'No clear winner' is a result
An inconclusive Test & Compare result is information, not a glitch. Why variants were too similar or impressions too thin, and what to change before you re-run.
An inconclusive Test & Compare run is not a broken tool. It is the tool telling you the variants did not separate on watch time hard enough, or that the video did not get enough impressions for a separation to count. YouTube will still put something on the video: the first title or title-and-thumbnail combination you uploaded. That fallback is order, not merit. Treating it as a win is how a coin-flip default becomes "the thumbnail that tested best."
YouTube Help names three endings, all based on watch time share: Winner, Performed Same, and Inconclusive. "No clear winner" in hallway speech covers the last two. They are not the same instruction, and they are not a reason to immediately ship a fourth variant. This post is how to read the result, why the first-upload default is a trap, and what to change before you run the test again.
The three endings, in YouTube's words
Winner. One option outperformed the others on watch time share, and YouTube is willing to call the difference statistically significant. That option is shown to all viewers. This is the only ending that is a ranking.
Performed Same. The test ran, and the options performed about the same. There may be small differences. There is no clear winner. YouTube's instruction is to pick the option you prefer.
Inconclusive. There was no strong statistical difference in engagement. In this case, the first title and thumbnail you uploaded become the default. You can still change them manually after the test.
Both of the non-wins are information. Performed Same says the treatments did not matter at a level the tool will act on. Inconclusive says it could not tell: similarity, thin impressions, or a test that did not finish cleanly.
YouTube lists two reasons a video would not get a Winner:
- Minimal difference in titles or thumbnails. The gap you designed was not big enough to move watch time.
- Not enough impressions. Low-reach videos are less likely to get a Winner. More views make a Winner more likely, which is a sample-size statement, not a quality statement.
A third, operational reason sits elsewhere in the same article: if you change the title or thumbnail during the test, the test automatically stops and you have to restart. A stopped test is not Inconclusive in the interesting sense. It is an abandoned sample.
The first-upload default is not a finding
When the result is Inconclusive, YouTube applies the first-uploaded variant. That is a tie-break, the same way a config file has a default. It is not evidence that variant 1 held viewers better. If you uploaded the face-forward thumbnail first because it was ready, and the process-shot second because the still had to be exported, Inconclusive plus fallback will make the face-forward image look like the winner in every downstream screenshot.
Performed Same is slightly kinder: YouTube tells you to pick the one you prefer. That is still a judgment call. It is at least labelled as one.
After either non-win, do these three things before anyone updates the brand kit:
- Write down which variant was first. If you cannot remember, look at the test setup, not at the live video. The live video is the fallback, not the log.
- Do not copy the live packaging onto the next five videos. That is how a default becomes a house style.
- Decide whether you are going to pick by preference (Performed Same) or re-run with a bigger gap (either ending). Re-running the same three images is how you get the same ending.
You can always change the title or thumbnail manually after the test. Use that. The tool will not be offended.
Why the variants failed to separate
Most "no clear winner" runs I have watched fail in one of four ways. Only one of them is "the video is too small."
The variants are the same picture. A colour grade, a different font on the same still, a title that swaps a synonym. YouTube warns that testing titles and thumbnails that are too similar makes tests run longer, because there may not be enough difference to decide a winner. Watch time is a coarse instrument. It will not detect that you moved the face six pixels.
The variants change the click, not the session. All three images attract the same people, who then watch the same video the same way. CTR might have moved. Watch time share does not. If you designed a CTR test, you ran it on a watch-time instrument. That is an Inconclusive or Performed Same waiting to happen, and it is not the tool's fault.
The video did not get the impressions. YouTube says this plainly. New uploads, limited traffic sources, and quiet catalogue pieces can all finish without a Winner. YouTube's own tip is to test older videos first: a bad variant does less damage, and older videos already have an impressions engine. A two-week test on 400 impressions a day is a small sample dressed as an experiment.
The test was disturbed. Title or thumbnail edited mid-flight (YouTube says the test automatically stops). Format identity changed, for example a vertical re-upload that gets treated as a Short (you lose access to tests on that video). Those are process failures, not statistical endings.
A fifth, quieter issue: thumbnail resolution. If any experiment thumbnail is below 720p (1280×720), YouTube downscales all experiment thumbnails to 480p. A sharp variant and a soft variant become the same muddy variant, and it shows up as Performed Same.
What to change before you re-run
Do not re-run until you can point at the cause above. Then change one of these, not all of them.
Make the gap a category gap. Face versus process versus text-led. Promise A versus promise B. A still that exists in the video versus a still that does not. Batching creative variants is useful here only if each variant is a different idea. Three crops of one frame are one idea. Generating thumbnails and pulling a still from the timeline are the two cheapest ways to get a real composition change. Extracting a frame from a moment people already rewatch is how you keep the image honest to the watch, which is the metric the test uses.
Move the test onto a video that actually gets impressions. YouTube recommends older videos first. Pick one that already has a Reach tab worth reading. A packaging question you care about ("does the red-arrow thumbnail hold strangers") should be asked on a video strangers already see, not on yesterday's upload that has not left subscribers yet.
Wait the window. Tests take a few days to two weeks. Peeking at day two and stopping because you are impatient produces a sample that cannot win. If you must stop, stop. Do not call it Inconclusive and then treat the fallback as a finding.
Keep every thumbnail at or above 1280×720. If one file is a 720-wide export from a vertical frame, you have just downscaled the set. Generate or extract at the size the test will actually show. The YouTube thumbnail craft still applies: the picture has to read at small sizes. It also has to enter the test at the resolution YouTube asked for.
Do not touch title or thumbnail until the report is in. If a variant is embarrassing, you live with it for two weeks on an older video (another reason to test catalogue first). Mid-test edits reset the work.
Pre-commit what you will do on each ending.
| Result | Action |
|---|---|
| Winner | Ship that variant. Do not "correct" it back to the higher-CTR loser. |
| Performed Same | Pick by preference, or by which image you can reproduce. Record that it was a pick. |
| Inconclusive | Treat the live packaging as a default. Either leave it, or manually set the one you prefer. Re-run only after you change the gap or the video. |
The surrounding A/B discipline still holds: one variable, a stop rule, no peeking. "No clear winner" means the test did its job and the world refused to rank your three pictures. That is a result. Design a different comparison.
FAQ
If the first-uploaded variant becomes default, should I put my favourite in slot one?
Only if you are trying to win the fallback, which is not the same as running a test. Upload order should be arbitrary, or recorded, so that an Inconclusive ending does not secretly encode your preference as a finding. If you already know which one you prefer, YouTube's Performed Same instruction is "pick the option you prefer." You do not need the tool for that.
Can I re-run the exact same three variants on the same video?
You can start another test after one finishes. If nothing else changed, expect the same class of ending. Statistical variation exists (YouTube compares it to flipping a coin) and audience mix shifts over time, so you might get a Winner on a second run by chance. That is not a reason to keep rolling the same images until one of them lucks out.
How is Performed Same different from Inconclusive in practice?
Performed Same: the test ran, the options looked alike, you should choose. Inconclusive: the test could not establish a difference, and the first upload becomes default unless you change it. In both cases you are allowed to set the packaging manually afterwards. In neither case did a variant "win."
Should I test brand-new uploads at all, or only catalogue?
YouTube recommends testing older videos first to reduce the impact on overall views. New uploads can be tested; they are more likely to end Inconclusive, and a weak variant costs you the first days of a video that will not come back. Catalogue first is the order the Help Center already suggests.