How an entity ID works
Everything you create inside an ad account is stored as an object with an identifier: the uploaded image, the video, the post the ad serves, the creative that binds copy to media, and the ad itself. Those identifiers are how the platform recognises things. What the file looks like to you is irrelevant — two pictures that a person would call identical are the same thing only if they resolve to the same stored asset, and two ads that read identically are the same ad only if they point at the same creative object.
This matters most when you copy something. Duplicating an ad produces a new object with a new identifier, so the history attached to the original — its accumulated reactions and comments, its reporting row — belongs to the original and does not travel with the copy. Re-uploading a media file you already used generally resolves back to the stored asset rather than creating a second one, but the ad wrapped around it is still new.
Why entity IDs matter
They quietly decide two things you care about. The first is continuity of proof: engagement lives on a specific object, which is why reusing one post identifier across ad sets keeps a conversation together while duplicating the ad scatters it.
The second is honest measurement. If you rebuild an ad in a new ad set and treat its results as a continuation of the old one, you are joining two rows that the platform never joined. Delivery for a new object starts from nothing, and a material change inside an ad set can send that ad set back into learning while the system re-estimates who to show it to. Reading the first days of a rebuilt ad as though it were mature performance is how good creative gets killed early.
Where entity IDs go wrong
The commonest error is assuming that because a copy looks the same, it behaves the same. It does not carry proof, it does not carry history, and it competes against the original for the same people if the audiences overlap.
The second is losing track. Teams re-upload the same media under different names, build near-identical creatives across campaigns, and end up unable to say which version actually produced the result they are reporting. When someone asks which ad won, the answer has to be an identifier, not a description, because half a dozen ads in the account match the description.
How to act on it
Decide before you copy anything whether you want continuity or a clean slate. If you want the proof and the history, reuse the object — point the new ad set at the same post rather than rebuilding it. If you want an honest test, build something genuinely new and let it start from zero, knowing the early numbers are not comparable to a mature ad’s.
Keep one library of approved media with a naming convention you can search, and record the identifier of anything you report on, so a result can always be traced back to the exact object that produced it. That discipline is what makes ad creative work reviewable months later instead of a folder of similar-looking files.
