QUICK ANSWER
What should a decision matrix template include?
A decision matrix template should freeze one question, option set, criteria, exactly 10,000 basis-point weights, and a score scale before scoring. Each score needs option-specific evidence and confidence. Keep non-negotiable gates outside the weighted total, block totals when evidence is unknown, show ties and sensitivity, date the review, and send the final decision to a named human authority.
A decision matrix is useful when one bounded option set must be compared against the same frozen criteria. It becomes misleading when weights drift, missing evidence receives a default score, a failed requirement disappears inside a total, or the highest number is presented as an automatic choice.
This downloadable pack keeps those boundaries visible. It includes a fictional completed example, an empty starter, Markdown and CSV projections, a closed JSON Schema, and dependency-free validation. The matrix produces evidence-bound analysis only. A separate named human authority owns the actual decision.
How to use a weighted decision matrix without automating the decision
Treat the matrix as a dated comparison model. Freeze its boundary first, connect every score to reviewable evidence, calculate once with a declared formula, and expose the cases where the ranking is incomplete, ineligible, tied, or sensitive.
Freeze the decision question and option set
State one bounded question, scope, assumptions, constraints, review horizon, and option set. Include the status quo when it is a real alternative. Do not add or remove options after scoring without opening a new revision, because a changed choice set changes the analysis boundary.
Sources: [matrix-pack], [nasa-decision-analysis], [gao-alternatives-analysis]
Define criteria, evidence expectations, and exact weights
Write criteria that distinguish the options, say what evidence would support each score, and assign positive integer basis-point weights totaling exactly 10,000. The pack uses one fixed 1-to-5 scale and local weights. Neither the criteria nor their weights are universal defaults.
Sources: [matrix-pack], [asq-decision-matrix], [nasa-decision-analysis]
Record option-specific evidence before scores
Give every evidence item a stable ID, option, criterion or gate, status, confidence, reserved reference, checked time, and summary. A supported score must cite available evidence. Missing or disputed evidence stays unknown; it must not become zero, average, or another convenient substitute.
Sources: [matrix-pack], [rfc2606]
Keep non-negotiable gates separate from totals
Evaluate each option against the same named pass, fail, or unknown gates. A failed gate makes an option ineligible and an unknown gate blocks eligibility, but neither rewrites the weighted score. This keeps a strong total from concealing a failed mandatory condition.
Sources: [matrix-pack], [nasa-decision-analysis]
Reconcile totals, ranks, ties, and sensitivity
Normalize supported integer scores, sum the weighted numerator, and round once after the sum. Rank only complete, gate-eligible options. Show exact ties, label locally defined near ties, and rerun explicit weight changes to reveal whether the order is stable, reversed, or tied.
Sources: [matrix-pack], [asq-decision-matrix], [gao-alternatives-analysis]
Review, expire, export safely, and hand off authority
Record a reviewer, revision, change trail, review time, expiry, gaps, and limitations. Treat CSV as an interchange projection and protect downstream spreadsheet workflows from formula-like cells. Keep accessibility review separate, then link the analysis to a distinct human decision record.
Sources: [matrix-pack], [owasp-csv-injection], [wcag22], [json-schema-2020-12]
The decision-matrix boundary
This page owns a generic, transparent, multi-criteria comparison for one bounded option set. Adjacent formats keep their own questions, evidence, and authority.
Included
- Frozen question, scope, assumptions, constraints, option set, criteria, review horizon, and revision IDs
- Exactly 10,000 integer basis-point weights, a fixed integer scale, one declared formula, and one rounding point
- Option-specific evidence status, confidence, checked time, reserved references, gaps, and review expiry
- Separate non-negotiable pass, fail, or unknown gates plus complete and eligible ranking rules
- Exact ties, local near ties, explicit sensitivity cases, review records, change history, and a separate human decision reference
Not included
- Software-specific build, buy, or hybrid requirements, total cost of ownership, migration, and exit analysis; use the build-versus-buy owner
- Investment rationale, benefits, cost-benefit analysis, funding request, and recommendation; use the business-case owner
- Branching conditional logic or diagnostic paths; use a decision tree
- Final rationale, approval, and outcome history; use a decision log owned by the authorized decision-maker
- Vendor or product rankings, market claims, procurement selection, legal review, regulatory interpretation, compliance, certification, or safety decisions
- Eisenhower quadrants, personal task or time management, backlog assignment, roadmap commitment, universal weights, universal thresholds, or automatic priority
DOWNLOADABLE RESOURCE
Download the decision matrix template pack
Start with the Markdown starter for a facilitated review, use CSV for spreadsheet import, and use canonical JSON plus the closed schema when a tool needs a bounded contract. The example demonstrates complete, incomplete, failed-gate, near-tie, reversal, and exact-tie states.
