SwiftCNS
Get started

Guide

How to test an idea before you commit

Name the assumption, set success criteria before anything runs, run the smallest experiment that could prove you wrong, then decide and record why.

  1. 01

    Find the assumption the idea rests on

    Every idea depends on something being true: that customers have the problem, that they will pay, that the new flow is short enough. Write down the one that would sink the idea if it turned out false. Be specific: “new customers will pay $49 a month”, not “pricing”.

  2. 02

    Set the success criteria before anything runs

    Decide what result would count as success: a number and a threshold. Agree it with whoever will make the decision, write it down, and don’t change it once results start coming in. Criteria set afterwards can make almost any result look like a win.

  3. 03

    Run the smallest experiment that could prove you wrong

    Test the assumption on a slice: half of new sign-ups, twenty customers, one market, two weeks. Pick something cheap enough to run now and real enough that a miss would change your mind. Put an end date on it.

  4. 04

    Decide, and record why

    The result is met or not met against the criteria. Then a person decides: move forward, change course, drop it or test again. Write down the assumption, the criteria, the evidence, the decision and who made it, so the next person, or agent, starts from it instead of from zero.

Common mistakes

  • Setting the success criteria after you’ve seen the data.
  • Testing the comfortable assumption instead of the one that would sink the idea.
  • Treating “we shipped it” as evidence that it worked.
  • Running the experiment with no end date, so it never forces a decision.
  • Deciding without writing down why, so the same idea comes back in six months.

One worked example

You think the product is underpriced and want to move from $29 to $49 a month. Get it wrong and you either leave money on the table for another year or stall sign-ups when you need them most. Once a new price is public, it is hard to walk back.

Read the full use case →

Decision Recordworked example

  1. assumption

    New self-serve customers will pay $49 a month for the same plan without a big drop in conversion.

  2. experiment

    Show $49 to half of new trial sign-ups for three weeks. The other half sees $29. Same product, same trial.

  3. success criteria · locked

    Trial-to-paid at $49 is no more than 2 points below the $29 rate. Locked before the first sign-up sees the new price.

  4. evidence

    $29: 12% converted. $49: 5%.not met

decision

Change course

Keep $29 for solo users and test $49 as a team plan, where the extra value is easier to see.

Founder decides; SwiftCNS proposes the assumption and drafts the experiment.

Where SwiftCNS fits

You can run this method with a spreadsheet. SwiftCNS runs it in your browser or inside Claude, ChatGPT, Cursor and other AI tools: it finds the assumption and drafts the experiment with you, you approve the success criteria and they lock, and every decision stays on record for your team and your agents.

The method is open. For the full version, read the Innovation Playbook.

Connect your AI tool