Beguiled by Bananas : A retrospective study of usage & breadth of patron vs. librarian

acquired ebook collections

Jason Price & John McDonaldLibraries, Claremont University

ConsortiumNovember 5, 2009

(with data & discussion from Kari Paulson & Alison Morin of EBL)

Bananas tipped the boat:

a cautionary tale• Early patron-driven deal with a

major platform• Assignment on economics of banana

plantations• UC Boulder ‘bought every book

with banana in the title’ • Used by librarians & vendors(!) as

evidence that user-driven selection is a bad idea

Patron-driven model objections:

straight off the boat• Books will be selected based on click-

thrus that don’t indicate interest• Users will select ebooks that no one

(else) is interested in• User selected collections will be

unbalanced turkeys

DefinitionsPurchase type• Patron selected = Demand Driven = User-

selected• Librarian selected ≈ Library selected ≈

Pre-selectedEbook usage measured conservatively• Use data gathered post-purchase

– Did not count uses that lead to user-selection• 1 use ≈ 1 ‘read online’ ≈ 1 ‘download’

– read online = >10 min w/click thru OR copy OR print

– download = to adobe Digital Editions for multiple days

• Transaction level data – each use recorded separately with user anonymously identified

Questions we’ll address1. Are user-selected ebooks used

less than pre-selected ebooks? 2. Do user-selected ebooks have a

narrower audience?3. Are user-selected collections

less balanced?Do we have anything to fear in

patron-initiated selection?(Can we use this to build better

acquisition models?)

Overall Scope of the dataset• 1 Ebook Vendor – EBL (Ebook

Library)• 11 Libraries• 28,322 ebooks bought from 2006

- 2009• 212,887 uses• Purchase Models: User Selected,

Pre-Selected, or Mixed

Total Books & UsageLibrary Model Library

PurchasedPatron

PurchasedUsage by Download

Usage Read Online

A MIX 552 1131 6773 9888B MIX 2612 5246 42880 38329

C USER 102 2198 0 11801D USER 48 3010 697 15126E MIX 909 4159 17396 25604F PRE 1451 0 4905 3082G PRE 2154 31 7001 4459H USER 0 801 556 415I MIX 336 305 3334 2568J USER 53 2799 5 13349

K MIX 276 147 2436 2283 TOTAL   8,496 19,831 85,983 126,904

Total Books & UsageLibrary Model Library

PurchasedPatron

PurchasedUsage by Download

Usage Read Online

A MIX 552 1131 6773 9888B MIX 2612 5246 42880 38329

C USER 102 2198 0 11801D USER 48 3010 697 15126E MIX 909 4159 17396 25604F PRE 1451 0 4905 3082G PRE 2154 31 7001 4459H USER 0 801 556 415I MIX 336 305 3334 2568J USER 53 2799 5 13349

K MIX 276 147 2436 2283 TOTAL   8,496 19,831 85,983 126,904

Total Books & UsageLibrary Model Library

PurchasedPatron

PurchasedUsage by Download

Usage Read Online

A MIX 552 1131 6773 9888B MIX 2612 5246 42880 38329

C USER 102 2198 0 11801D USER 48 3010 697 15126E MIX 909 4159 17396 25604F PRE 1451 0 4905 3082G PRE 2154 31 7001 4459H USER 0 801 556 415I MIX 336 305 3334 2568J USER 53 2799 5 13349

K MIX 276 147 2436 2283 TOTAL   8,496 19,831 85,983 126,904

Total Books & UsageLibrary Model User-

SelectedPre-

SelectedUsage by Download

Usage Read Online

A MIX 1131 552 6773 9888B MIX 5246 2612 42880 38329

C USER 2198 102 0 11801D USER 3010 48 697 15126E MIX 4159 909 17396 25604F PRE 0 1451 4905 3082G PRE 31 2154 7001 4459H USER 801 0 556 415I MIX 305 336 3334 2568J USER 2799 53 5 13349

K MIX 147 276 2436 2283 TOTAL   19,831 8,496 85,983 126,904

Scope of this study• 5 libraries• Books owned more than 6

months

Data• 1 Ebook Vendor• 11 Libraries• Full purchase history• Bibliographic Data• Models: User Selected, Pre-

Selected, or Mixed• Transaction level usage data

Analysis Levels• Usage

• By selection model• By library profile• By unique users• By time since owned• By publisher • By LC class

Library BooksAverage Unique Users

Average Usage

Average Time

Available

A 1683 5.63 9.90 475.28B 7859 5.69 10.09 512.33

C 2300 3.64 5.13 289.86D 3058 4.51 5.19 528.08E 5068 5.00 8.49 440.30F 1451 2.96 5.50 496.38G 2185 2.76 5.24 424.26H 801 0.71 1.17 522.94I 641 4.56 9.21 766.38J 2853 2.66 4.68 327.66

K 423 5.28 11.13 784.81

Usage & Unique Users

Total UsagePurchase Model Library Mean Std. Dev BooksUser-selected A 10.12 13.95 778

B 8.39 19.31 4723E 8.14 15.25 3111I 14.36 42.49 277K 7.09 13.16 147

Total 8.61 18.74 9036Pre-selected A 3.67 5.89 498

B 4.60 11.55 2475

E 4.61 18.55 766I 3.21 8.69 331K 3.08 5.47 217

Total 4.31 12.25 4287Total A 7.60 11.92 1276

B 7.09 17.14 7198E 7.44 16.02 3877I 8.29 29.88 608K 4.70 9.56 364

Total 7.23 17.04 13323

Librarian Acquired

ANOVA of uses per yearSource

Type III Sum of

SquaresMean

Square F Sig.Partial Eta Squared

Observed Powera

Intercept Hypothesis 191873 191872 295.056 .000 .979 1.000

Error 4207 650       

Purchase Model

Hypothesis 35509 35508 26.796 .004 .843 .982

Error 6615 1325       

Library Hypothesis 3984 996 .431 .782 .301 .088

Error 9243 2310       

Purchase * Library

Hypothesis 9243 2310 8.092 .000 .002 .998

Error 3801351 285       

Subject Area Analysis• Pie charts of each discipline by

model (or bar charts• Another thing – is the

collection too skewed towards one LC class or subject areas or do demand-driven selection result in a good collection. Do ratios of each discipline as a proportion of total books bought by model.

• Is publisher content skewed as well?

• What about price/cost?

User-selected collections have similar

subject profiles

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

A B E I K

Sci & Tech

Soc Sci

Arts & Hum

User Pre User Pre User Pre User Pre User Pre

Prop

ortio

n of

col

lecti

on

Library

User-selected collections have similar

LC profiles A B E I K

lc USER PRE USER PRE USER PRE USER PRE USER PRE

H 20% 20% 20% 9% 28% 7% 13% 14% 7% 5%

R 14% 4% 8% 4% 10% 2% 2% 7% 3% 5%

Q 7% 2% 5% 4% 7% 3% 4% 5% 8% 28%

T 5% 5% 4% 3% 7% 2% 6% 3% 2% 7%

L 4% 1% 4% 2% 6% 1% 1% 1% 0% 1%

G 3% 2% 2% 2% 3% 0% 2% 5% 2% 3%

B 2% 1% 5% 2% 4% 1% 3% 6% 4% 2%

User-selected collections have similar

LC profiles

A B E I K0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

HRQTLHRQTL

Library

Prop

ortio

n of

col

lect

ion

Blue = User selectedGreen = Pre selected

User-selected collections have fewer

unused titles

USER PRE USER PRE USER PRE USER PRE USER PREA B E I K

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

Library and Purchase (Selection) Type

% o

f Unu

sed

book

sbo

ught