Every cataloger has a rough sense of whether a catalog is "in good shape" — but a sense is not a number, and a number is what lets you tell whether a cleanup project actually worked or just felt productive. Here is what genuinely counts toward catalog health, in order of how much it actually affects a patron finding what they want.
What counts against a catalog
- Missing core fields. A record with no usable title, author or classification is close to unfindable no matter how good your search is.
- Likely duplicates. Two records for the same title split your holdings count, confuse hold queues, and make "how many copies do we have" the wrong answer twice.
- Weak or missing classification. A book with no call number, or one that clearly does not match its subject, is a book that is hard to browse to and hard to reshelve correctly.
- Incomplete records left half-finished from a rushed import or a quick manual add, never circled back to.
- Thin or absent subject headings, which quietly breaks subject browsing without ever producing an error anyone notices.
Why "roughly fine" is not good enough to act on
A catalog that has been growing for a decade accumulates all of the above gradually, and gradual problems are exactly the kind nobody schedules time to fix — there is never a single moment that feels urgent. A large MARC import is usually the moment it becomes visible: a batch of new records with thinner metadata than the existing catalog drags a quality measure down all at once, which is uncomfortable but is actually useful information about the import, not a sign anything is broken.
Measuring it instead of guessing
A single score — built from the checks above, tracked over time rather than taken once — turns "cataloging could be better" into something you can act on and verify: run the check, see the ranked list of problems, work through the highest-severity ones first, and re-run it to confirm the number actually moved. That is the entire discipline. It does not need to be fancy; it needs to be repeated.
What a good process does not do
It does not change your classification behind your back. Finding every record with a missing or malformed call number is a mechanical check, and a good tool can suggest the missing author marks and numbers from fixed rules — but a cataloger still reviews them before they are saved — see our piece on where AI cataloging genuinely helps and where it does not.
A reasonable cadence
- Monthly, for an actively growing catalog — enough to catch an import's effect before it is forgotten which batch caused it.
- Quarterly, for a stable collection — mostly to confirm nothing is drifting.
- Immediately after any large import, migration, or batch edit — this is when a real problem is easiest to trace back to its cause.
Want a health score for your own catalog? Our AI cataloging assistant runs this check on request, and the knowledge-base article explains exactly what it looks for.