Awareness of alternative sources• Attributes of alternative sources• Purposes of use
“Last Reading” is a variation of the Critical Incident Technique that:
• Permits observation of any combination of:– Sources of articles read– Means of identification– Time spend reading– Age of articles read– Format of articles read– Outcomes and value of reading
• Provides comparison:– Over time– Among disciplines– By age, sex/gender– Other types of users
Examples of Observations Over Time (1977 to 2004)
• Medical faculty read most articles (3 times more than humanities or engineers)
• Personal subscriptions and readings from them continue to go down
• Total amount of reading continues to go up• Readings from libraries continue to go up and are
more valuable to purpose and are more often for research
• Both print and electronic sources are used
Average Reading per Faculty member
186215
197216
175206
219
0
50
100
150
200
250
Num
ber o
f Rea
ding
s
U ofTennessee
U ofPittsburgh
Drexel AllScientists
All-Non-Scientists
All US UNSW
Years of Observation
150172
188216
0
50
100
150
200
250
1977 1984 1993 2000-03
Average Articles Read per year per University Scientist
Ave
rag
e n
um
be
r o
f art
icle
s re
ad
pe
r sc
ien
tist
Year of Studies
Source of Additional Readings
37
113
52
120
9296
101
115
0
20
40
60
80
100
120
1977 1984 1993 2000-03
LibraryCollections
Other Sources
Academic Library Collections Source for Increased Readings
• 66 increased total readings; 64 from library collections
• When identified from searches, citations, etc., articles must be located and obtained
• Libraries the logical choice for faculty and students
Factors Leading to this Phenomenon:
• Number of personal subscriptions decreased (on average from nearly 6 to under 2; university scientists in U.S. from 4.2 to 3.5)
• Number of articles identified by searching increased (3 to about 50 articles per scientist)
• Breadth of journal reading increased, due in part to e-journal collections (13 to 23 journals from which at least one article read annually on average)
Critical (Last) Incident Method Can Show Usefulness and Value of Academic Library
Collections• Saves faculty time (15 min/reading)• Library reading rated higher in importance (5.5
vs. 4.7 in 1-7 scale)• Readers take more time reading library articles
(39 vs. 33 minutes)• Achievers read more and use library collections
more than non-achievers• Articles from libraries yield more favorable
outcomes• Articles from libraries help achieve greater
productivity
33.5%
10.3%
56.3%
1st1st YearYear
28.8%
18.1%
53.2%
Library
Personal
Separate
2-5 Years2-5 Years9.2%
17.5%
73.3%
Over 5Over 5 YearsYears
Older articles are judged more valuable Older articles are judged more valuable & are & are more likely to come from more likely to come from librarieslibraries
What is Expected From You:
• Obtain any necessary Human Subjects Permission (or waivers) from your institution
• Review the questionnaire and make suggestions to fit your specific situation
• Decide whether your survey should be web or paper• Send an email or cover letter to your faculty and
students describing the survey • Post a link to the survey on your website if you wish• Publicize to help response rates• Help us identify your i.p. addresses (broadly) for usage
logs if you can
What Is Expected From Us
• Obtain Human Subjects permission at UT (done)• Design and test the questionnaire• Receive responses at the UT secure server• Analyze results• Present survey results to each library• Compare survey results with deep log analysis
of OhioLink logs (with your name removed) in IMLS reports
• Show how the various user methods can be used together