Revenue Experiment

The thing that prevents the operator from building 6 half-finished side hustles.

Core rule

Maximum 2 active experiments at any time. If the operator wants a 3rd, force a kill or a ship of an existing one first.

State file

~/Vault/Vantage-OS/experiments.md

Schema:

## Active

### [Experiment name]
- Hypothesis: [If I do X, then Y will happen because Z]
- Investment cap: $___ AND ___ hrs
- Kill criteria: [Specific measurable conditions that end the experiment]
- Success criteria: [Specific measurable conditions that scale it]
- Started: YYYY-MM-DD
- Review date: YYYY-MM-DD
- Status updates:
 - YYYY-MM-DD: [what happened]

## Killed (last 90 days)

### [Experiment name]
- Killed: YYYY-MM-DD
- Reason: [hit kill criteria / ran out of budget / pivoted]
- Lesson: [what the operator learned]

## Scaled

### [Experiment name]
- Scaled: YYYY-MM-DD
- Promoted to venture in ventures.md

Hypothesis template (mandatory)

Every experiment must fit this template before being logged:

If I do [specific action] then [specific measurable outcome] will happen within [timeframe] because [reasoning grounded in evidence, not vibes].

Reject anything that doesn't fit this format. "Try TikTok" is not a hypothesis. "If I post 3 reels per week for 30 days about my rental-fleet ops, then I'll get 200 followers and 5 DMs about my services, because that's roughly what comparable fleet operator accounts averaged in their first month" is a hypothesis.

Investment caps (mandatory)

Every experiment has two caps:

• Money cap: max $ the operator will spend. Default $100 unless he argues up.
• Time cap: max hours of the operator's time. Default 10 hours over the experiment window.

If the operator hits either cap before kill or success criteria are met, force a decision: extend (with new caps), kill, or pivot.

Kill criteria (mandatory)

Pre-commit to what would kill the experiment. Common patterns:

• "If <50 people sign up after $X spent, kill"
• "If conversion rate <2% after 100 visits, kill"
• "If MOQ negotiation can't get under $X, kill"
• "If 0 paying customers after 30 days, kill"

The kill criteria must be set before running. No retroactive moving of goalposts.

Process

Starting a new experiment

1. Force the hypothesis statement.
2. Set investment caps.
3. Set kill and success criteria.
4. Set a review date (default: 14 days out, max 30).
5. Append to experiments.md Active.
6. Add a recurring "review experiment" entry to weekly venture-triage.

Reviewing an experiment

When review date hits:

1. Pull status updates from the file.
2. Compare to kill / success criteria.
3. Force one of three outcomes:

• Kill: move to Killed. Log lesson.
• Continue: only allowed if criteria are partially met AND the operator funds new caps.
• Scale: promote to a real venture in ventures.md. Migrate to that venture's specialist skill/agent set.

Mid-experiment status update

the operator can drop a note ("had this happen today"). Append to status updates with date. Do not re-evaluate yet.

Patterns to flag

• the operator has had 3+ experiments killed for similar reasons (e.g., no audience). Surface the pattern: "the last 3 things you killed all assumed you'd build an audience. That's expensive and slow. Want to bias toward experiments that work without audience?"
• the operator keeps extending an experiment past kill criteria. That's sunk-cost behavior. Call it out explicitly.

Output examples

Logging:

Experiment logged: "Hair product micro-test at a partner's salon"

Hypothesis: If we put a 1-SKU sample basket of [white-label leave-in conditioner] at a partner's checkout for 30 days at $X markup, then ≥20 units will sell and ≥30% of buyers will repurchase or refer, because that's roughly what comparable salon-retail products achieve in concept-fit cases.

Investment cap: $500 and 15 hours
Kill: <10 units sold in 30 days
Success: ≥20 units AND ≥30% repurchase intent → scale to multi-SKU
Review: 2026-06-16

Killing:

Killed: "Hair product micro-test at a partner's salon"
Reason: 6 units sold in 30 days, well below kill criterion of 10.
Lesson: Existing salon clients liked their current products. Need a wedge product that solves a problem they're complaining about, not a parity product. Next experiment should target a specific unmet need at the salon.