· 3 min read · Jason Youngjin Koh
Test the product before you make it
The most expensive way to learn nobody wants a product is to manufacture it first. A few days of ads can tell you which idea to make.
Create realistic images of each product concept with AI, run them as ads to the real target customer for three days, and count how many people take a committed action: joining a waitlist, placing a refundable pre-order, asking about price and delivery. The concept that wins on committed action is the one to manufacture.
Why test demand before manufacturing?
Minimum order quantities turn a guess into stock. A brand ordering a run of a new colourway, flavour or product line is betting cash on the team's taste. A demand test moves that bet onto evidence from strangers who have to decide whether to spend their own money.
How does a three day product test work?
- Pick three to five concepts you could realistically make, at a price you could realistically charge.
- Create the imagery. AI places each concept in studio and lifestyle settings in minutes, so every concept gets the same quality of presentation and the test measures the idea rather than the photography.
- Build one landing page per concept with the real price and one clear action: join the waitlist, or pre-order with a stated ship date and a full refund option.
- Run the ads with the same budget, audience and format for every concept, so the only difference is the product.
- Read the result after three days. Extend only the concepts that are too close to call.
What should you measure?
The action that costs the customer something. Clicks and likes measure curiosity. A waitlist sign up costs an email address. A deposit costs money. The closer the action sits to buying, the more the result is worth.
| Signal | What it tells you | Weight |
|---|---|---|
| Click through rate on the ad | The image and hook stopped the scroll | Low |
| Waitlist sign ups | People want it enough to be told when it exists | Medium |
| Refundable pre-orders | People will pay for it now | High |
| Questions about price, sizes or delivery | People are already imagining owning it | High, read every one |
Is it honest to advertise a product that does not exist yet?
Yes, when the ad and the page say so. Coming soon, join the waitlist, and pre-order now for March delivery are all honest offers. Taking full payment with no production date is not. Australian Consumer Law prohibits accepting payment for goods you do not supply within the time promised, and Vietnam's consumer protection law expects the same honesty about what is being sold.
Ad platforms review landing pages too. A page that clearly says pre-order passes. A page that pretends the item is in stock puts the ad account at risk, and the account is worth far more than any single test.
How big does the test need to be?
Big enough that the winner is obvious. If one concept draws several times the committed actions of the others at the same spend, three days is enough. If two concepts land within a handful of sign ups of each other, the test has not answered the question. Extend it, or call it a tie and let the written questions from customers break it.
What happens after the test?
The waitlist becomes the launch list. Everyone who signed up hears first when the product exists, so launch day opens with warm buyers instead of cold ads. The winning images and angles go straight into the launch campaign. The losing concepts go into the research file with the reason they lost, which is often more useful than the win.
Product imagery is one of six jobs where agents now carry the volume. What AI agents do in marketing