UniqueizerPROWhy identical files hurt creative test cleanliness
You uploaded one video into three ad sets and the numbers came out three times apart. It is not always the audience: the platform sees the same file and treats it as one object.
One file across several ads is one creative to the platform, not three. The test turns into an audience comparison. Variants have to differ technically: crop, colour, speed, grain, metadata.
What the platform sees
Ad systems fingerprint the uploaded file. Two byte-identical files are one object to them, even in different campaigns under different names.
The consequences follow from that: history can carry over, impressions can be distributed unevenly, and a "new" creative may not start from a clean slate.
How it breaks the test
| What you think you are testing | What is actually tested |
|---|---|
| Three creatives | One creative across three audiences |
| Different angles | Different bids and segments |
| A fresh launch | The continuation of an old one |
The result looks convincing and does not reproduce: move the winner to another offer and it fails.
How variants should differ
Renaming is not enough — the name is not part of the content.
- Micro-crop and slight rotation — shift the whole pixel grid;
- Colour — brightness, contrast, saturation, hue;
- Grain — light noise, invisible at low levels;
- Speed — for video, with audio kept in sync;
- Compression — different JPEG quality or bitrate;
- Metadata — EXIF and service fields stripped.
How far to go
- LowChanges are barely visible, for delicate offers.
- MediumThe balanced default.
- HighThe file changes visibly.
The working order
Build one master creative. Validate it on a small budget. Multiply it into three or four technically different variants. Then the test result is about the angle, not about a file collision.
Questions people ask
Is renaming the file enough?
No. The name is not part of the content, so the fingerprint stays the same.
Does processing degrade the creative?
At the low level the change is barely visible. High levels are visible, which is why they are chosen deliberately.
Is this an attempt to bypass moderation?
No. The content does not change and platform rules apply exactly as before. This is about test cleanliness.
How many variants should I make?
Three or four per ad set. Enough for a test and for rotation.