A library collection is never finished.
New books are published. Research fields change. Communities grow older or younger. Languages become more or less important locally. School syllabuses change. Databases become expensive. Digital formats appear. Physical shelves fill. Some titles become indispensable. Others become obsolete. A once-unfashionable subject can suddenly become central to public life.
Collection development is the library function that manages this change deliberately.
This article is part of eduKateSG’s How X Works programme and the How a Library Works series.
The shortest useful answer
Library collection development works by repeatedly comparing the library’s mission and users with the collection it actually has, then deciding what to strengthen, acquire, retain, replace, move, share, digitise or remove.
It is not the same as acquisition. Acquisition is the transaction that brings a particular resource under library control. Collection development is the larger strategy that decides why one resource deserves that transaction and another does not.
If acquisition is buying and receiving, collection development is portfolio management for knowledge.
A collection is a model of institutional purpose
No library can collect everything. The collection therefore reflects choices.
A primary school library may prioritise age-appropriate reading, curriculum support, broad curiosity and reading motivation. A university engineering library may require deep technical monographs, standards, journals and databases. A national library may carry preservation responsibilities for the documentary record of a country. A public library may need broad coverage across recreation, lifelong learning, employment, culture and community information.
The collection is thus a material expression of mission. If the mission changes while the collection does not, misalignment grows.
The collection development policy
A collection development policy turns institutional purpose into a repeatable decision framework.
It can define:
- communities served;
- subject priorities;
- desired depth of coverage;
- languages;
- formats;
- selection responsibility;
- donation rules;
- replacement principles;
- criteria for withdrawal;
- special-collection boundaries;
- approaches to controversial material;
- preservation responsibilities;
- cooperative collecting arrangements with other libraries.
The policy matters because a collection can otherwise drift toward whoever makes the loudest request, whichever vendor markets most aggressively, or whatever happened to be fashionable during a particular budget year.
Policy creates continuity without removing professional judgement.
Selection begins with need, not with titles
A weak collection process begins with a stream of titles and asks, “Do we want this?”
A stronger process begins with capability: “What should this library allow its users to learn, research, read, compare or remember?”
Once the need is clear, candidate resources can be compared against it.
This prevents a common error: building a collection that contains many individually good items but has major structural gaps.
Breadth and depth
Every collection makes a trade-off between breadth and depth.
Breadth gives users access to many subjects. Depth gives serious users enough material to go far inside a subject.
A small public library may sensibly prefer broad coverage with selected depth in high-demand local subjects. A research institute may deliberately sacrifice breadth in order to build exceptional depth in one field.
Neither model is inherently superior. The correct balance follows mission.
Core, supporting and peripheral material
Collection development becomes easier when material is not treated as equally important.
A library can think in layers:
- Core: resources essential to the library’s mission and expected to remain reliably available.
- Supporting: material that broadens, updates or deepens the core.
- Peripheral or opportunistic: useful additions that are desirable but not mission-critical.
This layering helps when budgets tighten. The library knows which capabilities must be defended first.
Demand is evidence, not the whole answer
Circulation counts, hold queues, database usage and user requests are valuable signals.
But popularity alone is not enough to define a library collection.
A local history collection may have low everyday circulation yet be irreplaceable. A specialised reference work may serve only a few users but answer questions no general source can. A classic may circulate less than a current bestseller while retaining long-term educational value.
Collection development therefore combines observed demand with mission, rarity, significance, curricular value, research need and expected future usefulness.
The invisible signal: failed demand
Loan statistics show what users borrowed. They do not directly show what users wanted but could not find.
Failed catalogue searches, unanswered reference questions, interlibrary-loan requests and repeated requests for unavailable titles can reveal unmet demand.
This is a powerful distinction. A collection may appear quiet not because users lack interest, but because users have learned that the library does not contain what they need.
Good collection assessment looks for absence as well as use.
Collection mapping
Libraries can map collections by subject, age, format, language, publication date, usage, condition, cost, curriculum relevance and other dimensions.
The purpose is to make imbalance visible.
A subject may appear large in total volume but consist mostly of old editions. Another may have many titles but very little introductory material. A language collection may have strong fiction and weak nonfiction. An academic collection may support one research speciality well but have almost nothing in an emerging adjacent field.
Counting items is therefore only the first measurement. The structure of the count matters.
Age profile and obsolescence
Old does not mean bad. New does not mean good.
But publication age interacts strongly with subject.
A hundred-year-old novel may remain central literature. A ten-year-old guide to a discontinued software platform may be useless for current training but valuable to a computing-history collection. A medical handbook can become dangerous if treated as current clinical guidance after its evidence base has moved.
Collection development therefore asks not “how old is this?” but “how does age affect the function this resource is expected to perform?”
Format is part of the collection decision
The same intellectual content can exist in print, ebook, audiobook, database, streaming media, microform or archival digital file.
Format affects cost, accessibility, preservation, licensing, discoverability and use.
A print reference set may be stable but difficult to update. A database may update continuously but create recurring subscription risk. An ebook may improve remote access but impose simultaneous-user restrictions. An audiobook may be essential for accessibility or preferred learning styles.
Collection development therefore manages capabilities across formats rather than treating format as an afterthought.
Duplicate copies are a capacity decision
Why buy three copies of one title instead of three different titles?
Because sometimes demand intensity matters more than breadth.
Long hold queues, course reading, book-club use or curriculum requirements can justify duplication. Multiple copies reduce waiting time and increase service capacity.
But duplication consumes money and shelf space. The decision is therefore a queueing and portfolio problem at once.
Replacement decisions
When an item is lost or worn out, the library does not automatically replace the exact same thing.
It asks:
- Is the title still used?
- Is a newer edition available?
