Dimensioned Comparison Matrix¶
Comparison artifact — instantiates Dimensioned Comparison Framing
Lays comparands out as rows and dimensions as columns, scores every cell on a common scale, and reads a ranking or dominance relation off the completed grid.
The Dimensioned Comparison Matrix is the archetype's central working artifact: a grid whose rows are the comparands, whose columns are the chosen dimensions, and whose every cell holds a value measured and rescaled so that all columns can be read together. Its defining move is that it forces completeness — each comparand must be scored on each dimension on a commensurable scale, which is precisely what a side-by-side glance quietly skips. Where a checklist asks whether items are legitimately comparable and an annotation reports what the result means, the matrix is the machine that turns heterogeneous specs into one grid you can rank, Pareto-filter, or scan for dominance. It is the place where "which is better?" stops being a hunch and becomes a computation whose every input is visible.
Example¶
A logistics manager must pick among four electric delivery vans for a city fleet. The vendor sheets are useless side by side: range is in miles, payload in kilograms, charge time in hours, total cost of ownership in dollars-per-mile, and the service network in "number of certified shops." She builds a matrix — four van-rows, five dimension-columns — and drops each raw spec into its cell. Then she rescales every column to a 0–1 band so a range advantage and a cost advantage can be summed rather than compared in incompatible units, applies weights reflecting the route profile, and reads a ranking off the totals.
The grid earns its keep at the moment it contradicts the pitch. Van A tops the weighted score, but the matrix shows why: an unusually low cost-per-mile figure is doing almost all the work, and Van A's payload cell sits below the fleet's hard minimum. She flags payload as a non-negotiable floor rather than a tradeable column, re-reads the grid as dominance-with-a-veto, and Van C — second on the naive total but the only one clearing every floor — becomes the defensible choice. The matrix didn't pick for her; it made the basis of the pick auditable in one screen.
How it works¶
- Populate rows and columns. Comparands become rows, the selected dimensions become columns; the intersection is the unit of work — one measured value per cell.
- Score under a scale protocol. Each cell is filled by the same measurement rule per column (same instrument, same window, same missing-cell convention), so cells within a column are mutually comparable before anything is combined.
- Commensurate the columns. Raw units are rescaled — min–max to a 0–1 band, z-scores, or ordinal ranks — so that a step on one dimension is comparable to a step on another. This is the step that makes cross-column reading legitimate.
- Read a relation off the grid. Apply the chosen readout: a weighted sum yielding a rank, a Pareto sweep yielding a non-dominated set, or a dominance check with vetoes for non-tradeable floors.
Tuning parameters¶
- Normalization method — min–max vs. z-score vs. rank-transform. Min–max preserves spread but is hostage to outliers; ranks are robust but discard magnitude. The choice can reorder the winner, so it belongs on the record, not in a footnote.
- Aggregation rule — compensatory (weighted sum, strengths offset weaknesses) vs. non-compensatory (dominance, veto floors). Compensatory reads cleanly but can bury a fatal weakness; non-compensatory protects hard constraints but may return "no clear winner."
- Cell granularity — a single scored value vs. a value-plus-uncertainty in each cell. Finer cells resist false precision but slow the build.
- Show-both toggle — display raw units alongside normalized scores, or normalized only. Showing both lets a reader audit the rescaling that produced the rank.
When it helps, and when it misleads¶
Its strength is consolidation: every comparand, every dimension, and every rescaling decision live in one auditable object, so a disputed ranking can be traced to the exact cell or weight that drives it. It is the artifact reviewers can actually argue over.
Its central failure mode is the compensatory illusion — a weighted sum lets a towering score in one column silently compensate for a disqualifying score in another, so the grid crowns an option that should have been vetoed.[n1] The tidy normalized decimals also invite false precision over inputs that were really guesses, and the matrix is easily run backwards: the weights and normalization tuned until the pre-chosen favorite tops the column. The guarding discipline is to mark non-compensatory dimensions as vetoes rather than tradeable columns, and to re-read the ranking under at least one alternative normalization before trusting it — a job the Dimension Weight Sensitivity Panel exists to do.
How it implements the components¶
The matrix realizes the scoring core of the archetype — the components that turn a justified set of items into a computed relation:
comparand_set— the rows; each item enters as one row scored across all columns.comparison_dimension_set— the columns; the chosen dimensions become the axes the grid is built on.measurement_and_scale_protocol— the per-column rule that fills each cell with a like-measured value.normalization_or_commensuration_rule— the column rescaling that makes cross-dimension reading legitimate; the matrix's signature step.relation_readout_rule— the rule (weighted sum, Pareto, dominance-with-veto) that converts the completed grid into a rank or a non-dominated set.
It does not settle the comparison question, the shared frame, or which items legitimately belong — that framing is done by Comparison Basis Checklist and Comparator Set Audit — nor does it test the grid's stability under alternative weights, which is the Dimension Weight Sensitivity Panel's job.
Related¶
- Instantiates: Dimensioned Comparison Framing — the matrix is the concrete grid the archetype's scoring machinery produces.
- Consumes: Comparison Basis Checklist supplies the question and frame; Comparator Set Audit supplies the justified rows.
- Sibling mechanisms: Comparison Basis Checklist · Pairwise Comparison Protocol · Comparator Set Audit · Dimension Weight Sensitivity Panel · Matched Case Comparison Sheet · Comparison Readout Annotation · Counterbalanced Comparison Display
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: The mechanism measures comparands under consistent column protocols, rescales dimensions to a common basis, and reads ranking or dominance from the completed grid, so its operative form is multidimensional comparison analysis.
Nearest alternative: Representation, Specification & Plan — The matrix persists cells and scores, but commensuration and relation inference rather than the table artifact are the defining work.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Operations Research
Origin pattern: Single lineage
Present-day reach: Multi-domain
Rationale: Multi-criteria decision analysis established option-by-criterion matrices, normalization, weighting, dominance checks, and transparent ranking.
Review resolution: Multi-criteria decision analysis established option-by-criterion matrices, normalization, weighting, dominance checks, and transparent ranking. The scored comparison matrix cohered in operations research; organizational use is applicability, and the method is broadly portable without being universal.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] A compensatory decision model (the family that includes weighted-additive scoring and multi-attribute utility theory) assumes a deficit on one attribute can always be offset by surplus on another. That assumption fails whenever a dimension is really a hard constraint — a payload floor, a safety minimum — which is why such dimensions must be modeled as vetoes rather than columns in the sum. ↩