Decision matrix template pack
A fictional four-option comparison with five weighted criteria, two non-negotiable gates, evidence confidence, reconciled results, three sensitivity tests, review history, and a separate pending human decision reference.
Format: Markdown, CSV, JSON, JSON Schema, and dependency-free Node.js tests in one ZIP archive
Locally reproduced August 1, 2026. SHA-256: cc6df7d59a07b6fca9245987b5c507dbb8dd58e035361f2a501c3f762215f9ad
Included
- Editable Markdown template and blank two-option JSON starter
- Canonical four-option JSON example plus human-readable Markdown projection
- Criteria, options, and scores CSV projections that open in spreadsheet applications
- Closed Draft 2020-12 JSON Schema for structural validation
- Dependency-free validator, 65 positive and negative tests, deterministic builder, README, and allowlisted ZIP
Verification boundary
The archive allowlist, deterministic bytes, schema closure, stable IDs, UTC chronology, exact 10,000 bp weights, score normalization, one-point rounding, evidence references, gate eligibility, incomplete totals, rank, ties, near ties, sensitivity recomputation, authority, reserved URLs, privacy checks, and CSV parity were checked locally.
FICTIONAL REVIEW SNAPSHOT
One total is not the whole result
Fixed 1-to-5 scale. Weights total exactly 10,000 bp. Analysis only.
| Criterion | Weight | Alder | Birch |
|---|---|---|---|
| Evidence strength | 2,800 bp | 4 | 5 |
| Need fit | 2,600 bp | 4 | 3 |
| Change affordability | 1,800 bp | 3 | 4 |
| Delivery confidence | 1,600 bp | 4 | 4 |
| Accessibility review readiness | 1,200 bp | 3 | 3 |
Option Alder
7,050 bp
Rank 2
Option Birch
7,150 bp
Rank 1
Option Cedar
Blocked
Evidence unknown
Option Dune
9,300 bp
Gate failed, no rank
SENSITIVITY
Test the order, not just the total
- Evidence +500 bpOrder staysGap widens
- Need fit +500 bpOrder reversesModel is sensitive
- Need fit +400 bpExact tieNo tie-break
Priority matrix template: rank options without hiding uncertainty
The same artifact works as a priority matrix when the options are bounded initiatives or proposals and the criteria are local to one review. It still must preserve evidence gaps, mandatory gates, ties, sensitivity, and human authority.
Completed comparison with a near tie
Use when: Two or more options have complete evidence and pass every local non-negotiable gate, but their reconciled totals are close.
Fictional Options Alder and Birch receive 7050 and 7150 basis points. Both remain visible, their 100 bp gap is labeled a near tie under this revision, and no winner is declared.
Structure
- Exact option, criterion, weight, score, evidence, total, and eligibility references
- A local 250 bp near-tie threshold that does not create a universal tie-break rule
Watch for: A narrow numeric gap may be fragile, but the chosen threshold does not prove the options are equivalent or tell a human which one to select.
Sources: [matrix-pack], [nasa-decision-analysis]
Bounded initiative priority matrix
Use when: A review group needs to compare a frozen set of initiatives against the same evidence expectations and local criteria.
Rename the options and criteria, then record local weights and evidence before scoring. Keep unknown work unranked and send any backlog assignment or roadmap commitment to its real owner.
Structure
- One frozen initiative set and criteria set for one dated review
- Evidence-bound totals plus separate gates, gaps, near ties, and sensitivity checks
Watch for: The matrix is not an Eisenhower tool, task manager, automatic backlog ranker, or roadmap approval workflow. Do not reuse sample weights as a universal prioritization formula.
Sources: [matrix-pack], [asq-decision-matrix]
High score with a failed mandatory gate
Use when: An option scores strongly on weighted criteria but does not satisfy one stated non-negotiable condition.
Fictional Option Dune keeps its calculated 9300 bp total for traceability, while the failed gate makes it ineligible and removes its rank.
Structure
- Weighted result remains reproducible and is not rewritten to zero
- Gate status, evidence, note, eligibility, and null rank remain explicit
Watch for: A failed local gate is not a legal, compliance, procurement, accessibility, or safety determination. Those reviews remain separate.
