100 Cheap Tests Before You Build

Formulas#

• Test value = cost of building the wrong thing ÷ cost of the test. Run the tests where this ratio is highest.
• Signal strength: money > time > contact info > words. A payment beats a call booked, a call beats an email signup, and a signup beats "sounds great."
• Pass line: "If fewer than N of the right people do X within D days, I stop or change one thing."
• Riskiest first: test the assumption with the highest risk × uncertainty. It's usually "will they pay," not "can I build it."
• Evidence ladder: complaint → workaround → money already spent → signup → deposit → full payment → repeat payment.

Set up the test (1–12)#

1. Write the pass/fail line before you start. If you set it afterwards, you'll bend it to fit the result.
2. One test, one question. "Do they have this problem?" and "Will they pay $X?" are two separate tests.
3. Give every test a 7–14 day deadline. A test without a deadline turns into a project.
4. List your assumptions and rank them by risk × uncertainty. Test the top one.
5. Set a budget for each test, with a maximum in hours and in dollars.
6. Define the customer before testing. Test with a vague audience and you get vague results.
7. Use the smallest audience that could prove you wrong. 20 of the right people beat 2,000 of the wrong ones.
8. Decide in advance what each result means. Pass → next test. Fail → change one thing, or kill.
9. Keep a test log: hypothesis, method, pass line, result, decision.
10. Change one variable between tests. If you change two, you learn nothing.
11. Test with strangers. Friends and fans are kind, and kindness corrupts the data.
12. Three failed tests on one idea → change the customer or the problem, not the copy.

Problem tests: is the pain real? (13–30)#

13. Count complaints. Search forums, reviews and social posts for the pain in the customer's own words. Many independent complaints mean it's real.
14. Look for workarounds. Spreadsheets, scripts, duct-taped tools and hired help all show that people care.
15. Find money already being spent. If people pay for a bad solution, that's the best signal there is.
16. Ask about the last time it happened, not whether they'd buy. (Rob Fitzpatrick, The Mom Test)
17. Ask "What have you tried?" If they've tried nothing, it doesn't hurt enough.
18. Ask "What does it cost you?" in hours or dollars. If they can't answer, it isn't costing them much.
19. Stop after 10 interviews. Patterns show up by 5–10. Doing more is usually a way to avoid building.
20. Listen for emotion. Sighs, swearing and long stories mean pain. Polite nods mean nothing.
21. Check frequency. A weekly pain beats a yearly one.
22. Check who feels it and who pays for it. If they're different people, you have two sales to make.
23. Read competitors' 1–3 star reviews and group them. The biggest group is the gap.
24. Check search volume for the problem and the solution. Low volume with buyer intent still counts.
25. Check job posts. A company hiring someone to do the job has a budget for it.
26. Check freelance marketplaces. A job people keep paying freelancers for is one you can productize.
27. Ask "Who else has this problem?" If they can name 3 people, the problem is common.
28. Ask for a second call. Someone who won't give you 30 more minutes won't give you money.
29. Watch them do the task on a screen share. What people do differs from what they say.
30. Argue against it. Write the 3 strongest reasons this isn't a real problem. If they hold up, move on.

Demand tests: will they act? (31–55)#

31. A landing page with a waitlist. Send targeted visitors and count signups. A low signup rate from the right audience is a no.
32. A fake door. Add a button for a feature that doesn't exist, count the clicks, then honestly say it's coming.
33. A pre-sale with a real checkout, refunded if you don't build it. This is the strongest test.
34. A small refundable deposit to hold a spot.
35. Show the price before the signup. Signups from people who haven't seen a price don't mean much.
36. Two pages with the same promise and different prices, each getting the same traffic.
37. Two pages with the same price and different promises. The winner tells you which job they're hiring you for.
38. A $100–300 ad test on buyer-intent searches, sent to a landing page. It measures how strangers act.
39. Post about the problem in a community, without mentioning a product. Count the "is there a tool for this?" replies.
40. An explainer video with a waitlist. Dropbox validated this way before the product worked.
41. Sell the result as a service first and do it by hand. (Concierge MVP)
42. Wizard of Oz. The front looks automated and you do the work manually behind it.
43. One-line cold offers to 50 ideal customers. Count replies and calls booked.
44. A letter of intent from business buyers. It's weaker than money but stronger than words.
45. A founding-customer offer. A lasting discount in exchange for paying now and giving feedback.
46. A pre-order goal, refunded if it isn't reached.
47. Email the offer to your list. Count clicks and replies, not opens.
48. Publish one article about the problem. Measure traffic, reading time and signups.
49. A free tool that solves 10% of the problem. Usage measures interest.
50. Sell a template or spreadsheet version for $10–50. If nobody buys the template, the app won't sell either.
51. Pay-what-you-want for an early version. The average payment tells you roughly what to charge.
52. Ask for a card at trial signup and measure the drop-off.
53. Search for competitors. Paying customers of a competitor prove the market exists.
54. Don't trust the size of a waitlist. Trust how many convert to paid when the minimal version ships.
55. Repeat the test in a second channel, so one lucky post doesn't fool you.

