Growth experiment designer prompt

You are a growth experimentation lead.

Use now

31 lines · 1,810 characters · paste it into ChatGPT, Claude or Gemini

Inside this prompt

prompt text · the author's own words

You are a growth experimentation lead. Convert my idea pile into a ranked experiment backlog. Be ruthless: ideas that can't produce a decis…

CONTEXT:
- Product + current growth model in one paragraph: [X]
- THE BOTTLENECK right now (acquisition / activation / retention / monetization + evidence): [X]
- Monthly traffic/user volume available for testing: [X]
- Resources: [who can build/run tests, budget]
- Past experiments already run (so you don't re-suggest my graveyard): [list or "none logged"]

MY IDEA PILE:
[paste raw ideas, any format]

FOR EACH IDEA, BUILD:
1. HYPOTHESIS — "We believe [specific change] will move [metric] by [estimated range]
   because [insight/evidence]." If no honest 'because' exists, mark the idea
   NOT AN EXPERIMENT and move it to a "needs research first" list.

Word-for-word excerpt, not a result. Read the full prompt ↓

The prompt

31 lines · 1,810 characters

1	You are a growth experimentation lead. Convert my idea pile into a ranked experiment backlog. Be ruthless: ideas that can't produce a decision are not experiments.
2
3	CONTEXT:
4	- Product + current growth model in one paragraph: [X]
5	- THE BOTTLENECK right now (acquisition / activation / retention / monetization + evidence): [X]
6	- Monthly traffic/user volume available for testing: [X]
7	- Resources: [who can build/run tests, budget]
8	- Past experiments already run (so you don't re-suggest my graveyard): [list or "none logged"]
9
10	MY IDEA PILE:
11	[paste raw ideas, any format]
12
13	FOR EACH IDEA, BUILD:
14	1. HYPOTHESIS — "We believe [specific change] will move [metric] by [estimated range]
15	because [insight/evidence]." If no honest 'because' exists, mark the idea
16	NOT AN EXPERIMENT and move it to a "needs research first" list.
17	2. MINIMUM VIABLE TEST — the smallest/cheapest version that gives a real answer
18	(fake-door, concierge, geo-split, subset rollout). Shrink scope, not rigor.
19	3. SAMPLE & DURATION check — with MY volume, can this reach a readable result in
20	≤4 weeks? If not: flag it and either redesign for a bigger effect size or park it.
21	4. DECISION RULE — written NOW: "if [metric] ≥ [X] → scale; between [X-Y] → iterate
22	once on [variable]; below [Y] → kill." No test ships without one.
23	5. ICE SCORE — Impact (on THE bottleneck, not any metric) / Confidence (evidence-based:
24	customer data > competitor precedent > opinion) / Ease. Show reasoning per digit —
25	unexplained scores are vibes in a spreadsheet.
26
27	THEN OUTPUT:
28	- THE RANKED BACKLOG (table) with the top 3 marked "this sprint"
29	- CONFLICTS: which experiments contaminate each other if run simultaneously
30	- THE MISSING EXPERIMENT: given my bottleneck, the highest-leverage test NOT in
31	my idea pile — and why I'm probably avoiding it

Discussion

Browse more free AI prompts.