Duplicate-looking catalogue records can represent different works, editions or incomplete metadata matches. A shared title or poster is not enough to merge them safely. Start by identifying what each record actually represents and what useful relationships are attached to it.
A cleanup should preserve meaningful choices. Removing one row is not an improvement if it also removes a distinct cut, another language edition or the episode associated with a household’s viewing history. Keep identity and local preferences in view together.
Use several independent clues
Compare the work type, supported provider identifier, release context and explicit version information. A remake and its earlier production can share a title; alternate versions can share a work identity. Those are different relationships and may need different grouping.
Artwork can be wrong or reused. Treat it as a recognition aid rather than the deciding key. Likewise, a year can narrow the search without being globally unique. A good comparison combines the metadata rather than elevating one convenient field into proof.
Review a suspected pair
Do not normalize private resource URLs destructively to force a match. Query parameters and paths can be meaningful. A row-level comparison should not silently change account access data while attempting to organize metadata.
- Record the two entries and their supported identity fields.
- Compare work type, source identifiers and edition labels.
- Check whether the application already groups versions beneath one work.
- Identify local progress, favourites or collections attached to either record.
- Merge, hide or correct only through a supported reversible method appropriate to the actual relationship.
Keep uncertainty visible
If the source lacks enough information to distinguish the records, ask for a specific correction rather than deleting the one with less attractive artwork. An uncertain identity is a data-quality gap. Pretending certainty can make later recovery harder.
After a merge or correction, verify the surviving entry’s details and local relationships. A familiar title remaining in search does not prove that the right version or progress record survived. Check the aspects that mattered before the cleanup.
A useful report says why the records appear equivalent and which fields disagree. That gives a maintainer a verifiable question. The aim is a clearer library with genuine identity differences preserved, not a lower item count obtained by collapsing everything that looks similar.
Useful answers
Is the same poster enough to prove duplicate works?
No. Artwork can be reused or incorrect.
Can a cleanup affect watched progress or favourites?
It can if those preferences attach to one of the removed record identities.
What if the metadata does not establish whether the records differ?
Keep the uncertainty documented and request a specific correction instead of guessing.