Shelf Sleepers: A Two-Year Circulation and Collection-Age Study with a CREW-Based Weeding Plan

Shelf Sleepers is a B.Lib.I.Sc project that answers a question every librarian quietly carries: how much of our collection actually circulates — and what should we do about the rest?

The study is set in Sudarshan Degree College Library (fictional), a 28,000-volume college library where shelf space has run out and new titles wait in cartons. Using two full academic years of circulation data (2023–24 and 2024–25) from Koha and the library's catalogue export, the project builds a title-level usage dataset: issues per title, zero-use titles, and the DDC class, publication year and accession date of every record. Borrower identities are removed at extraction — only transaction counts are analysed.

The statistical core is deliberately honest: a chi-square test of whether use differs across DDC classes, a Mann–Whitney U test comparing the age of used versus unused titles, and a Spearman correlation between title age and circulation — plus the simple, hard-hitting concentration figure (what share of issues goes to the most-used 20% of titles). Zero-use titles are then screened through the CREW method's MUSTIE criteria (Misleading, Ugly, Superseded, Trivial, Elsewhere) into a candidate weeding list, and a shelf verification sample of 60 zero-use titles checks whether they are really on the shelf at all.

The deliverable is a defensible weeding plan, an estimate of shelf space reclaimed, and a one-page collection-policy statement the library committee can adopt.

Sudarshan Degree College Library (fictional) holds roughly 28,000 printed volumes in a building sized for fewer. New syllabus-recommended titles wait un-accessioned because there is no shelf room, while older editions, superseded guides and never-borrowed gifts occupy the stacks. The library has never measured circulation at title level: stock verification counts presence, not use, and committee discussions about "clearing space" stall because no one can say which titles are genuinely dead and which are simply new or held in reserve.

The problem is the absence of an evidence base for deselection: which subject classes and age bands actually circulate, how old unused titles are, how concentrated use is, and what proportion of zero-use books would qualify for weeding under accepted professional criteria (CREW/MUSTIE) once verified against the shelf. This project builds that evidence base from two years of the library's own circulation data and turns it into a weeding plan and policy the committee can act on.

Abstract

This project measures the circulation performance of a degree-college library's print collection over two academic years and converts the measurement into a weeding plan. Title-level issue counts from the Koha circulation module are joined to catalogue fields (DDC class, publication year, accession date) to profile the collection and its use. Statistical tests examine whether use differs by subject class, whether unused titles are older than used ones, and whether age predicts circulation. Zero-use titles are screened against the CREW method's MUSTIE criteria to produce a candidate weeding list, verified on the shelf by sampling, and costed in shelf space.

Introduction

Ranganathan's Fifth Law — a library is a growing organism — has a corollary librarians say less often: growth requires pruning. Most Indian college libraries have acquired decades of textbooks, guides and superseded editions while shelving filled, budgets shifted and syllabi changed. Collections are steadily added and almost never reviewed, because deselection feels like loss and nobody has data to defend it.

Literature gap

Weeding methodology is well documented in Western practice — the CREW manual, standard collection-management texts, and studies of "dead stock" in university libraries. Indian LIS literature on college libraries, however, leans heavily on user-satisfaction surveys and automation studies; usage-based collection evaluation of small affiliated colleges, where the shelf-space crunch is sharpest, is rare. This project supplies a method sized for exactly that setting: one library, two years of transaction data, standard statistics, and a defensible selection framework.

Existing practice vs proposed practice

• Existing: stock verification counts what is missing, not what is unused; weeding, when it happens, is driven by physical condition and hearsay; no usage data is consulted; the catalogue grows and the shelves clog.
• Proposed: an annual, repeatable circulation analysis; subject-wise and age-wise usage profiling; CREW-based deselection criteria written into the collection policy; a shelf-verification step that separates "unused" from "missing".

Feasibility

• Technical: all data comes from Koha's circulation module and catalogue export; analysis runs in Excel and jamovi (free); the statistical tests are standard B.Lib/M.Lib curriculum content.
• Operational: extraction and profiling take about four weeks; shelf verification of the sample takes two; the whole project fits comfortably in 22 weeks alongside coursework.
• Ethical: borrower identifiers are hashed at extraction; no personal reading records are analysed — only counts per title; the library committee's written permission is obtained.

