| name | growth-engine |
| description | Design, run, and interpret growth experiments with funnel evidence and scale discipline. Use when the agent needs to identify a growth bottleneck, prioritize experiments, analyze conversion movement, or decide whether a startup has earned the right to scale a channel or tactic. |
Growth Engine
Purpose
Scale only after finding a real causal improvement. Treat growth as a disciplined experiment system, not a collection of hacks.
Operating Frame
- Start at the bottleneck instead of brainstorming channels blindly.
- Change one important variable at a time when causality matters.
- Verify that gains survive a hold period before scaling spend or headcount.
- Label major claims as evidenced, inferred, or unverified.
Minimum Inputs
- Funnel stages and baseline metrics
- Current bottleneck
- Candidate channels or interventions
- Guardrail metrics and test window
If critical facts are missing, ask for the single highest-leverage unknown or mark the assumption as provisional and continue.
Workflow
- Map the funnel from acquisition to retained value. Identify the current bottleneck and quantify its baseline.
- Build an experiment card for each idea: hypothesis, target metric, guardrail metric, segment, intervention, expected mechanism, timebox, and stop rule.
- Prioritize experiments by expected impact, confidence, and speed to learning. Penalize ideas that create activity without new information.
- Run the smallest credible test. Capture raw results, confounders, and what changed in the environment during the test.
- Classify outcomes as signal, noise, mixed, or broken instrumentation. Do not scale from mixed or noisy results.
- Promote only the experiments that clear both the target metric and the hold-period check, then fold the winners back into the standing growth system.
Required Outputs
- Funnel map
- Ranked experiment backlog
- Experiment cards
- Result readouts with causal confidence
- Scale or stop decisions
Default Output Format
Use this section order unless the user asks for a spreadsheet or experiment tracker:
- Funnel map and bottleneck
- Baseline metrics
- Ranked experiment backlog
- Experiment cards
- Result interpretation
- Scale or stop decisions
- Recommended next skill
Verification Gate
Do not recommend scaling because a chart moved once. Require baseline comparison, guardrail review, and a believable mechanism for why the lift happened.
Handoff
- Use
$gtm-launch when a growth lesson should become part of a broader launch motion.
- Use
$fundraising-kit after the growth system produces defensible traction proof.