Phase 3a - Validation

How to read this page
When to enter this phase
Step 1 - Design Thinking
Step 2 - Hypothesis Register (Plan mode)
Step 3 - Run experiments (you, not AI)
Step 4 - Hypothesis Register (Results mode)
What you exit with
Phase exit gate

Design and run experiments to validate (or invalidate) your riskiest assumptions. The phase ends with a hard Go/No-Go verdict.

Duration: Externally paced - tempo is set by the market, not by AI · Gate type: Hard gate - only GO advances · Playbooks: Greenfield

How to read this page

This phase cannot be accelerated by AI - the signal comes from the real world on its own schedule. AI's job here is designing the experiment and structuring the hypothesis, not producing the verdict.

When to enter this phase

Enter after Phase 2 (or when a Problem Validation Summary exists). Can start in parallel with late Phase 2.

"Done elsewhere" rule: if Design Thinking and hypothesis validation were done outside the framework, import the results directly (Go/No-Go verdict, evidence per hypothesis, riskiest assumptions) - Phase 3a is then complete, you don't re-run the experiments.

What you need before entering: Problem Validation Summary (Phase 2 exit artifact), and willingness to run real-world experiments (landing pages, smoke tests, pre-orders, concierge MVP).

Step 1 - Design Thinking

When to run / skip: always run, unless Design Thinking was already done outside the framework (see "done elsewhere" above).
Gather first: Problem Validation Summary.
Command: /design-thinking
What you get: Problem Statement, POV, HMW ideation synthesis, Elevator Pitch, draft Validation Hypotheses.
What it does NOT give you: a formal, testable hypothesis register - that's Step 2. This is ideation, not yet structured for experiments.
Done when: you have draft hypotheses ready to formalize.

Step 2 - Hypothesis Register (Plan mode)

When to run / skip: always run after Step 1.
Gather first: Validation Hypotheses draft from Step 1.
Command: /pm-hypotheses (Plan mode)
What you get: a formal register - ICP, assumption map, experiment plan, success criteria, all defined before experiments run (so you can't move the goalposts after seeing results).
What it does NOT give you: the results - that's Step 4, after you've actually run something.
Done when: every hypothesis has a pre-defined success criterion.

Step 3 - Run experiments (you, not AI)

When to run / skip: always. This is the step AI cannot do for you.
Gather first: the experiment plan from Step 2.
Activity: landing page, smoke test, pre-order campaign, concierge MVP - whatever the plan specified.
What you get: real-world evidence: did people show up, convert, pre-order, confirm the pain.
What it does NOT give you: a verdict yet - that requires structuring against the criteria (Step 4).
Done when: you have real data against every hypothesis's success criterion, or a clear reason a hypothesis couldn't be tested.

Step 4 - Hypothesis Register (Results mode)

When to run / skip: always run after experiments conclude.
Gather first: experiment results from Step 3.
Command: /pm-hypotheses (Results mode)
What you get: results recorded against the pre-defined criteria, and the verdict:
- GO → advance to Phase 3b
- PIVOT → loop to targeted re-validation with refined hypotheses
- STOP → terminate the project
What it does NOT give you: a way around a bad result. There is no FORCE bypass - the verdict must be genuine, based on what the experiments actually showed.
Done when: the verdict is recorded. GO is the only one that unlocks Phase 3b.

What you exit with

Hypothesis Register - ICP, assumption map, experiment plan, success criteria
Experiment results - evidence collected from real-world validation
Go/No-Go verdict - the Phase 3a exit decision

Phase exit gate

Gate type: Hard. Only GO advances to Phase 3b. PIVOT loops to targeted re-validation. STOP ends the project. There is no FORCE bypass.