Thumbnails
Three variants, three concepts, three titles, so that a test has something to learn from and a channel has something to choose between.
The short version
Is this for you? For the channel shipping one thumbnail with no rival and losing the click at first glance. It needs enough traffic for a test to mean something; on a channel that posts rarely, we will tell you the sample is too thin to learn from.
What leaves your plate. Three genuine concepts rather than three shades of one, a face, the subject, the parts laid out, each with its own title, all read at the size a thumbnail is actually seen.
What starts working. Packaging a channel can test rather than admire, in a visual language a returning viewer knows before reading a word, with the source frames kept where a future editor can find them.
The smallest way in. A free audit: three fields, and a written reading back, usually within about three working days. It commits you to nothing.
What to read next. The five levels to see where this fits, or the casework to see it done.
Three concepts, read at the size of a thumbnail
Three concepts, so the test can tell you something
Not three shades of the same idea. A face, the subject itself, and a flat lay of the parts, each with its own title, its own type treatment, its own button and its own colour combination. One thumbnail is a preference with a publish date on it, and a channel that ships one has no way of finding out whether it was right.
A thumbnail is read at the size of a thumbnail
Title of four words or fewer, legible on a phone held at arm's length. A face at roughly a third of the frame, because smaller stops registering and larger stops leaving room for anything else. Corners kept clear so a duration stamp or a platform badge lands on empty space and leaves the working word alone. These are where we start, not where we insist: where your own click data prefers a different face size or a longer line, that is the rule for your channel.
Stills go through a proper development pass
Frames are pulled and developed through Camera Raw or Lightroom before anything is designed on top of them, which gives the exposure and colour latitude a thumbnail wants, even though a pulled video frame is not a camera raw file. Backgrounds get blurred where a photograph runs full bleed. Processed files are stored in the video's own folder, named for the video, so the third variant can be found in a year by somebody who was not there.
At one partner, thumbnails are contractually not ours
Written into that partnership's standard as an explicit exclusion, and it stays excluded. The packaging decision belongs to them, they have their own view of how the channel should look, and a studio that quietly does the work anyway is overriding a partner's judgement about their own brand while calling it a favour.
What has to survive at phone size
| Title length | four words or fewer |
|---|---|
| Face | about a third of the frame |
| Corners | left clear for timestamps and platform badges |
| Variants per video | three concepts, three titles |
| Source stills | developed from pulled frames, never screen grabs |
| Storage | in the video's own folder, named for the video |
| Overridden by | your own click data, where it prefers a different face size or a longer line |
Where the variants come from, and what they are for
Where the variants come from
Concepts are drawn from the channel's own visual identity, from what its near competitors have made ordinary, and from the part of the video a viewer would actually want. Type and colour follow the brand book and never the current fashion, and the design language stays clean and vector-led so it survives being scaled down twice.
What the test is for
Three genuinely different concepts answer one question, and only one: which whole package an audience wants. Subject, title, type and colour all change together between them, so a winner tells you the concept won, not which part of it did the earning. Naming the part, title or image, is a second and narrower test with exactly one difference left in it, run on the concept that already won. Where the channel is eligible, the three concepts run together in YouTube's own tool, which settles them on share of watch time rather than click-through alone; where it is not, the reading is click-through and retention held side by side. Either way the loop is the same: find the concept that earns attention, then, where it is worth knowing why, isolate the one thing that did. Designing that comparison as a method, one honest variable and one question at a time, is the A/B testing page; here the job is producing the three concepts worth putting into it.
A thumbnail is not a poster; it is a decision taken at the size of a fingernail by somebody who is already scrolling past it.
It is read at the size it appears, not the size it was designed at, which is where most of them are lost.
What changes for you
Packaging a channel can test rather than admire, in a visual language a returning viewer recognises before reading a word of it, with the source frames kept somewhere a future editor can find them.
Related: A/B testing · Titles, descriptions and metadata · Reading a video · Between the uploads
The audit reads the channel the way a manager would: cadence, packaging and what the analytics are asking for next. What we would change comes back in writing, before any commitment.
Request your free audit.
Three fields, and a written reading back: a fit snapshot, two or three prioritised opportunities, and a recommended next step.
A person reads the channel and writes the audit by hand: a considered read typically takes three working days. That is the usual shape, not a promised turnaround. We use these details only to reply to you: no lists, no lurking.
What you will get
A fit snapshot: where your channel stands, and whether we are a match.
Two to three opportunities: specific, prioritised, yours to keep.
A recommended next step, even if that step is not us.
The audit is free and commits you to nothing: nobody follows up with a call you did not ask for.
The casework shows all of this done.
The case studies are the same work with the numbers, the sources and the dates attached.
See the casework →