I spent a Tuesday last month watching a creator record a 22 minute video about a topic he could not summarise in one sentence. He had lights, a decent mic, a second camera angle. What he did not have was a reason for anyone to click. When we got to the thumbnail stage, there was nothing to hold onto. No claim, no number, no face doing anything, no conflict. Just a man in a room who had made something.
That is the whole problem, and it is not a thumbnail problem. It is a packaging problem that got postponed until the thumbnail stage, where it becomes my problem and it is far too late to fix cheaply.
My position, and I know it annoys people: decide how the video will be packaged before you shoot a single frame. Title first, thumbnail concept second, script third. If you cannot package it, you should not produce it. Not yet.
Packaging is not decoration, it is a specification
Most creators treat the title and thumbnail as a wrapper applied at the end, like gift paper. I treat them as the spec sheet. The title is the promise. The thumbnail is the proof that the promise is worth 30 seconds of a stranger's evening. Everything in the video exists to pay that promise back.
When you write the promise last, you end up reverse engineering a claim from footage that was never built to support it. That is where the dishonest thumbnails come from. Not from bad designers, from bad sequencing. A designer handed 40 minutes of unfocused footage has two options: make something vague and truthful, or something punchy and slightly false. Both lose. The vague one gets no clicks. The punchy one gets clicks and then a retention cliff at 40 seconds, and the algorithm reads that cliff far more clearly than it reads the click.
A thumbnail cannot promise something the video never intended to deliver. It can only advertise what you already built.
What the pre-production version looks like
It is not complicated and it takes about 15 minutes. Before shooting, write the title. Then sketch the thumbnail in words, not pixels. Three lines is enough:
- The visual subject. What is physically in frame, and what is it doing.
- The text, if any. Maximum four words, ideally two.
- The tension. What question does this force in the viewer's head.
If you cannot fill line three, stop. You do not have a video, you have a topic. Topics do not get clicked. Tensions do.
Then, and this is the part almost nobody does, go shoot the thumbnail. Deliberately. Set up the shot the concept needs, take 20 frames of it, and move on. You just saved yourself the miserable ritual of scrubbing an export at 3am looking for a frame where your face is doing something interesting. It never is. Faces in mid-sentence look like faces in mid-sentence.
The cost of doing it backwards
I can put rough numbers on this because we run the same pipeline both ways depending on how the client works.
When a client sends us the title, the concept note, and two or three deliberately shot stills, the first draft is usually close. We spend our revision rounds on the things that actually move click through rate: contrast, crop, how the eye travels, whether the text survives at phone size. Turnaround sits comfortably inside our 48 to 72 hours per thumbnail.
When a client sends us a finished export and says "pull something from this", we spend the first two rounds discovering what the video is even about. The design work does not start until round three. Same price, since our pack is $250 for 8 custom thumbnails with unlimited revisions, so the money cost is on me. The cost to the creator is time and, more importantly, ceiling. The best possible thumbnail from bad source material is still a compromise. We are choosing between mediocre options faster, not finding a great one.
On one finance channel around 80k subs, moving to a title-first workflow was worth more than any single design change we made that quarter. Same designer, same style rules, better inputs. I will not pretend the lift was purely packaging order, they also got more consistent with upload timing, but the videos where the concept existed before the camera turned on outperformed the retrofitted ones consistently enough that we stopped arguing about it.
Design decisions get easier when the concept came first
There is a practical reason this matters beyond honesty. Almost every design rule I care about becomes trivially easy to apply when the concept is decided early, and nearly impossible when it is not.
Take contrast. Separating a subject from a background is the single most reliable lever we have, and I have written before about how contrast is doing most of the work in any thumbnail that performs. If you knew the concept before shooting, you wore a colour that separates from your wall. If you did not, I am cutting you out and rebuilding a background, which looks like a cutout and reads as cheap on a small screen.
Or composition. The rules for where the subject sits, how much empty space you leave, and how many focal points survive at 320 pixels wide are all covered in the basics of thumbnail composition, and every one of them is a shooting decision as much as a design decision. Framing yourself dead centre at chest height gives me nowhere to put text. Two steps to the left when you shot it and I have a clean right third.
The series argument
This gets sharper when you are not making one video, you are making a set. Packaging before production is the only way a series looks like a series. If you plan four episodes with a shared visual system before shooting, the colour, the crop, the position of the episode marker, all of it holds together. If you produce four videos and then ask a designer to retrofit a system onto four unrelated sets of footage, you get four thumbnails that vaguely rhyme.
We wrote up the approach in how a set of thumbnails signals it belongs together, and the short version is that consistency has to be a production constraint, not a post-production wish. Same for rebrands. Everything I have learned about changing a channel's visual identity without losing the audience starts from the same place: decide the new system, then shoot into it. Rebranding in post is repainting a car with the engine running.
Where this is hardest, and what to do instead
I will be fair to the objection. Some formats genuinely cannot be packaged in advance. Interviews are the obvious one. You do not know what your guest will say, and the best thumbnail moment is often a reaction nobody could have scripted.
The answer is not to abandon pre-production packaging, it is to change what you decide in advance. For interview and podcast formats we lock the system, not the specific frame. Two-shot or single, where the guest name sits, how much of the quote appears, what the colour rules are. Then during the shoot you capture deliberate stills at the start and end, and you flag two or three moments live when something lands. That is most of what we do in thumbnail design for interview shows, and it works because the decisions were made when they were cheap.
Same logic applies to formats with strict conventions. Legal and professional channels have credibility requirements that make certain visual moves unusable, and if you find that out at the design stage you have wasted a week. Decide the boundaries first.
The other place this bites is when the video and the packaging live in different heads. If an editor is cutting and a separate designer is packaging and nobody talked before the shoot, you get two interpretations of one video. Our sister studio edits video podcasts and clips and the same rule holds on that side: the editors who know the packaging plan cut differently, because they know which moment has to survive.
What I would actually change on Monday
If you take one thing from this, make it a checklist you run before you press record.
- Write the title. Not a working title, the real one, the one that would appear on the homepage.
- Write the thumbnail in three lines. Subject, text, tension.
- If the tension line is empty, kill the video or change the angle until it is not.
- Shoot the thumbnail on purpose. Twenty frames, two minutes, done.
- Only then write the script.
You will notice that steps one to four take about a quarter hour and step five gets easier because of them. A video with a clear promise is a video with a clear structure. The packaging work is not overhead, it is the outline.
I know this reads as a designer telling you to do the designer's job before you hire the designer. It sort of is. But it is also the difference between paying about $31 a thumbnail for something that had a chance, and paying the same for something that was always going to be a rescue job. If you want to see what the first category looks like across different niches, our work is the fastest way to judge whether we are the right studio for yours.
The uncomfortable truth is that the thumbnail is not where clicks are won. It is where the decisions you already made either show up or do not. Make them earlier.