College and Research Libraries


A Least Cost Searching Sequence 

BY GERALD J. LAZORICK and THOMAS L. MINDER 

IN EVALUATING the procedures of the ac-
quisitions department of the Pennsylvania 
State University library system, the cost 
of searching information to order a book 
was found to be $.73. In addition, an or-
der form had a rather high in-process 
time. An analysis of the flow of order 
forms through the department showed 
that a bottleneck was occurring with the 
searching procedure. As a result of this 
many order forms were being marked 
"rush," and routine items were faced 
with an even higher in-process time. The 
library was faced with adding searchers 
or with accepting the fact that there 
would almost always be a backlog of un-
processed orders. The latter would result 
in a high in-process flow time thus reduc-
ing service to patrons and requiring that 
needed course material be ordered well 
in advance of the time when it was ac-
tually needed. 

The technical services librarian, the 
catalog librarian, an industrial engineer, 
and the librarian in charge of library sys-
tems research met and compared the 
amount of searching needed to identify 
adequately a publtcation to be purchased 
with the amount of searching needed for 
complete bibliographic information. At 
the time, it was decided that the library 
would adopt for a trial period an "ade-
quate information" philosophy. The pos-
sibility that duplicate copies would oc-
casionally be received was considered to-
gether with the increased probability of 
obtaining the wrong books. However, the 
fact that searching time would be con-
siderably reduced thus increasing the 
number of possible searches in a given 
period of time was assumed to outweigh 

Mr. Lazorick is Industrial Engineer, and 
Mr. Minder is Engineering Librarian, in 
Pennsylvania State University. 

the disadvantages in accepting the philos-
ophy. 

The definition of adequate information 
was also discussed. It was decided that 
adequate information would include the 
author, title, publisher, and date of publi-
cation. It was also assumed that the latest 
edition would always be ordered unless 
otherwise specified. 

With this in mind, a study of the 
searching procedure was made by the in-
dustrial engineer. Out-of-print items and 
foreign language items were not consid-
ered in the study. Out-of-print items re-
quired special processing by acquisitions 
personnel, and foreign items required 
searching in a large variety of bibliog-
raphies. In addition, the bulk of the books 
ordered by the acquisitions department 
were in the English language. 

The searching sequence was considered 
to be an area for study. What was the 
optimum searching sequence? Realizing 
that one bibliographic source would con-
tain more information than another and 
that the time for searching varied con-
siderably with the source used, a method 
was needed to determine in what se-
quence bibliographic information should 
be sought. In other words, should a 
searcher first look in the LC catalog, and 
if adequate information was not found 
then continue the search in Books In 
Print, and again if not found, continue in 
the Cumulative Book Index, etc., or 
should some other sequence be used? 

126 COLLEGE AND RESEARCH LIBRARIES 



It was decided that Books In Print, 
Publishers' Trade List Annual, LC proof 
slips, Cumulative Book Index, and the 
LC catalog would be studied as the sourc-
es of bibliographic information. The two 
variables which had to be measured were 
the per cent of time in which adequate 
information was found in each of the 
sources and the average time to complete 
a search in each of the sources. The first 
of these, the percentage, is really the 
probability of finding adequate informa-
tion in each of the sources. 

THE STUDY 

A random sample of twenty-five Eng-
lish language order slips were selected. A 
searcher was asked to take the twenty-
five slips and check each of the fi~e so_urc-
es to see if the items could be Identified 
with adequate information. The percent-
age of time adequate information was 
found was recorded for each bibliographic 
source. Another searcher was then given 
five random samples of ten order forms 
and was asked to search the samples in 
the five bibliographic sources while being 
timed separatedy for each source. The 
outstanding order file and the public cata-
log were not studied, since previous stud-
ies indicated that about 20 per cent of 
the orders ~eceived in acquisitions were 
for books on order or already in the col-
lection. These sources would always be 
searched first. Books In Print and Pub-
lishers Trade List Annual were both in-
cluded for control and comparison pur-
poses. The results of the study are sum-
marized in Table 1. 

The second column in Table 1 shows 
the average time of a search in minutes. 
The third column shows the probability 
of finding the item searched. It should be 
noted that the probability of not finding 
adequate information after searching all 
five sources is .03. That is, we can expect 
that 3 per cent of the books searched will 
not be identified if we search only the five 
sources. 

