The test is supposed to measure demand. Most of the time it measures how homemade your ad looked.
You have an idea. Before building it, you do the sensible thing: put $100 behind a Facebook campaign, send the clicks to a landing page, count the signups. If nobody signs up, you saved yourself six months.
Except that is not what you measured. You measured whether a stranger, scrolling at speed, believed your thing was a real company. A stock photo with a headline typed over it in a default font, pointing at a page with a purple gradient and a bare email field, gets scrolled past by people who would have bought the actual product. The idea did not fail. The surface did.
And the honest reason you shipped that surface is the chicken-and-egg problem: it is irrational to commission a brand for something you are about to find out nobody wants.
Not a brand. A brand is a long-lived asset with a strategy behind it, and you cannot justify one yet. What you need is a believable visual surface — enough consistency that a stranger stops reading the design and starts reading the offer.
That is a short list:
Nothing on that list requires knowing who you are yet. It requires making one decision about how the thing looks, and then not contradicting it.
This is the part people get backwards. A scrolling stranger is not grading your kerning. They are pattern-matching against every real company they have ever seen, and real companies are internally consistent: the ad, the page, and the product all look like one decision.
Fakes are inconsistent. A slick stock ad leading to a template page with different colours and a different mood is the tell. Two mediocre-but-matching assets beat one beautiful asset and one that contradicts it — every time.
A mood board does not make your ad look real. What you pull off the board does.
Describe the idea in a sentence or two and MoodyBoards generates a board: a palette, type, and six original images that agree with each other. From that one artifact you take:
The board’s images are original generations, not stock everyone has already scrolled past. Pick the two strongest, crop to 1:1 and 4:5.
A different image from the same board. Same world as the ad, so the click feels like an arrival rather than a bait-and-switch.
Already decided, already on the board. This is what stops your landing page defaulting to purple-gradient-and-Inter.
Hit Build your website and the board becomes a brief — colours, type, mood, every image — that Claude Code turns into the actual page.
If you have never used Claude Code, the free ten-minute course takes you from install to a live page.
Making an unbuilt product look credible and lying about it are different things, and the line is easy to stay on the right side of.
With a consistent surface in place, a failed test finally means what you wanted it to mean: people saw a credible offer and did not want it. That is worth the hundred dollars. A failed test on a homemade surface tells you nothing at all, and the real cost is that it talks you out of an idea that was fine.
The other outcome is the useful one. When it works, you already have the direction — the board is the thing you hand to whoever builds the real version, and the ad that won is evidence about which images did the work.