Startup Experiment Card

Use this card to answer a specific question: “Is the next investment of effort worthwhile?” Write the assumption, scope, and decision conditions before approaching users. This is a working template suggested by the guide, not a market validation standard. The two-week period is adjustable, and the examples are entirely hypothetical.

Test one main assumption at a time. Record questions outside the scope without quietly expanding the card into a complete business plan. For the method, see From a Real Problem to First Customers. For working time and operational observations, see A Sustainable Small Business.

Copy the Blank Card

Experiment name / identifier:
Owner / start and end dates:
The next decision this experiment will inform:

[COMPLETE BEFORE STARTING]
Target users and specific task:
User / adoption decision-maker / payer:
Evidence that the task occurred recently:
Current alternatives and their costs:
Main assumption (one statement that could be disproved):
Strongest counterexample or alternative explanation:

Contact channel / eligibility conditions:
Planned invitations / interviews / pilots:
Possible recruitment bias (such as friends or volunteers):
Pilot deliverables / explicit exclusions:
Price / revision scope / delivery and acceptance method:
What people do and what tools do:
Material permissions / recording consent / retention and deletion:

Old-process baseline (quality, time on both sides, spending):
Behavioral evidence to record and where it will be stored:
Advance conditions for continuing:
Advance conditions for adjusting:
Advance conditions for stopping or pausing:
Time limit / cash spending limit / deadline:
Who can stop the experiment / how unfinished commitments are handled:

[COMPLETE AFTER ENDING]
Actual invitations / responses / interviews / quotes / payments / acceptances:
User behavior (anonymous ID + date + evidence summary):
Counterexamples / refusals / withdrawals and reasons:
Actual quotes / payments received / refunds / spending:
All hours (recruitment, communication, preparation, production, revisions, support):
Old/new process comparison and conditions that remain incomparable:
Departures from the original plan and why:

What the evidence supports / what it does not support:
Decision: continue / adjust / stop / insufficient evidence, pause
Reason for the decision:
Unfinished commitments and responsible person:
The single main change in the next round:
What new fact would change this judgment:
Next review date:

Each field can be brief. An evidence summary might say, “User A showed de-identified examples of two tasks from last week.” Do not paste complete customer materials into a public repository. Keep only the anonymous summary needed to verify payment; private contact details and transaction receipts belong in an appropriately restricted location.

Worked Example: Organizing Product Information

This example tests only whether shop owners will pay for a limited information-organizing service and whether delivery can stay within a time allowance. It is not a plan to launch an AI product. All people, behavior, and numbers below are fictional illustrations.

Anchor the baseline in a particular task too. Suppose the two pilot users supply timing summaries of their old process: 90 and 60 minutes respectively. In the new process, user preparation, communication, and review take 30 and 70 minutes, while provider work takes 90 and 210 minutes. There is one signal of user time saved and one counterexample where user time increases. Do not showcase only the first user.

Example Results and Interpretation

Suppose 10 people were invited, 8 responded, and 6 completed interviews. Four showed recent recurring tasks. Three received the same quote; two paid and accepted delivery, while one declined because the existing template was sufficient. Payments totaled 400 yuan, with no refunds and 40 yuan of cash spending. All recruitment and delivery work took 11 hours. Provider work on the two orders took 1.5 and 3.5 hours, with 6 hours for other work: still 11 hours in total.

The second order involved inconsistent input formats and extra checking. The user’s own preparation and review also took longer than the old process. Although payment and acceptance counts met the conditions, the per-order time condition did not. The decision is therefore adjust. The next round will accept only materials that follow the input template, keeping the price and quantity limit unchanged, and test working time again. That proposed improvement is still an assumption; it cannot already be described as a solved problem.

This round supports “These two users accepted a paid pilot under these conditions.” It does not establish regular repeat purchases, acquisition costs through unfamiliar channels, or sustainable profit. 400 - 40 = 360 yuan is only the remainder after the listed cash spending. It excludes eleven hours of labor and other costs, so it is not a claim of 360 yuan in profit.

If only two interviews were completed, record “The planned evidence volume was not reached; judgment is paused,” explaining whether recruitment was difficult or too little effort was available. Do not lower the threshold to two after seeing the results and claim success. If a plan must change, retain the original version, change date, and reason, and apply new conditions to the next round.

Finish with a Checkable Conclusion

Within two weeks and limits of twelve hours and 100 yuan of spending, we observed two paid and accepted deliveries, but one exceeded its delivery-time allowance. We will narrow the accepted input format and run another round. We are not yet increasing order volume.

For a shorter English update:

Two customers paid for and accepted the pilot, but one delivery exceeded our time limit. We will test a stricter input format before taking more orders. Repeat demand and sustainable costs remain unproven.