MARCH 1964 

TABLE 1 

SEARCHING SUMMARY 

Time Probability of 
to a Successful 

Source Search Search 

Library of Congress Catalog 3.24 .72 
Cumulative Book Index . . 2.14 .60 
Books In Print . . . . .56 .40 
Publishers' Trade List Annual 1.09 .36 
Library of Congress Proofs . .94 .28 

There are two commonly accepted 
philosophies in this type of sequencing 
situation. One person would argue that 
we should put the least time-consuming 
source first in the sequence, the second 
time-consuming source second, and the 
most time-consuming source last; whereas 
others would argue that the source with 
the best chance of a successful search 
should be put first, and the source with 
the least chance of a successful search 
last. However, neither of these philoso-
phies will necessarily result in the opti-
mum sequence, that is, the least time-
consuming sequence. One could argue 
that all combinations should be evalu-
ated. However, there would be 51 or 120 
possible combinations. Fortunately, op-
erations research and industrial engineer-
ing have solved this sequencing problem 
for an analogous industrial application. 
The optimum sequence in this instance 
can be determined by taking the ratio of 
the time consumed to the per cent of suc-
cess and ordering the tests such that the 
ratios will be in increasing sequence. The 
ratio for each of the bibliographic sources 
is shown in Table 2. 

TABLE 2 

SEARCHING SEQUENCE RATIOS 

Source 

Library of Congress Catalog 
Cumulative Book Index . 
Books In Print . . . . 
Publishers' Trade List Annual 
Library of Congress Proofs . 

Ratio 

4.50 
3.56 
1.40 
3.03 
3.36 

The optimum searching sequence to-
gether with the "least time first" sequence 

127 



TABLE 3 
PossiBLE AND OPTIMUM SEQUENCES 

Least Most 
Time First Success First Optimum 

BIP LC Catalog BIP 
LC Proof . CBI PTLA 
PTLA . BIP LC Proofs 
CBI PTLA. CBI 
LC Catalog LC Proof . LC Catalog 

and the "most success first" sequence are 
shown in Table 3. 

Based on this study, it was pointed out 
that adoption of the optimum searching 
sequence would result in 27 per cent 
more searches per year of English lan-
guage publications. Certain exceptions to 
the procedure should be followed based 
on certain idiosyncrasies of the sources. 
For example, the publications listed by an 
obscure scientific society would rarely be 
listed in Books In Print. Therefore, one 
would be wasting his time searching in 
Books In Print. 

Special considerations were also given 
to the LC proof file, a rather recent ad-
dition to the Pennsylvania State acquisi-
tions department. The other sources are 
rather stable bibliographic sources which 
are not likely to change in the searching 
sequence in the years to come. Since the 
use of the LC proof file is still in its in-
fancy at Pennsylvania State, it is expect-
ed to grow in size and usefulness. It was 
therefore recommended that the LC proof 
file be re-evaluated annually as a search-
ing tool. . 

One weakness of the study should be 
noted. In order to make the times and 
percentages more accurate, a larger sam-
ple would have to have been taken. This 
was not done because neither the indus-
trial engineer nor the searchers had the 
time to gather as much information as 
would have been necessary. Further data 
is being collected, and the searching pro-

cedure will continuously be re-evaluated. 
The least-cost searching sequence 

method is derived from a methodology 
described by L. G. Mitten. 1 This method 
also has promise in other areas of library 
work. For example, the reference staff 
might establish a card file of searching 
sequences for certain general types of 
questions. The searching sequence pre-
ferred would be determined by the prob-
ability of searching success in specific 
sources and the searching time through 
these sources. The sequence might also 
be determined by the other variables such 
as the level of competence of the request-
er or the librarian. 

To carry this one application a step 
further, one would expect its most fruitful 
area of use in the middle ground between 
the very special narrow topic and the 
extremely broad "encyclopedia" topic. 
These two extremes usually direct the 
searcher to the obvious sources. There is 
usually not a choice of six or eight differ-
ent sources and sequences for this type of 
search. 

Another possible application lies in the 
area of "state of the art" searches. The 
questions, "What is the probability of 
finding useful information in source X 
since date Y?" and "What is the expected 
redundancy between indexes V and W?" 
are tied in very closely with the economics 
of literature searching versus laboratory 
experimentation. 

Least-c·ost searching sequence method-
ology might . profitably be applied to the 
oft-stated proposition that it is cheaper to 
re-invent than to search the literature. It 
might also reduce the cost of searching. 

•• 
1 L. G. Mitten, ·"An Analytic Solution to the Least 

Cost Testing Sequence Problem," Journal of Industrial 
Engineering, XI (January-February 1960), 17. 

128 COLLEGE AND RESEARCH LIBRARIES 

I