College and Research Libraries JAY B. CLARK An Approach to Collection Inventory All large libraries eventually face the problem of determining catalog accuracy. This paper recommends taking pilot samples of the collec- tion to be followed by a selective inventory. The procedures and task times of the recent inventory of the Houston Public Library are in- cluded. LARGE ACADEMIC AND PUBLIC LffiRARIES considering an inventory should be in- terested in the planning and results of the inventory of the Main Library of the Houston Public Library-the deci- sion to inventory, the procedure, the time estimates, and the results of the in- ventory. The library had taken its last com- plete inventory in 1924 and partial in- ventories in 1934 and 1943. But by 1968 staff and user complaints about the ac- curacy of the public card catalog led the library to consider the possibility of tak- ing the first complete inventory in fifty years. PILOT INVENTORY The first step was to define book losses more accurately by checking a random sample of the shelflist against the shelves. The collection to be inventoried consisted of 236,519 titles held in 357,350 volumes. For the sample, a loss rate of 30 percent was assumed. Sam- pling tables indicated 600 titles needed to be pulled at random from the shelf- list.1 The sample cards were duplicated and sent to subject departments where their staff compared the shelflist cards with the titles on the shelves, marking Mr. Clark is chief, Technical Services, Houston Public Library, Houston, Texas. 350 I the accession numbers located. Cards not matched with books initially were periodically checked again over five months. The results of the survey indi- cated that 41 percent ± 3.5 percent of the volumes and approximately 35 per- cent of the titles in the shelflist for the Main Library were missing. While there was a backlog of book cards for vol- umes reported lost that had not been withdrawn from the catalog, the back- log could not explain the losses indicat- ed by the sample inventory. The library decided to inventory the Main Library. PROCEDURE The first step was to divide the collec- tions into manageable sections for in- ventory and prepare a procedure. The chapter by R. E. Beck and J. R. McKin- non on the "Development of Methods and Time Standards for a Large Scale Library Inventory" was used in design- ing a procedure and estimating time. 2 The inventory of each collection di- vided itself into three parts: ( 1) the physical inventory of the volumes on the shelf; ( 2) post-inventory follow-up to check volumes in circulation during inventory; and ( 3) withdrawing the ti- tles from the shelflist and pulling card sets from the public catalog. In the physical inventory each book was com- pared with its shelflist card, checking An Approach to Collection Inventory f 351 the accession number. One member of the team read the call number from the book as well as title and accession num- ber. Another staff member, holding the shelflist tray, located the correct sheH- list card and wrote the year in red be- side the accession numbers located. The staff member holding the book marked the book pocket and sheHlist accession number with a date ( '70) in red. When a shelflist could not be found for a ti- tle, the staff prepared a duplicate shelf- list on a three-by-five-inch slip. Later, the cataloging staff either found or re- created the shelflist and other necessary r ecords. Our first estimate for the time to do the shelf inventory was fifty vol- umes per man hour. The first inventory indicated that forty-two volumes per man hour was more realistic. The inventory follow-up lasted for 120 days after the inventory. The staff checked volumes returning from circu- lation for evidence of inventory. For each returning book that missed the shelf inventory, the staff marked the book pocket with the inventory date and wrote a three-by-five-inch slip including author, title, accession number, and call number. This slip was used later to add the inventory date to the shelflist. The last phase of the inventory was divided into three steps. The first was marking «withdrawn" on the shelflist card after the accession numbers not found in inventory, pulling the shelf- list card if dead for the system, and turning up in the tray those cards dead at Main Library only. This was done at the rate of 300 per hour. These cards were revised by a cataloger at the rate of 900 cards per hour, and finally copy slips were typed for the standing shelf- list cards to be used to pull cards from the public catalog. The second step in the last phase of the inventory was to pull catalog cards from the official catalog, main public catalog, and departmental catalog. Staff members from technical service and public service joined in pulling these cards. This was done at a rate of forty cards per hour. Finally, the cards pulled were checked for complete card sets at a rate of 1,600 cards per hour, retrieving any that were overlooked and refiling cards pulled by mistake. One unit card for each dead title was then sent to the departmental librarian to consider for replacement. ACTUAL INVENTORY The Business Technology collection was selected as the first room for inven- tory. It was relatively small ( 147 sec- tions), and the inventory sample indi- cated the room had a high loss rate. A map of the room was prepared that showed shelving, the call numbers for each range, and the number of sections per range. At a staff meeting before the inventory, procedures were explained and staff members were given copies of the procedures, a room map, and their team