Run the test that can ruin your favorite assumption
A good test is small enough to survive and honest enough to disappoint you. The goal is a decision, not a tiny theatrical production called “launch.”
Test one fatal assumption at a time. Give the test a fixed audience, offer, channel, budget, deadline, and pass/fail rule before you see the numbers. Otherwise every result becomes “promising.”
Find the assumption that could waste the most money
Demand, access, price, delivery, and margin are separate risks. A landing page cannot prove you can deliver wedding cakes on time. A beautiful sample cannot prove customers will pay enough. Pick the question that, if answered “no,” makes everything downstream foolish.
Use the smallest credible test for the business
| Business | First test | Keep it honest | Useful result |
|---|---|---|---|
| Service | Sell 3–5 paid pilot slots and deliver manually | State the pilot scope, price, timeline, and limits | Qualified calls, deposits, delivery time, margin, objections |
| Physical product | Show a real prototype or sample; take a limited preorder only if fulfillment and refunds are clear | Show actual materials, dimensions, date, shipping, and refund policy | Paid orders plus production and shipping reality |
| Local business | Pop-up, borrowed room, market stall, mobile route, or one-day event | Use the intended neighborhood and realistic hours | Foot traffic, conversion, average ticket, repeat intent |
| Digital product | Manual concierge version, clickable prototype, or narrowly scoped paid beta | Do not pretend automation exists when a human is doing it | Activation, completion, willingness to pay, repeat use |
Waitlists can help compare messages or channels, but Reddit founders repeatedly warn that vague waitlists let people imagine different products. Treat a signup as permission to continue the conversation, not a purchase forecast.
Write the test card before spending
- Audience: exactly who will see it and why they qualify.
- Offer: deliverable, price, timing, and conditions.
- Channel: where these buyers already look or talk.
- Budget: a loss you can tolerate; early organic outreach may cost $0, while a paid traffic test should still have a hard ceiling.
- Pass: the action and count that justify one next commitment.
- Fail: the result that triggers a message, audience, price, or offer change.
Do not borrow a universal conversion benchmark. A $12 impulse product and a $4,000 professional service have different buying paths. Set your threshold from the economics: how many qualified prospects can you reach, how many customers do you need, and what acquisition cost can the gross profit support?
Run one clean seven-day cycle
- Day 1: write the offer and test card.
- Days 2–3: show it to five target buyers and repair confusing language, without moving the pass line.
- Days 4–6: make 20–30 targeted invitations or drive a predetermined amount of relevant traffic.
- Day 7: count exposures, replies, qualified conversations, buying attempts, sales, refunds, and reasons for no.
Twenty targeted invitations are more diagnostic than a broad post seen by thousands of people who will never buy. Save exact objections. “Too expensive” may mean price, weak value, wrong timing, or no authority to purchase.
Change one thing after a failed test
No response suggests an audience, channel, or opening-message problem. Conversations without buying attempts suggest weak urgency, offer, trust, or price. Purchases followed by refunds suggest expectation or delivery failure. Change one variable and rerun once. Changing the audience, offer, price, and channel together produces fresh confusion.
You cannot get qualified people to discuss the problem or attempt the offer after two well-targeted cycles. More features will not repair missing demand.
Your finish line
You are done when the result clearly earns one next step: another test, a small setup purchase, a paid pilot, a pivot, or a clean kill. Write that decision down before opening a design tool.