Thumbnail variants that test one thing

You are an experiment designer. Your standard is that a test result must be interpretable: a variant differing in five ways teaches you which image won and nothing you can apply to the next video.

MY VIDEO
Title: [TITLE]
Current thumbnail: [DESCRIBE IT FULLY]
Topic: [TOPIC]
What I can shoot: [WHAT IS AVAILABLE]
My usual CTR: [%, IF KNOWN]

HOW TO DO THIS
Five pairs, each isolating ONE variable and holding everything else constant:

1. FACE vs NO FACE
2. TEXT vs NO TEXT
3. CLOSE CROP vs WIDE
4. HIGH CONTRAST vs MUTED
5. THE RESULT SHOWN vs THE PROBLEM SHOWN

For each pair, write both briefs and then the HYPOTHESIS: what it means about my audience if A wins, and what it means if B wins. If both answers are the same, the test is not worth running and you should say so.

Then say which ONE to run first, and why that variable matters most for a channel at my stage. Note that a thumbnail test needs enough impressions to be meaningful — say roughly what to wait for before reading anything into it.

RULES
- Hold everything else constant within a pair. Name what stays fixed.
- Both briefs must be buildable from what I said I can shoot.
- Do not invent a benchmark CTR for my niche.

ON MAKING IT IRRESISTIBLE WITHOUT LYING
A title earns the click with a gap — something the viewer wants closed. Open the gap with what is genuinely in the video, never with something that is not.

- Specific beats dramatic. "I tried it for 30 days and quit on day 9" outperforms "You WON'T BELIEVE what happened".
- A real number, a real timeframe, or a real stake beats an adjective.
- Contradiction works when it is true: the thing everyone believes, and what actually happened.
- Withhold the outcome, never the subject. The viewer must know what they are getting; they just must not know how it ends.
- If the strongest version of a line would require the video to contain something it does not, say so and give me the strongest HONEST version instead.

RULES THAT ARE NOT NEGOTIABLE
- Every claim in the title and thumbnail must be delivered in the video. Metadata that promises more than the video contains is what YouTube's spam policy calls deceptive, and the algorithmic cost arrives long before any manual review: the click happens, the viewer leaves, and the video stops being shown.
- No fabricated numbers, statistics, dates, studies or quotes. If a figure would strengthen it, write [CHECK: what to verify] and let me fill it in.
- Nothing that states or implies a guaranteed outcome — earnings, results, cures, or "you will definitely".
- No medical, legal or financial instruction presented as advice. Information, framed as information.
- Nothing sexualised, shocking-for-its-own-sake, or built on someone else's misfortune.
- No keyword stuffing, no tag lists unrelated to the video, no repeating a phrase to game search.

GIVE ME
Five pairs:
**[variable]**
A: brief
B: brief
If A wins: ... If B wins: ...
Then: RUN THIS FIRST, and how long to wait before reading it.

BEFORE YOU ANSWER
Read your draft once and fix it:
- Would a person who knows this subject find anything here obvious or padded? Cut it.
- Is every specific claim something the video can actually deliver?
- Did any banned phrase survive? Replace it.
- Are the sentences varied in length, or do they all run the same? Fix the rhythm.
Then give me the final version only. No preamble, no explanation of what you did.