froggo-mission-control: Skill for Claude Code

Growth Experiment

Purpose

Design rigorous growth experiments that produce actionable, statistically valid results. Every experiment must have a clear hypothesis, defined success criteria, and documented learnings before it runs — not after. Guessing is not a growth strategy.

Trigger Conditions

Load this skill when:

• Designing a new A/B test, multivariate test, or growth hypothesis
• Evaluating whether an experiment has reached statistical significance
• Documenting the results of a completed experiment
• Reviewing an experiment backlog to prioritize what to run next
• Auditing an in-flight experiment for methodology problems

Procedure

Step 1 — Write the Hypothesis

Every experiment starts with a formal hypothesis in this structure:

If we [specific change or intervention],
then [measurable outcome],
because [underlying mechanism or user behavior we believe is true].

Examples of strong hypotheses:

• "If we add a one-click referral prompt on the post-trade confirmation screen, then referral sign-ups will increase by 15%, because users are most motivated to share immediately after a positive transaction."
• "If we reduce the onboarding flow from 7 steps to 4 steps, then D1 activation rate will increase by 10%, because friction is the primary drop-off driver based on funnel data."

Anti-patterns to reject:

• "We think the button color should be green." (No measurable outcome, no mechanism)
• "Let's try making the CTA bigger." (No hypothesis, no success criteria)

Step 2 — Define the Primary Metric

Choose exactly one primary metric. This is the single number that determines whether the experiment wins or loses.

• Must be measurable before the experiment runs
• Must be attributable to the change being tested (not confounded by other changes)
• Must be collected at the user level, not aggregate

Secondary metrics (guardrail metrics) are allowed but do not determine win/loss. They exist to catch regressions.

Primary metric: [metric name] — [how it is measured] — [current baseline]
Guardrail metrics:
  - [metric 1] — must not decrease by more than X%
  - [metric 2] — must not decrease by more than X%