Examples
Two thumbnails, fixed step by step
The guide explains what each factor means. This page shows the factors doing something — two realistic packages taken apart, rebuilt, and rescored.
Case 1 · Thumbnail craft
The thumbnail nobody can read
A tutorial channel exports a thumbnail that looks fine on a 27-inch editing monitor: a full sentence of explanation across a screenshot of the software, plus a corner badge, plus an arrow. Shrunk to 168 pixels wide, the size of the test, the sentence is a grey smudge, and three elements compete for a glance that lasts well under a second.
In your editor
Small-size test · 168px
A sentence, thin and mid-grey, over a busy screenshot. Nothing dominates.
In your editor
Small-size test · 168px
Three words, heavy, near-white on dark. One subject, one accent.
Same markup, same proportions — only the width changed. At this size, smaller than YouTube shows a thumbnail anywhere we measured in October 2026, only what survives a glance is left. The checker shows your own thumbnail at this size as soon as you drop it in.
The 3 factors that moved most
- Thumbnail Readability28 → 82 (+54)
- Title / Thumbnail Coherence45 → 78 (+33)
- Emotion40 → 58 (+18)
What changed
- Cut the overlay from eleven words to three. Anything longer than about four words stops being legible at feed width.
- Raised contrast: near-white type with a hard shadow on a dark base, instead of mid-grey on mid-grey.
- Deleted the corner badge and the arrow. Two competing elements became one subject plus one accent.
- Stopped the title from repeating the thumbnail text, so the two now carry different information instead of saying the same thing twice.
The takeaway
The gain came from more than one factor. Readability was one of the lowest scores on the board and the most visible fix, but coherence, once the title stopped repeating the thumbnail, added about as much. The lesson is not "do these four things"; it is "fix the low scores that matter, then re-measure before touching anything else."
Case 2 · The honest hook
The hook that beat itself
A creator has learned that tension drives clicks, and pushes every lever at once: an all-caps shock title, a devastated expression, an implied catastrophe the video never actually delivers. Every curiosity factor goes up. The score goes down — and the reason is the most useful thing this tool will ever tell you.
I LOST EVERYTHING… (I'm done)
Your channel
Maximum tension, minimum information. The video cannot pay this off.
Why I shut down my $40k/month channel
Your channel
Still a real hook — but it promises exactly what the video delivers.
The 3 factors that moved most
- Promise Clarity25 → 88 (+63)
- Clickbait Risk85 → 22 (-63)
- Title / Thumbnail Coherence35 → 82 (+47)
What changed
- Curiosity, cliffhanger, FOMO and urgency all went DOWN. The rewrite is deliberately less tense than the original.
- Promise clarity went from 25 to 88: a viewer now knows what they are about to watch before they click.
- Clickbait risk fell from 85 to 22, which shrank the penalty from −36% to −5%.
- The CTR layer got worse — 80 down to 68 — and the final score still went from 35 to 71.
The takeaway
This is the case that justifies the whole scoring model. The hyped version wins on almost every pre-click signal and still scores 35, because the two layers are multiplied rather than averaged: a package that earns the click and loses the viewer is worth less than one that does both moderately well. If you only ever take one thing from this tool, take this — you are not optimising for clicks, you are optimising for clicks that survive.
Now do it with yours
Run your own thumbnail and title through the analyzer, apply the first change under Fix these first, and run it again with "Test a new version". The factors that moved tell you what worked.