Skip to content

Multi-Criteria Decision Matrix

An analytical selection model — instantiates Divergence-Convergence Cycle Orchestration

Scores every shortlisted option against every weighted criterion in one matrix and sums to a ranking — making the trade-offs, the weights, and who owns them explicit and auditable rather than intuitive.

When a shortlist survives to the point where the choice turns on trade-offs — cheaper but slower, faster but riskier — Multi-Criteria Decision Matrix is the tool that makes those trade-offs explicit instead of adjudicating them by feel. Options run down the rows, weighted criteria across the columns; each cell scores an option on a criterion, and the weighted sum gives a ranking. Its defining feature is simultaneous, weighted, transparent comparison: every option is judged against every criterion at once, the relative importance of each criterion is stated as a number rather than smuggled in, and the finished matrix is a record anyone can inspect and challenge. It does not generate options and it is not a vote — it is the analytical convergence step that prices a shortlist against declared priorities.

Example

A logistics company must choose a site for a new distribution center from four finalists. The decision keeps oscillating in meetings because different people weight different things. They build a decision matrix. Criteria and weights (agreed before scoring, by the stakeholders who own each): land-and-build cost ×0.30 (finance), transit-time-to-market ×0.30 (operations), labor availability ×0.20 (HR), expansion headroom ×0.10 (strategy), incentive package ×0.10 (finance). Each site is scored 1–5 on every criterion, cells are weighted and summed.

Site B, the crowd favorite for its low sticker price, lands second once transit time is properly weighted; Site C wins on the balanced total. The matrix's real payoff isn't the winner but the record: it shows exactly which weighted criteria decided it, so a sensitivity check — "does C still win if we cut the transit weight to 0.20?" — can be run in minutes, and the rationale survives the meeting as an auditable artifact rather than a remembered argument.

How it works

What distinguishes it from a vote or a pairwise method:

  • Fix criteria and weights first. The columns and their weights are agreed before options are scored, so priorities can't be reverse-engineered to fit a favorite.
  • Score every option on every criterion. The full option × criterion grid is filled, giving a simultaneous comparison rather than a sequence of duels.
  • Weight and sum to a ranking. Weighted cell scores are totaled per option; the ranking falls out of the arithmetic, not the room's mood.
  • Keep the matrix as the record. The completed grid is the commitment record — it shows what was valued, how much, and why the winner won, ready for a sensitivity check.

Tuning parameters

  • Criteria weights — the relative importance assigned to each column, and the highest-leverage dial in the tool. Small weight changes can flip the winner, which is exactly why weights must be set before scoring and stress-tested after.
  • Scoring scale — coarse (1–3) versus fine (1–10) and whether scales are anchored with descriptions. Finer scales feel precise but invite false discrimination between near-ties.
  • Weight ownership — who sets each criterion's weight. Assigning weights to the stakeholders accountable for that dimension makes the priorities legitimate; letting one person set all of them collapses the multi-criteria premise.
  • Normalization — how raw scores across differently-scaled criteria are made comparable. The choice quietly affects results and should be stated, not defaulted.

When it helps, and when it misleads

Its strength is making a trade-off decision legible: it forces priorities into the open as weights, lets many criteria and stakeholders be reconciled in one view, and leaves an auditable trail that a later challenge or sensitivity analysis can interrogate. Descended from weighted decision matrices and Pugh's selection method, it is the workhorse for a bounded shortlist where the criteria are known.[1]

Its failure modes are precision and manipulation. The numbers look objective but encode subjective weights and scores, so it invites false precision — a decisive-looking total resting on arguable inputs. It is the mechanism most easily run backwards: set the weights after glancing at the options and the matrix becomes a machine for laundering a decision already made. It also flattens genuinely incommensurable values into one scale and can miss what resists scoring. The discipline that keeps it honest is fixing weights before scoring, assigning weight-ownership to the accountable stakeholders, and running a sensitivity analysis on the weights — if the winner flips under a plausible weight change, the matrix has told you the decision is close, not that its top number is truth.

How it implements the components

It owns the analytical-convergence machinery — turning a shortlist into a ranked, recorded choice:

  • convergence_criterion_set — the weighted columns are the explicit, agreed criteria the selection runs on.
  • option_comparison_frame — the option × criterion grid is the comparison itself: every option against every criterion, simultaneously and on the record.
  • convergence_commitment_record — the completed matrix is a durable artifact of what was valued and why the winner won, ready for audit and sensitivity checks.
  • stakeholder_dimension_assignment — criteria weights are owned by the stakeholders accountable for each dimension, tying priorities to the people responsible for them.

It does not generate the options it ranks (candidate_option_inventory, exploration_dimension_map — see Morphological Matrix, then Scoring and Assumption Reversal and Recombination), nor does it run the fast participatory shortlisting of Dot Voting with Evidence Gate.

References

[1] The weighted decision matrix and the related Pugh selection matrix (Stuart Pugh's controlled-convergence method) are the established lineage of this tool; sensitivity analysis on the criterion weights is the standard corrective for the false-precision and reverse-justification failures noted above. ↩