assignments. Teams of two staff members were assigned specific portions of the sheHlist. Responsibility for the teams was divided between two super- visors responsible for the teams in their respective areas. As each team completed its first assignment, a supervisor reas- signed the team another portion of the shelflist. The supervisor tried to reas- sign a team so that it would not run into another ·team in the narrow aisles. In this inventory of the Business Technology collection almost 12,000 vol- umes were found to be missing, repre- senting 21 percent of the collection's ti- tles. Two hundred volumes had errors in either call number or accession num- ber, and fifty volumes belonging to oth- er units were removed. Some problems with the inventory procedure were iden- tified. Too many teams had been as- signed to a supervisor. In following in- ventories as many as four area super- visors were used. Other changes includ- 352 I College & Research Libraries • September 1974 ed showing team leaders the ranges that they were responsible for prior to the inventory and marking inventoried shelves with white paper slips instead of turning books down. The average time for actually inventorying the books was .7 minute per volume. Following completion of the shelf in- ventory, volumes returning from circu- lation were checked as they came back to the library. At the end of the 120-day period, a check was made to estimate the error rate of the first inventory. A ran- dom group of 622 shelflist cards of the Business Technology Room was pulled and compared with books on the shelf. Nine volumes in the sample had not been inventoried, yielding an estimated error of 1.45 percent ± 3 percent. To confirm this, the shelves were checked a second time for the 5,200 titles that were indicated missing and only 36 titles ( .05 percent) were located. Inventory reliability was judged sufficient, and the second search was dropped for inven- tories of other collections. Over the next three years four other collections were inventoried. By the fall of 1972 shelf inventories were com- plete, but there remained a large num- ber of cards not yet pulled from the catalogs. To complete this task each technical service unit was assigned fif- teen hours per week for card pulling, and public service units later also made commitments. The inventory of the Main Library was completed in July 1973. REsULTS AND DISCUSSION What did the inventory accomplish? Removed from the public catalog were cards for 47,514 missing titles, represent- ing 20 percent of the titles in the Main Library catalog. Before the inventory two out of every five cards referred the user to books that were · not available. The ratio was probably higher since a large percent of requests are for materi- al published in the last five years, and a large portion of the lost titles fell in this category. The losses by collection are listed in Table 1. The catalogs of the inventoried rooms prior to the proj- ect contained 236,519 titles, and the holdings were 357,350 volumes. The inventory was done without the employment of a special inventory team and, consequently, at the expense of current operations. A better approach would have been one in which a full- time team of three were responsible for a selective inventory. The pilot inven- tory overestimated volume losses by 10 percent and titles by 15 percent. The error appeared to be the result of using many individuals unfamiliar with the shelflist to take the sample. However, the later survey in Business Technology indicated that much more precise results could be obtained by assigning it to a single staff member who was familiar with the project and shelflist. TABLE 1 INVENTORY RESULTS AT HousToN PUBLIC LIBRARY- MAIN LmRARY Missing Missing Titles Volumes Business Technology Room 5,273 11,918 Fine Arts Room 3,627 10,502 Children's Room 5,222 10,629 Literature & Biography 11,933 38,122 Remainder 000-500s 21,459 40,030 Total 47,514 20 percent of 111,201 236,519 titles missing 357,350 · 31.1 percent of volumes missing [ An Approach to Collection Inventory I 353 With random sampling, a library that is uncertain of the accuracy of its pub- lic catalog can with a small expenditure of staff time obtain a reliable estimate of its catalog accuracy. Such a survey is certainly justified, considering the cost of staff time in inventory. The error rate that justifies an inventory is a mat- ter of judgment, with libraries suggest- ing rates that vary from 1 percent to 5 percent.3 However, the interruption of normal processing or the cost of ad- ditional staff to inventory a large collec- tion, I believe, makes a loss rate of 10 percent a more realistic guideline. A pilot inventory can be used to identify those collections that have an unaccepta- ble discrepancy between the titles that the catalog shows and those actually available. Using the pilot inventory as a guide to the collections in need of in- ventory, libraries will :find selective in- ventories a more effective approach than general inventories. REFERENCES 1. R. Gene Brown and Lawrence L. Vance, Sampling Tables for Estimating Error Rates or Other Proportions (Berkeley: Public Ac- counting Research Project, Institute of Busi- ness and Economic Research, University of California, 1961 ). 2. Barton R. Burkhalter, ed., Case Studies in Systems Analysis in a University Library (Metuchen, N.J.: Scarecrow, 1968 ) , p.48- 75. 3. Irene A. Braden, "'Pilot Inventory of Library Holdings," ALA BuUetin 62:1129-31 (Oct. 1968); Pamela Bluh, "A Study of an Inven- tory," Library Resources & Technical Ser- vices 13:367-71 (Summer 1969) .