🎯 Comparison and decision
The commands in family 05: criteria before options, justified rankings, trade-off matrices, and bets with an exit condition.
📏 /criteria + /weights — define how to choose
The golden rule of family 05: the criteria and their weights exist before for the options to be evaluated. A yardstick set afterward is designed to favor the preferred measure.
📐 Anatomy of a good criterion
- •Measurable: “time to first paying customer,” not “agility.”
- •Weighted: impact 40%, cost 30%, risk 30% — stated weights, not implied.
- •Independent: criteria that duplicate the same factor inflate the score without adding information.
New here? Weight is the relative value of each criterion in the final score — the numerical translation of "this matters more to us." Undeclared weights don’t disappear: they become invisible bias.
🎯 Copy and run
/criteria /weights para escolher; proponha critérios mensuráveis e pesos com justificativa
How to verify: you agree with the weights after reading them on your own — if they surprise you, discuss them now, not later.
🥇 /rank — rank alternatives with justification
Sorts options by criteria and explains each position. A ranking without justification is a numbered preference; with parameters, it becomes an auditable team standard.
🎯 Copy and run
/rank criteria=impacto,custo,risco entre, e ; justifique cada posição
How to verify: swap two items and see whether the rationale still holds together — a fragile ranking falls apart under the test.
💡 Practical tip
Ask for sensitivity: "if the risk weight doubles, who wins?" A ranking that changes completely with small weight variations is a technical tie in disguise.
🔢 /score — score on an explicit scale
Scores by criterion in scale with anchors (“7 = we already have a pilot customer in this range”). Without an anchor, each evaluator uses a different yardstick, and the average becomes noise.
Set anchors for each level
Describe what scores of 3, 6, and 9 mean for each criterion—concrete examples, not adjectives.
Show the composition
Always show the total score alongside the partial scores — a win based on one dominant criterion should be visible.
🎯 Copy and run
/score as opções , ,em escala 1-10 por criterio com âncoras; mostre composição da nota final
How to verify: you can reconstruct the decision by looking only at the table — without reading the prose.
⚖️ /compare — compare without deciding too soon
Produces a side-by-side comparison by dimension, without forcing a verdict. Separating comparison from decision preserves the window in which alternatives can still surprise you.
✓ Open comparison
- ✓Same dimensions for all options—including those that favor the underdog.
- ✓"Differences that matter" highlighted separately from cosmetic differences.
✗ Premature conclusion
- ✗A comparison written to confirm the option the boss has already chosen.
- ✗Different dimensions for each option—a table that hides what’s uncomfortable.
🎯 Copy and run
/comparevs nas dimensões: custo, curva de aprendizado, lock-in e suporte. Só compare, não decida
How to verify: symmetric table, same set of dimensions, no recommendation embedded in the text.
🏁 /decide — make decisions with evidence and reversibility
Close the process with a single recommendation, evidence, and—crucially—the degree of reversibility: how much would it cost to reverse course if you’re wrong?
Illustrative: the first question for any decision is "which door am I going through?"
💡 Practical tip
The family’s complete chain: /chain(/criteria -> /rank -> /decide). The decide inherits the criteria set at the start—criteria that can’t be adjusted retroactively to favor the preferred option.
🎯 Copy and run
/decide entre , eusando os critérios definidos acima; declare reversibilidade e custo de mudar de ideia
How to verify: one recommendation (no “it depends”), with cited evidence and a stated plan B.
🔲 /tradeoffmatrix — compare alternatives × criteria
The complete matrix: each alternative assessed against each criterion, with weights applied and every cell justified. Well-crafted narratives fall apart in the face of a well-built matrix.
| Criterion (weight) | Option A | Option B | Option C |
|---|---|---|---|
| Impact (40%) | 8 · broad reach | 6 · medium term | 9 · new market |
| Cost (30%) | 5 · new team | 8 · reuses the team | 3 · high capex |
| Risk (30%) | 7 · proven technology | 8 · single dependency | 4 · regulatory uncertainty |
Teaching example: each cell = rating + reason on one line. The matrix shows that C wins on impact but bleeds on cost and risk.
🎯 Copy and run
/tradeoffmatrix entre; linhas=critérios com pesos, colunas=opções, cada célula com nota e motivo
How to verify: no empty cells and no generic "great" — every rating has a verifiable reason.
📈 /pareto — find the dominant frontier
Eliminates the dominated options — those that another option matches or beats on every criterion—and the frontier where improving one criterion requires sacrificing another.
New here? In Pareto frontier there’s no free lunch: to gain on one axis, you pay on another. A dominated option is one that loses to another option on EVERY axis — it should never have made it to the decision table.
🧹 What the command eliminates
- •Your most expensive and slowest internal competitor compared with another proposal of yours.
- •The tool that covers the other one in both price AND functionality.
- •The project someone else delivers just as well, sooner, and with fewer people.
🎯 Copy and run
/pareto sobre as opções listadas; elimine dominadas, mostre a fronteira e diga o trade-off de cada sobrevivente
How to verify: shortened list with justification based on the eliminated option's dominance — not preference.
🎲 /bet + /stop — place a bet and know when to stop
Every big bet raises three questions: how much could it yield (upside), how much it might cost (downside) e when we stop? /bet structures the first two; /stop locks the third before emotion kicks in.
✓ A grown-up bet
- ✓"If we don’t have 10 paying customers in 90 days, we’ll shut down the line."
- ✓Objective trigger, decided when no one is emotionally invested.
✗ Zombie project
- ✗"Let’s give it one more quarter" — the fourth in a row, with no new criteria.
- ✗A stopping criterion that shifts with the results (“when it’s good enough”).
🎯 Copy and run
/bet /stop para apostar em; upside, downside, exposição máxima e kill criteria objetivos com data
How to verify: kill criteria that a stranger could judge without emotional context — number + date.
⚠️ The stranger test
Show your stop criterion to someone uninvolved: looking at the numbers objectively, could they tell whether the time has come? If they need "context," the criterion is just for show.
Quick self-check (optional): why define criteria BEFORE seeing the options?
🎓 Module Summary
Next Module:
4.1 - Planning and architecture — turn the decision you made into an actionable plan.