Almost nothing you read about a film, a record or a broadcast originates with the person writing it. Entertainment information travels along a supply chain, and knowing the shape of that chain explains most of what is otherwise puzzling about it — why two sources disagree about a runtime, why a release date is wrong in the same way everywhere, why a correction never quite catches up with the error.
The chain starts with the people who made the thing. A distributor issues press materials: a synopsis, a cast and crew list, a running time, stills, a certificate of classification. These documents exist to be copied. They are written to be lifted verbatim, and they usually are.
From press kit to trade press
Trade publications take those materials and add reporting: box office figures, production histories, contract news. The trades are where most durable facts about the industry are first written down in a form outsiders can cite. They are also where the first errors enter the record, because a trade writer working to a deadline will reproduce whatever the distributor supplied.
Databases come next. They ingest the trade press and the press kits, normalise the fields, and expose them for reuse. Normalisation is where information quietly changes shape: a runtime given as "approximately 100 minutes" becomes 100; a credit that read "additional photography" becomes a cinematographer credit; a release described as "spring" acquires a specific day.
Everything downstream — retail listings, streaming catalogues, encyclopaedia entries, the summary a search engine shows you — draws on that normalised layer. By the time a fact reaches a general reader it has usually passed through three or four hands, each of which had a reason to make it tidier than it was.
Why the same fact appears in several versions
Disagreement between sources is rarely a sign that one of them is careless. More often the sources are describing genuinely different objects. A film exists as a theatrical cut, a festival cut, a television edit, a home video release and a restoration, and each has its own running time and sometimes its own credits. A record exists as a domestic pressing, an export pressing and a reissue with extra tracks. A broadcast exists as transmitted and as syndicated.
Databases usually pick one version and present it as the work. That is a reasonable simplification, but it means two databases can both be right and still contradict each other, because they picked different versions.
Circular citation
The failure mode worth watching for is circularity. A database records a date. An encyclopaedia cites the database. A news article cites the encyclopaedia. The database, updating itself later, cites the news article. Four sources now agree and none of them is independent. The agreement looks like corroboration and is nothing of the kind.
The defence is unglamorous: follow citations back until they stop, and see what is at the bottom. If the bottom is a press release, treat the fact as a claim by an interested party. If the bottom is a primary document — a copyright registration, a contemporary review, a broadcast log — the fact is worth relying on.
What to do with all this
Treat entertainment data as testimony rather than measurement. Ask who first recorded it and what they had in front of them when they did. For most purposes the answer is good enough and the exercise is unnecessary. For anything that will be quoted onward, it is the whole job.