- Has another format become preferable?
- Has the subject shifted?
- Do we already have enough copies?
- Is the item now rare or difficult to replace?
Replacement is therefore a fresh collection-development decision triggered by loss.
Weeding: why libraries remove books
Removing material can feel contrary to the idea of a library. In practice, responsible withdrawal is essential.
Collections become less usable when shelves are crowded with damaged, obsolete, duplicate or irrelevant material. Old editions can hide newer ones. Overfull shelves make reshelving harder. Users can lose trust when visibly outdated information dominates a subject.
Weeding creates space and signal quality.
But withdrawal must be governed carefully because low use is not the same as low value.
The retention question
Before removing material, a library can ask whether another institution has preservation responsibility.
A widely held mass-market title may be safe to withdraw locally. A rare local publication may need permanent retention even if it has not circulated for years.
This is where resource-sharing networks and cooperative preservation become important. A collection can be locally leaner if the wider network reliably preserves and supplies material.
Collection diversity
A library that serves a broad community needs to consider whether its collection reflects different experiences, viewpoints, languages and levels of difficulty.
This is not solved by simply counting demographic categories. The deeper question is whether users can find credible material that helps them understand both their own world and worlds unlike their own.
A strong collection contains disagreement. It does not require every resource to endorse the same view. What matters is that selection is governed by relevance, evidence, literary or cultural value, representation and mission rather than by accidental historical bias.
Controversial material and intellectual freedom
Libraries sometimes face pressure to add or remove material because a group strongly approves or disapproves of it.
A collection policy helps distinguish between professional review and pressure to erase legitimate viewpoints.
Libraries still make age, legal, curricular and community decisions, especially in school contexts. The important mechanism is consistency: similar decisions should be made under similar principles rather than according to who complains most loudly.
Budgets turn priorities into numbers
Collection development becomes concrete when money is allocated.
A budget may be divided by subject, format, branch, user group or funding source. Recurring digital subscriptions compete with one-time purchases. Price inflation can reduce purchasing power even when the nominal budget stays constant.
This means a collection can weaken without visibly shrinking. If subscription costs rise faster than budgets, the library may quietly lose breadth, cancel titles or stop buying monographs.
Collection development therefore watches not only spending but capability retained per dollar.
Cost per use—and its limits
For digital resources, libraries can divide subscription cost by recorded use to estimate cost per use.
This is useful for comparing large packages and identifying resources that are expensive relative to demand.
But it is not a universal value measure. A specialist database may be expensive per session and still be essential. A cheap high-use entertainment platform may produce excellent usage numbers but serve a different mission.
Metrics inform judgement; they do not replace it.
The long tail
Many collections have a small number of heavily used items and a very large number of rarely used items.
It is tempting to treat the long tail as waste. That can be a mistake.
The long tail is often where research depth, unusual questions, historical memory and serendipitous discovery live. A library is valuable partly because it preserves options before anybody knows exactly who will need them.
The correct question is not whether every item is frequently used. It is whether the low-use layer is still performing a defensible role.
Collections across multiple branches
A multi-branch system does not have to duplicate identical collections everywhere.
Branches can specialise while a shared circulation and transit network moves material to users.
This creates a system-wide collection-development problem: optimise the total network rather than every branch independently.
A title that appears underused at one branch may be highly useful when the system allows efficient requests and transfers.
Digital packages complicate selection
Publishers and aggregators often sell digital resources in bundles.
The library may want thirty journals but have to license three hundred. A platform may add and remove titles. A package may look economical overall while containing large amounts of irrelevant content.
Collection development therefore evaluates package-level economics, overlap with other subscriptions, licence terms, cancellation rights and long-term access.
This is one reason digital libraries require collection strategy as much as physical ones.
User requests and professional judgement
User recommendations are valuable because they reveal demand that selectors may not see.
But a request is one signal among many. The library checks fit, quality, cost, duplication and expected usefulness before purchase.
A mature process keeps the barrier to suggesting material low while keeping the standard for permanent collection responsibility appropriately high.
Collection development as feedback control
Mechanically, collection development is a feedback loop.
- Mission defines the desired capability.
- The existing collection is measured.
- Users generate demand, use, failed searches and requests.
- Selectors compare observed reality with desired capability.
- Gaps, excesses and risks are identified.
- Resources are acquired, retained, replaced, moved or withdrawn.
- The changed collection produces new patterns of use.
- The cycle repeats.
The collection is therefore a controlled evolving state, not a stockpile.
Collection development and AI
AI can scan publication streams, detect topic growth, cluster requests, identify subject gaps, compare overlap among packages and summarise evidence for retention or cancellation.
It can make the observation layer much stronger.
But the desired state still requires institutional judgement. A model can predict use; it cannot by itself decide what a school, university, nation or community should preserve for the future.
Collection development is therefore a good example of where AI can improve measurement without owning the mission.
The complete mechanism
- The library defines whom it serves and what capabilities its collection should provide.
- A collection policy translates that mission into durable principles.
- The current collection is mapped by subject, depth, format, age, language, condition, cost and use.
- User demand and failed demand reveal where reality differs from expectation.
- Selectors identify gaps, duplication, obsolescence and preservation priorities.
- Budget is allocated across competing needs.
- Acquisition brings selected material into the collection.
- Replacement, relocation, digitisation and duplication adjust capacity.
- Weeding and cooperative retention remove clutter without destroying important memory.
- Usage and research needs are measured again.
- The collection is continuously renewed rather than periodically rebuilt from scratch.
That is how library collection development works.
It is the discipline of keeping a collection aligned with a moving world: deciding what deserves to enter, what deserves to stay, what can safely leave, and what must remain reachable even before anyone knows who will need it next.