| name | expected-value-calculation |
| description | Use when evaluating marketing channels, partnerships, or budget allocation with asymmetric outcomes — especially KOL selection, audit vs rush launch, and any high-cost decision being made on gut feel. Compute EV = (P_success × Gain) − (P_failure × Loss) for each option, rank, enforce portfolio diversification across top +EV options rather than single-bet concentration, and always surface the probability-assumption layer. |
| composition_level | atom |
| extraction-lens | capability |
| source_attribution | Matt Bond (Hivemind Library) |
| license | pending-consent |
| status | candidate |
Expected Value Calculation
When to use
- Choosing between marketing channels or tactics with different cost/risk profiles
- Evaluating a KOL partnership, ad spend, sponsorship, or paid campaign
- Allocating a budget across multiple options
- High-cost / high-risk bets with asymmetric outcomes (e.g., audited launch vs. rush launch)
- Any decision where "gut feel" is currently the ranking method
When NOT to use
- Decision is cheap and reversible — just run it, don't calculate
- Probabilities are genuinely unknowable (use scenarios or red-teaming instead)
- User wants a creative direction, not a quantitative comparison
- Outcome is a single binary win/lose where EV collapses to simple decision theory
Core procedure
1. Identify the options
List 2-5 alternatives being compared. If the user has named only one, ask what they're comparing it against — EV is only meaningful relative to alternatives.
2. For each option, estimate four inputs
- P_success — probability of success (0.0-1.0)
- Gain — potential value if it works (dollars, or user-value equivalent)
- Loss — potential cost if it fails (same units as Gain)
- P_failure = 1 − P_success (derived)
If the user can't estimate probability, walk them through the probability estimation heuristics in reference.md before proceeding.
3. Compute EV
EV = (P_success × Gain) − (P_failure × Loss)
4. Present results as a comparison table
| Option | P_success | Gain | Loss | EV | Verdict |
|--------|-----------|-----------|----------|-----------|---------|
| A | 0.20 | $100,000 | $50,000 | −$20,000 | Avoid |
| B | 0.60 | $100,000 | $5,000 | +$58,000 | +EV |
5. Apply the portfolio rule
Even when one option has the highest EV, 100% concentration is usually suboptimal — it exposes you to that option's P_failure. Recommend diversifying across top 3-4 +EV options at weights roughly proportional to EV, capped at ~40% per option to limit single-point-of-failure risk. See reference.md for the portfolio allocation template.
6. Exclude −EV options by default
Only include a −EV option if there's an explicit strategic reason the numbers don't capture (brand-building, learning, optionality, asymmetric upside). If included, name the reason in the verdict column.
7. Always surface the assumption layer
The calculation is only as good as the probability estimates. Before presenting the final ranking, state:
"EV assumes P_success = X%. If that estimate is off by 2×, the verdict flips from +EV to −EV."
Run the assumption sensitivity test in reference.md before acting on any result.
Output format
Return the comparison table, the recommended allocation (if it's a portfolio decision), and the assumption-sensitivity note. Do not return a single "answer" without showing the working — the user needs to see the probability assumptions to calibrate them.
Supporting files
EXAMPLES.md — three worked examples (channel selection, KOL tier comparison, audit vs. rush launch) plus a portfolio allocation examplereference.md — +EV / −EV action patterns for crypto marketing, portfolio templates, probability estimation priors, sensitivity test checklist