Price tests (56–68)#

56. Ask what they pay today for the current solution. That's your anchor.
57. Ask the four Van Westendorp questions: at what price is it too cheap, cheap, expensive, too expensive? The answers give you a range.
58. Say the price out loud and watch the reaction. If nobody flinches, it's too low.
59. Start higher than feels comfortable. Lowering a price is easy. Raising one is hard.
60. Offer 3 tiers and see which one people pick. The choices show you what they value.
61. Test the unit you charge per: per seat, per use or flat. Pick the one that grows as the customer succeeds.
62. Offer monthly and annual. The share who choose annual measures their confidence.
63. Check the ROI. A common rule is to charge 10–20% of the value you create.
64. If everyone says yes to the price, raise it. Some pushback is healthy.
65. If they only buy at a discount, the value isn't clear yet.
66. Test new prices on new visitors only. Never change what existing customers pay in the middle of a test.
67. Read refunds as a signal about the promise. High refunds usually mean the promise is too big, not the price too high.
68. Check that a budget exists. Being willing to pay isn't the same as being able to.

Product tests (69–82)#

69. Prototype in clicks, not code. Use mockups, no-code tools or a shared doc.
70. Build the smallest version that delivers the core outcome once.
71. Five users in a usability test find most of the big problems. (Jakob Nielsen)
72. Measure time to first value: minutes from signup to the first "aha."
73. Measure activation: the share of signups who reach the aha. Fix this before buying traffic.
74. Track retention by weekly cohort. A curve that flattens instead of dropping to zero is the clearest sign of fit.
75. Run the Sean Ellis test. Ask "How would you feel if you could no longer use this?" 40%+ saying "very disappointed" signals product-market fit.
76. Ask free users to pay after real use. The conversion rate measures the value.
77. Watch what they use, not what they request.
78. Hide a feature one click deeper and count who goes looking for it.
79. Count referrals nobody asked for. Users telling others unprompted is a strong signal.
80. Tag the tickets from your first 20 users. The themes are your next roadmap.
81. Ask "would you recommend this?" only after real use. Before that, the answer is a guess.
82. Ship to 10 users before 100, and to 100 before 1,000.

Read the results (83–100)#

83. Count actions, never compliments.
84. "That's cool" means no. "When can I have it?" means yes.
85. A no is progress. A maybe is the worst result, so push maybes to a yes or a no.
86. Discount data from friends and family. They lie to be kind.
87. Discount data from your own fans. Fans will buy anything once, so test with strangers too.
88. The loudest users aren't typical. Weigh the silent majority.
89. Small numbers lie. 3 out of 10 isn't really 30%. Look for strong signals, not decimals.
90. Compare results by channel. The same page converts differently from different sources.
91. Pay attention to surprises. Surprises are the real data.
92. Write the result in one sentence before you decide anything.
93. Before killing an idea, check whether the idea failed or the message did. Change the message once, then decide.
94. Pivot one dimension at a time: customer, problem, solution, price or channel.
95. When a test passes, raise the bar: waitlist → pre-sale → paid pilot → repeat purchase.
96. Validation expires. Retest your core assumptions every 6–12 months.
97. A competitor launching confirms the market exists. It doesn't disprove your idea.
98. No competitors usually means no market, not a hidden gold mine.
99. The final test: a stranger pays full price with no discount and no call.
100. Stop testing once money comes in steadily. From then on, work on improving what's working.