How Clean Is Our Catalogue? A Quality Audit of 400 Retro-Converted MARC 21 Records in a College Koha Catalogue

Many Indian college libraries have recently moved to Koha and "retro-converted" their old catalogue: staff, trainees or outsourced data-entry operators keyed years of accession-register and card-catalogue entries into MARC 21 records in a short time. The result is a catalogue that works, but nobody knows how accurate it is. This B.Lib.I.Sc project audits that accuracy.

The study is set in Vidyanagar Arts and Science College Library (fictional), whose retro-conversion into Koha was finished last year. A random sample of about 400 bibliographic records is drawn from the catalogue and each record is checked against the book in hand (or its title page image) and against the rules: AACR2 (and RDA where the library has moved to it) for description and access points, MARC 21 for field and indicator use, DDC 23 for the class number, and the Sears List for subject headings. Duplicate records are searched separately.

Every error is coded into a simple error taxonomy — access points, title and statement of responsibility (245), edition (250), publication details (260/264), physical description (300), classification (082), subject headings (650), item data, and duplicates — and tallied in Excel. The report presents error rates by field and by type, identifies likely root causes by interviewing the staff who did the conversion, and ends with a practical quality-control checklist the library can use for all new cataloguing.

Abstract

Retro-conversion turns a library's legacy catalogue into machine-readable records. Done quickly and without review, it spreads small errors — a title in the wrong subfield, a missing author, a wrong class number — through thousands of records, and these errors silently hide books from users of the OPAC. This project conducts a sample-based quality audit of a college library's retro-converted Koha catalogue, measures error rates by field and type, investigates causes, and proposes a quality-control checklist.

Introduction

A catalogue is only as good as its records. Ranganathan's "every book its reader" depends on each book being findable by author, title and subject. In automated systems, where searches depend on specific MARC fields and indexes, a mistake in the wrong field can make a book practically invisible. Yet most small libraries never audit their records after migration.

Review of literature (gap)

LIS literature on retro-conversion in India focuses mainly on methods (in-house keying versus outsourcing versus copy cataloguing) and on software. Studies that measure record quality tend to address large university catalogues or union catalogues. Small affiliated college libraries, where conversion is often done by temporary staff, are rarely studied. This project provides a simple, repeatable audit method suited to such libraries.

Existing practice vs proposed practice

• Existing: records entered from registers with minimal review; no written local cataloguing policy; no routine check of new records; duplicates noticed only by chance.
• Proposed: periodic sample audits with a fixed error taxonomy; a short local cataloguing policy; a quality-control checklist applied to every new record; a monthly duplicate check using Koha reports.

Feasibility

• Technical: the audit needs only Koha's staff interface, MarcEdit for viewing records in bulk, and Excel.
• Operational: checking about 400 records takes roughly three to four weeks for one student working a few hours a day.
• Ethical: no personal user data is involved; staff interviews are anonymous and focus on processes, not individuals.

Vidyanagar Arts and Science College Library completed the retro-conversion of its catalogue into Koha in a short period using temporary staff and trainees working from accession registers and old catalogue cards. Since then, library staff and students have noticed books that exist on the shelf but cannot be found in the OPAC, the same title appearing as several records, and class numbers that place books in the wrong section. No one knows how widespread these problems are or which kinds of errors are most common.

The problem this project addresses is: what is the level and pattern of errors in the library's retro-converted MARC 21 records, what causes them, and what practical quality-control measures can prevent them in future? The answer must be based on a systematic sample rather than anecdotes, so that the library can decide where to spend limited correction effort.

In scope

• Books (monographs) catalogued during the retro-conversion; one library.
• Bibliographic fields 020, 082, 100/110, 245, 250, 260/264, 300, 490, 650 and the item field 952.
• Duplicate record detection within the catalogue.
• Staff interviews about the conversion process.

Out of scope

• Serials, theses, non-book material and e-resources.
• Authority control files (noted as a recommendation only).
• Correcting the entire catalogue; only the sample is corrected, and the method is handed over.

The study uses a descriptive, sample-based audit design with a small qualitative component (staff interviews).

Error measures: percentage of records with at least one error; mean errors per record; error rate for each field (records with an error in that field ÷ records where the field applies); share of errors that affect retrieval. Sample size logic: about 400 records gives a margin of error of roughly five percentage points at 95% confidence for a large catalogue — explain this in the report with the standard formula.