GuidesHow to2 min
How to test an idea before you build
A test is a user, a thin lie, and a number. If you cannot fake it this week, you are not ready to spend on the real thing.
A prototype that cannot fail a named question is a first version wearing a costume. Write the question on the box. Throw most of the box away on purpose. If you cannot fake the job this week — paper, a jig, a person behind a curtain — you are not ready to spend on the real thing.
People will say yes to a story. Count what they do with an object. SINLE will not take an Invent until a thin lie has been on a dirty Tuesday: night shift, bad inputs, the skeptical operator. Demo data is a lie. If they will not live with the slow version, they will not pay for the automated one.
The sequence below is the test. A meeting is not a decision. Build, rewrite, or stop on the date you wrote.
A test that can actually fail
Belief: night shift will log a stop reason on paper if the sheet is at the cell. Number: four of six leads do it for five shifts. Lie: a paper log, not an app. Date: next Friday, build or stop. If you cannot name the number, you are not testing. You are hoping.
People say yes to a story. Count what they do with the paper. If they will not live with the slow version, Invent will not save it.
The sequence
Do these in order. Skip a line and you are back to a workshop.
- 01
State the belief in one sentence
“They will wait for a stamp at the cell if we put a gauge here.” Not “build a platform.”
- 02
State the number
Time, scrap, or a yes/no from five users. “Feedback” is not a number.
- 03
Use the cheapest lie that can be wrong
Paper log, a person behind a curtain, wood instead of steel, a one-page site.
- 04
Put it on the worst real Tuesday you can afford
Night shift, dirty inputs, the skeptical operator. Demo data is a lie.
- 05
Decide on the date you wrote
Build, rewrite, or stop. A meeting is not a decision.
Keep these
- If the lie does not change the job, the expensive version will not either.
- People will say yes to a story. Count what they do with an object.
Do this before you spend
- 01Write the belief, the number, and the date on one card.
- 02Run the cheapest lie on a real shift this week.
- 03Decide on the date. Build, rewrite, or stop. Do not add a meeting.
If the lie already failed, do not book a build. If it held, send the number. We will scope what survived.
Questions
- How do you test an innovation before building it?
- Do the job with a thin version: paper, a concierge, a script. If users will not live with the slow version, they will not pay for the automated one.
Next how-to
- How to go from prototype to productionProduction means auth, real data, a kill switch, and a URL the company can send. The path in between, without a two-year programme.
- How to find the job worth fundingWalk the work. Score waits, rework, and hero jobs. Fund a constraint with a user — not a whiteboard of ideas.
- How to kill a project that is not workingUse the stop rule you wrote on day one. A clean kill is a result. A zombie project is how next month disappears.
Next
Bring the test, not the vision.
If the thin version held, Spark turns it into a written scope. If it failed, you just saved Invent.