Science-backed
team decisions.

You define alternatives (the options being compared) and criteria (the considerations that matter).

Each criterion carries a priority (0–10, 0 = “doesn’t apply to me”) and, for every alternative, a performance range (1–10, min–max, optionally with a triangular peak) that represents uncertainty.

A priority is a relative weight. In every trial the engine combines criteria into an outcome, and the weight scales how much a criterion steers that outcome. Only the ratios between priorities matter: with {5, 7, 9}, the 9 outweighs the 7 by 9/7 ≈ 1.29× and the 5 by 1.8×. Rescaling all priorities together (e.g. to percentages that sum to 100) does not change the winner.

In team runs, every member’s importance ratings are first made relative to their own top criterion: their #1 becomes the reference (1.0) and their other ratings become fractions of it. So a member who tops out at 7 and one who tops out at 10 contribute equal top-weight — nobody’s personal number scale silently dominates the average.

A 0 means “does not apply to me”: it is excluded for that member and never drags the group weight down. If more than half the members mark a criterion 0, that criterion is dropped from the run entirely (the strongest criterion is always kept so the run stays meaningful).

The members’ normalized weights are averaged and mapped to a 0–10 priority for the engine. Dispersion is measured on the same normalized values, so disagreement reflects real differences in preference — not differences in how big a number each person happens to write.

DCZion applies Condorcet's rule to the winner check, not to ballot collection. In a team run, members' ratings are first averaged into a single aggregated model — individual members never go head-to-head against each other.

On that aggregated model, every alternative is compared pairwise against every rival: the option that beats all others head-to-head is the Condorcet winner; if preferences cycle, the option with the most pairwise wins takes the recommendation. Method B (criteria-bloc) is the closest to classic Condorcet voting — each criterion casts a vote as a bloc sized by its priority.

Each option accumulates a weighted score per simulation trial. Then, within every trial, options are ordered by score, and the winner of each head-to-head pair gets a tally. After thousands of trials the pairs become a probability: “A beats B in 94% of runs.” The option that beats every other one over 50% is the Condorcet winner; if none (a score-based cycle), the option with the most pairwise wins wins. Uses performance intensity, uncertainty ranges, and the optional downside-risk penalty.

This reconstructs classic Condorcet voting. A priority distribution like {19%, 19%, 27%, 35%} is treated the way one treats voter blocs: 19% of the decision weight prefers one ordering, 35% prefers another. Criteria genuinely disagree (cost loves C, quality loves B), so real majority conflict — and genuine cycles — can emerge, and a majority of weight resolves them. This is the fairest “whose priorities win” reading, but it ignores performance intensity and near-ties.

Method A = richer decision model: uses magnitude, uncertainty, and risk — but is not majority voting, and its priorities are compressed.

Method B = faithful voting model: proportional to priority spread, surfaces true disagreement — but discards intensity and near-tie detail.

They answer different questions. Method A asks “what is the best weighted recommendation?”, Method B asks “which option wins a fair contest of the team’s preferences?”

Method A — weighted score (default). Use when:

• Magnitude is the point — “how much better” matters: cost savings, revenue, TCO, impact size.

• Your criteria are quantifiable, with real numbers on cost, time, or risk.

• The decision is hard to undo — only A applies the downside-risk penalty.

• You want one stable, defensible recommendation with win odds.

Weaknesses: the weighted average can hide genuine preference conflict, and a mediocre option nobody loves can still win on summed scores.

Method B — criteria-bloc Condorcet. Use when:

• Criteria are competing factions — cost vs quality vs speed — and the real question is “whose priorities win.”

• You expect the team to fight: B surfaces genuine cycles and disagreement instead of averaging them away.

• You want robustness to score-gaming: only the ordering matters, so inflated magnitudes can’t buy a win.

Weaknesses: discards intensity (slightly cheaper counts the same as half the price), has no risk penalty for irreversible decisions, and with few criteria a tiny “electorate” can produce noisy cycles.

Quick heuristic: quantified criteria (cost, time, revenue) → Method A. Value-judgment criteria held by different factions (strategy, culture, risk appetite) → Method B. Hard to undo → Method A. Expect a fight → run B to expose the conflict, then A to quantify the tradeoff.

Performance ranges are sampled thousands of times; each sample set is one trial. This turns confidence into probabilities and makes results reproducible with a fixed seed.

Method A samples a shared weighted score; Method B re-ranks each criterion’s bloc per trial, so wide ranges weaken conviction without inventing certainty.