Barcoding: The Essential Link

Larry N. Osborne

"A chain is only as strong as its weakest link." "For want of a nail..." "Spoil the ship for a hap'or'worth of tar..." The old wives and the speakers in parables are often at pains to instruct us to beware of falling into the trap of overlooking seemingly small things that can cause big disasters.

Librarians have often been guilty of just that, to their chagrin. We have spent time and money selecting (or designing) the best available automated systems, and we have spent untold millions of dollars retrospectively converting our records to machine readable form, only, in many cases, to find that connecting the one to the other is indeed the weak link in the chain.

The link between the collection and the database is usually provided by some sort of code ... in manual systems this is often the call number (class number + book number + copy number), or a special accession number. Such a code is unique, relatively easily understood by users, and connects the item in hand solidly with the database, i.e. the shelflist. While it would be possible, with today's technology, to use the call number as the link to a machine readable database, such a link would be slow, expensive, error prone, and generally not desirable. The most common linking medium in modern systems is the barcode, with the optical character recognition (OCR) label a distant second.1 In effect this is a kind of accession number, one that is machine readable and links books to a computer file rather than to the shelflist.

Unfortunately, like accession numbers, barcodes are simply not very exciting. They don't have the whistles and bells of an integrated system, nor do they have the grandeur of mass presented by a retrospective conversion process; they're simply essential. They are also ultimately the library's responsibility. We can buy a turnkey system, and ship our shelflists out for the attention of experts, but we (or our subordinates) are the ones who get to stick the barcodes on the books.

But just because barcodes lack glamour doesn't mean they aren't important. Choosing the wrong kind, or the wrong time to attach them to the books or bibliographic records can be a costly error, and can delay startup of an otherwise operational system. Such delays can be avoided by the foresighted. Unfortunately the appearance of simplicity has led to a dearth of literature describing the options available, and resulted in most managers depending on anecdotes and maxims. And there are decisions to be made: what form of label to choose, what code to use, when to put them on the book, at what point to link them to the bibliographic records, and even where to put them on the book itself.

In order to understand the confusion surrounding barcodes it is necessary to realize that there is no single standard, and that there are several potentially useful types for libraries. Indeed, there isn't even agreement on whether to use barcodes at all, some libraries opting for OCR coded labels. This multiplicity of options, coupled with the fact that not all systems are capable of functioning optimally with all forms of labels leads to the first factor to be considered in any barcoding project: does the library know which vendor's system it will be adopting, or at very least, does it know the likely contenders.

If the library has selected a vendor, or if it has narrowed the selection to a few possibilities, inquiry into preferred form of barcode is both possible and necessary. If, on the other hand, the library is at the very beginning of the search, it must either decide to use one variety of code (and then make the ability to handle it one of the system requirements in any request for proposals), or it must delay any barcoding project until a system is selected.

Basing the format and code decision on the system's requirement rather than selecting a system which will handle the form of label used makes good sense, unless the delay will mean that the library must pass up an opportunity to start out with a converted, labeled, and linked collection. If the vendor has not yet been selected, and there is good reason to barcode before one is chosen, some factor other than vendor preference must perforce be used as a basis for choosing a code type. The likely possibilities boil down to four: OCR, CODABAR, Code 39, or code 2 of 5 labels, although some libraries have also used the universal product code (UPC).

The first, optical character recognition labels, have the undoubted advantages of being easy for humans to interpret, simple to produce in-house, and often physically smaller. Unfortunately, some major corporations which have adopted this technology have found that the scanning speed is slow and the error rate high. One need only compare observations made in a supermarket using barcodes to those in a store where many of the clerks routinely key in stock numbers rather than accept the poor performance of the OCR readers.2

Barcodes are a faster and more reliable (although not absolutely accurate.3) technology, indeed they have been described as "...the most accurate, least expensive way to get data into a computer, period",4 but come in a bewildering variety of formats. The most commonly used in libraries is the CODABAR format, which is also known as the CLSI standard after its largest library user or the Monarch standard after the Pitney Bowes subsidiary which developed it. Like many de facto standards which developed along with computers it has several variations and some severe limitations including difficulty in printing, a limited character set (primarily numbers and punctuation), and occasional difficulty in scanning. In addition to the data, stop and start characters are added to the ends, and a check digit is incorporated to decrease scan errors.5

The three of nine code, or code 39, overcomes some of these limitations. It is an alphanumeric code (not particularly important in most library applications), has superior internal error checking because of its invarying structure of nine elements (bars and spaces) per character, three of which must be thick and six of which must be thin (and thus is able to function without a separate check digit), and can more easily be produced locally using standard dot matrix printers.6 It appears that because of its advantages, code 39 labels are the label of choice for libraries which have not yet started to barcode their collection. The ability of code 39 to encode alphabetic characters is more a matter of interest than a practical consideration for libraries, because the software of few if any library vendors can interpret such data.

The 2 of 5 code is of interest only to those who use Follett’s Circulation Plus and Catalog Plus software. As a pioneer in the small-computer automation field, Follett selected 2 of 5 code before an industry-wide standard became obvious. Unfortunately it guessed wrong. Newer versions of the Follett packages will eventually accept longer 3 of 9 codes. Thus 2 of 5 is not a viable choice for new library barcoding projects.