In scope

• One library: print monographs catalogued in Koha, including textbooks, reference duplicates and gift books.
• Two academic years of issue and return transactions; circulation counts per title, not per borrower.
• Statistical profiling by DDC class, publication year and accession date.
• A candidate weeding list screened with CREW/MUSTIE criteria, verified on the shelf by sample.

Out of scope

• Periodicals, theses, e-books and the digital library.
• Borrower-level analysis of any kind — reading histories are never examined.
• Actual withdrawal of books: the committee executes the plan; this project only proposes it.
• Acquisition budgeting and supplier evaluation.
• Inter-library loan and departmental library collections.

Design

A descriptive-analytical case study of one library's collection and circulation record, with a documentary-data core and a small verification component on the shelf. The unit of analysis is the bibliographic title; the observation window is two academic years.

Data sources

1. Circulation data: Koha SQL report of all issues and returns by biblionumber for 2023–24 and 2024–25 (including renewals, marked as such). Member IDs are dropped at extraction; only per-title counts survive.
2. Collection profile: catalogue export with biblionumber, DDC class (080 field), publication year (260/264$c), accession date and item type.
3. Shelf reality: barcode-scan verification for the zero-use sample.

Procedure and timeline

Measures and analysis plan

• Usage: issues per title per year; share of titles with zero issues; median issues among used titles.
• Concentration: share of all issues accounted for by the top 20% of circulating titles (a simple, reportable inequality figure).
• Chi-square: titles classified used/unused × DDC division (10 classes); expected counts checked, sparse divisions merged with the merger reported.
• Mann–Whitney U: publication age (years since publication) of used versus unused titles — chosen because age is skewed and the zero-use group's distribution shape is unknown; normality is checked (Shapiro–Wilk) and a t-test reported as sensitivity analysis if assumptions hold.
• Spearman: title age vs total issues among used titles.
• Weeding logic: candidates = zero-use in both years and MUSTIE-positive (superseded edition, damaged beyond repair, trivial content, or held elsewhere in the cluster) — each with its reason code; verified-missing titles move to the write-off list instead.

Ethics and permissions

Written permission from the library committee; borrower identifiers hashed and discarded at extraction; no individual reading record is accessed at any stage; the weeding list is a recommendation for committee approval, not an act of withdrawal.

The study is a pipeline from transaction records to a committee-ready plan. The figure belongs in the Methodology chapter.

flowchart TD
  A["Koha circulation report: issues/returns by title, 2023-24 and 2024-25"] --> B["Anonymise: drop member IDs, keep per-title counts"]
  C["Catalogue export: DDC class, publication year, accession date"] --> D["Join: title-level usage dataset"]
  B --> D
  D --> E["Profile: issues per title, zero-use list, class and age bands"]
  E --> F["Statistics: chi-square, Mann-Whitney U, Spearman"]
  E --> G["Zero-use titles"]
  G --> H{"MUSTIE screening"}
  H -->|Misleading / Ugly / Superseded / Trivial / Elsewhere| I["Candidate weeding list with reason codes"]
  H -->|Fails all criteria| J["Retention review list"]
  I --> K["Random shelf verification sample of 60"]
  K --> L{"On shelf?"}
  L -->|Yes| M["Confirmed weeding candidates"]
  L -->|No| N["Write-off / missing list"]
  M --> O["Weeding plan, space estimate and collection policy"]
  N --> O
  F --> O

MUSTIE criteria as operationalised here

Each criterion is applied only after the title has failed to circulate in both study years, so popular-but-tired titles are separated from never-used ones.

• Repeat annually so the library accumulates a multi-year usage series and the weeding policy matures from one-off cleanup to routine collection management.
• Add a user-side arm: brief in-library interviews on titles consulted but not issued (reference use is invisible to circulation data).
• Compare print circulation with e-resource downloads from the library's e-subscriptions for the same subjects.
• Extend the design to a cluster of colleges for comparative benchmarks (usage by class, age at first weeding).
• Prototype a Koha saved-report dashboard so the profiling tables refresh automatically each session end.