Other forms of machine readable media include the magnetic strip (familiar from bank and credit cards), which is seldom if ever used for book identification, but has been used by some systems for patron ID cards. The other bar code formats: the Universal Product Code, the two of five, the two of seven, code 11, etc. are not commonly used in libraries.7 Some libraries do use the UPCs which are pre-printed on periodicals and paperbacks, especially for circulation purposes when a full record for the item is not to be entered into the database.

Having selected a format, libraries must next decide when to barcode. The question of when to actually put the barcode onto the item is not one which can be answered with a pronouncement suitable for all situations, but three basic alternatives exist. First, in most small to medium sized libraries (up to say a quarter of a million volumes) any retrospective conversion project should probably begin with or include an inventory of the collection. Obviously, if each item must be handled anyway to verify its presence, one way to indicate that a book has been inventoried is to place a barcode label on the book. Sheets of barcodes are available which have duplicate barcodes side-by-side in columns. One of these may be used on the book, and the other placed on the shelflist card next to the accession, copy, or volume number. The advantage of this is that the barcode can be added to the record when the shelflist is converted to machine readable form, resulting in a collection already linked to the database. On the other hand the bulk of the shelflist is increased by the thickness of the barcodes, and the cost of the barcodes is higher (since two are required per item). For medium or smaller libraries these disadvantages are overcome by the advantages of one-handling inventory and barcoding, and production of a pre-linked database.

The second possibility is that retrospective conversion is complete, but barcoding has not been done. In such cases a library can either (1) go ahead and inventory the collection and label the items, (2) label but don't inventory them, or (3) label items only at the next time they must be physically handled, usually upon circulation. Labeling during circulation (normally upon return) has the advantage of being client-friendly (since those items which circulate first, and hence generally more often, are barcoded first) but it has the dual disadvantages of never establishing true holdings (since books which do not circulate are never linked to the database, making them indistinguishable from items which are lost or stolen), and of requiring additional time for applying and linking the code to the record. Applying the barcodes after retrospective conversion is complete requires an additional linking step in which the item record is retrieved and the barcode number is entered, but must eventually be done if a true online shelflist is to be established. Neither the possibility of yet another project, nor that of putting off the mass barcoding until (or if ever) inventory time comes around again is attractive, and libraries are well advised to barcode and inventory before the retrospective conversion itself is done, if they possibly can.

Obviously the third possibility is that the library is in the middle of a retrospective conversion and/or inventory project. Since it is generally preferable to label and inventory simultaneously, libraries may wish to begin labeling all books at the point of inventorying they have reached (with the result of an incomplete online shelflist), to go back and bring the labeling and the inventory to the same point (at increased time and cost), or to not label at all during the inventory (and accept the problems discussed above).

Unfortunately, even if the library is at the optimal position for labeling (at the beginning of the inventory), it still may not wish to label the books if it has not selected a system vendor. As noted above, while most vendors' software can operate with most code formats, use of the format for which the system was designed will generally result in optimal operation. Of course, if the library is at the beginning of a long inventory and retrospective conversion project, with no system in sight for years, management may feel that the benefits of having a database with unique item identifiers linked (via barcodes) to the physical items overrides the potential for incorrect code selection. Eventually the library may change systems anyway, and the barcodes may be optimal for the second system, even if they weren't for the first. In any case, should the library decide to begin to label books before it selects a system, it should choose one of the two commonly used formats, CODABAR or code 39.

Of course converting the bibliographic record and labeling the item are only the end links in the chain. Somehow a specific item must be linked to a specific record. There are several ways to do this. If the library has labeled its items and shelflist with duplicate barcodes (cleverly remembering to indicate copy or volume numbers next to the appropriate label where necessary), and this has been done before retrospective conversion, then it may simply add the number as another item in the record as it (or its conversion contractor) enters the record.

Libraries which find themselves in the position of having a database without barcodes as part of the records have no easy way out. While it would theoretically be possible to take each book to a terminal and (calling up the appropriate record by ISBN, author/title, or some other key) link the two by adding the barcode number to the record, such an approach would be outrageously expensive. Most libraries in this difficult situation will be forced to resign themselves to linking at the point of circulation.

Several other techniques have been used to link already converted records to books. For example, if the collection has already been converted, a list of items from the database can usually be produced in shelf-list order. This can then be annotated with the barcodes by taking the list to the stacks.

Another option, if the items have not yet been barcoded, is to produce a consecutively numbered printed shelflist, order corresponding barcodes from a barcode supplier, and, using the printed list as a guide, apply the appropriate barcode to the corresponding book. Retrospective conversion vendors usually prefer this approach, since it moves the responsibility of accurate linking from them (if they have to key in pre-linked barcodes) to the library (which has to make sure the correct barcode goes on each book). As an aid to matching the barcode and the book, retrospective conversion vendors usually print identifying data, for example call number, author, and title, on the label itself. Such labels are sometimes called “smart” barcodes 8, These, and similar, approaches, are de facto inventories, and should be viewed in that light when cost/benefit analysis is performed. They may be necessary, but they are definitely not desirable, since they increase the amount of precision work necessary to produce a clean and correct database, and hence the likelihood for errors.

Two other variables in any labeling project bear discussing: what goes on the label, and where it goes.

Barcode labels can be printed with as much information as the library desires: just the code, with the code and its translation into human readable digits, with this plus call number, and author, and title, and page numbers, and publisher, and (if the library is willing to accept a big enough label) with all the information normally found on a catalog card. As noted above, labels with more than just human readable numbers and the barcode itself are sometimes called "smart" barcodes, while ones with just the code and its plain text equivalent are referred to as "dumb".

Just because it's possible to produce such a label does not, however, mean its worthwhile. A barcode's sole function is to serve as a link between the physical item and the item record in the bibliographic database. If our equipment was 100 per cent reliable all that would be necessary to have on a label would be the code itself. Unfortunately, our equipment does sometimes fail, and then humans must keep track of the data by hand. Thus the optimal (and really the only rational) content for a label is the code itself and its numeric equivalent.

Fancier labels are not worthwhile because they don't significantly increase the amount of useful information on the label (since the use of the code is to link the record and item) while significantly increasing the amount of collection maintenance. Labels, while they are relatively durable, are subject to wear and vandalism. If a label must be replaced and it consists only of a number, then it is a simple task to take the next available label, replace the number in the record with the number on the new label, and circulate the book. If, on the other hand, the book's call number is also on the (missing) label then a new label will have to be custom produced. For a code 39 library which produces labels in-house this could be a fairly rapid operation (although not the sort of thing you would want to do while the patron waits). If, however, the library must order labels from a specialty printing house the wait can be days or weeks (or longer if the library waits to collect an order of "worth while" size. Such inconvenience is seldom offset by the redundancy which comes from having the call number two places on the book (three, if technical services also writes it in the book as part of the cataloging process).

The final consideration in barcoding items is where the label should be placed on the item. Barcode labels have been affixed to many parts of a book, but in any given library having a single predictable location, even if it proves less than optimal, is better than having many. Chose a location thoughtfully and stick to it.

If, as part of the cataloging process, book jackets are removed, the task of deciding where to put the label is easier, since at least the library need not consider putting the label on the dust jacket, but for most public and many academic libraries this is not the case. While it is obvious that the dust jacket is not a durable location, placing the label there does allow the item to be circulated with minimal handling, and if "dumb" labels are used, they can easily be replaced if the cover is removed.

More common places for labels are on the title page or end-papers (which necessitates opening each book in order to check it out, not necessarily a disadvantage if a due date is inserted or stamped...if the barcode is in the same place as the due date goes), and on the front or back cover (which is prone to more wear, but simplifies inventory, since books need to be removed from the shelves if a portable barcode reader is available, merely tilted back).9 Libraries which chose this last option may wish to cover the label with clear tape to reduce wear, especially if less durable locally-produced dot matrix labels are used. Of course the library should first make sure that the barcode readers will scan through the tape, although the author has not found this to be a problem. Scanning difficulties are more common with light pens than laser readers. Hand-held laser guns are becoming increasingly available.10

Few things in automation are simple or straightforward. Even in a task as seemingly trivial as barcoding a collection, several factors must be considered: format of the code, data content of the label, point of labeling and point of linkage, and position of the label on the item. None of these are particularly difficult in themselves, but remembering to consider them can make the difference between a successful and satisfying system and a painful disaster.

Larry Osborne, is Associate Professor of Library and Information Studies at the University of Hawaii at Manoa in Honolulu. His research interests include both marketing information storage and retrieval systems, and expert systems for information storage and retrieval. Email: osborne@uhunix.bitnet Mailing address: School of Library and Information Studies, Hamilton Library, 2550 The Mall, Honolulu, Hawaii 96822 USA.


1. Rahn, Erwin. "Bar Codes for Libraries." Library HiTech. 6;2,2(October 1984)73.

2. Personal observation and discussion at local Sears and J.C. Penney stores in the Honolulu area. readers are encouraged to confirm this observation for themselves, although this will not be possible locally, since both companies have abandoned OCR technology in favor of barcodes.

3. Wray, Bruce R. "Bar Codes: Basic Principles." Computers and People. 34,1-2(January-February 1985)12.

4. Berenstein, Amy. "Bar Codes Earn Their Stripes". Business Computer Systems. 4,2(February 1985)68.

5. Solomon, Leslie. "Once Codes Earn Their Stripes". Business Computer Systems. 22,4(April 1984)70.

6. Rahn, po cit.

7. Evans, Peter W. "Barcodes, Readers and Printers for Library Applications." Program. 17,3(July 1983)161.

8. Retrospective Conversion: Guidelines for Libraries. Dublin, Ohio: OCLC, 1987.

9. deKlerk, Ann. "Barcoding a Collection - Why, When and How." Library Resources & Technical Services. 25,1(July/March 1981)84.

10. Sanger, Elizabeth "No Magic Wands." Barron's. 63,14 (4 April 1983)15.