Sources: [matrix-pack], [nasa-decision-analysis]
Sensitivity reversal and exact tie
Use when: Reviewers need to know whether small, declared changes to local weights would alter the eligible order.
The pack includes one stable case, one reversal, and one exact tie. Each test preserves the 10,000 bp total and recomputes only the complete, gate-eligible options.
Structure
- Named criterion deltas, adjusted weight sum, reconciled totals, ranks, and finding
- No hidden tie-break, fallback score, automatic recommendation, or selection field
Watch for: Sensitivity reveals model dependence under the exact tested changes. It does not prove robustness against every assumption, evidence gap, or future condition.
Sources: [matrix-pack], [gao-alternatives-analysis]
Decide what the comparison needs next
Use these checks to pause weak analysis and route unresolved work. They do not choose an option or assign authority.
The question, option set, criteria, or weights changed after scoring
Choose: Open a new revision, freeze the updated boundary, reconnect evidence, and recompute every affected result.
Tradeoff: The review restarts, but people do not compare numbers produced under different models.
An option has unknown or disputed evidence for any criterion
Choose: Keep the score unknown and block its total and rank until a reviewer accepts option-specific evidence.
Tradeoff: The matrix may remain incomplete, but it does not invent precision or quietly penalize missing evidence.
A non-negotiable gate fails or remains unknown
Choose: Preserve any complete weighted total for traceability, mark eligibility separately, and withhold rank.
Tradeoff: Readers see more than one state, but a strong total cannot conceal a failed or unresolved mandatory condition.
Eligible options tie, nearly tie, or reverse under a tested weight change
Choose: Present the sensitivity result and evidence gaps to the human decision authority instead of adding an unreviewed tie-break rule.
Tradeoff: The matrix may not produce one preferred option, but the uncertainty stays visible and reviewable.
The review expired or source conditions materially changed
Choose: Stop using the rank, refresh evidence and gate states, append the change record, and complete a new evidence review.
Tradeoff: The decision pauses while the model catches up with current facts.
START WITH THE COMPARISON CONTRACT
Freeze the model before you score the options
Download the pack, replace its fictional question and evidence, keep every unknown and gate visible, and show sensitivity before the analysis reaches a human decision-maker.
Download the decision matrix packThe ZIP is locally reproduced. Public availability and adapted source accuracy require separate verification.
Record the final decision in a decision logThe matrix owns comparison analysis. The authorized decision, rationale, and outcome history belong in a separate record.
What a decision matrix cannot prove
Structure makes a comparison inspectable. It cannot turn assumptions into facts, assign real authority, or make the underlying judgment objective.
- Weights, score anchors, evidence confidence, gates, and near-tie thresholds are local judgments. A reproducible formula does not make them universal or unbiased.
- A total summarizes the declared model and available evidence. It does not prove quality, value, feasibility, accessibility, compliance, safety, or future success.
- Sensitivity tests cover only their explicit changes. Untested criteria, correlations, missing options, changing evidence, and different human values can alter the conclusion.
- CSV and JSON validation cannot verify that source evidence is true, current, complete, authorized, or interpreted correctly.
- A real workflow still needs authenticated identity, authorization, source access, concurrent-write handling, review assignment, audit integrity, retention, export, recovery, monitoring, and human governance.
- This ordinary informational article does not grant AI signup credits. Linked commercial pages follow their own current eligibility rules.
Sources and verification record
The same-release pack is the exact source for its fields, examples, formulas, and tests. Primary and specialist references support the broader decision-analysis, schema, export-safety, reserved-domain, and accessibility boundaries.
[matrix-pack] Playcode:Decision matrix fictional example
Checked August 1, 2026. Supports: The locally reviewed four options, five criteria, 10,000 bp weights, evidence records, gates, results, sensitivity cases, authority boundary, CSV projections, schema, and deterministic tests. Public availability remains unverified until deployment.
[nasa-decision-analysis] National Aeronautics and Space Administration:NASA Systems Engineering Handbook: Decision Analysis
Checked August 1, 2026. Supports: A documented decision need, criteria, alternatives, uncertainty, mandatory criteria, normalization, sensitivity analysis, and a distinct decision-maker. It does not prescribe this artifact or its weights.
[gao-alternatives-analysis] U.S. Government Accountability Office:GAO-15-37: Amphibious Combat Vehicle Acquisition
Checked August 1, 2026. Supports: Alternatives-analysis practices including a defined mission need, avoiding a predetermined solution, considering the status quo, independent review, and sensitivity analysis. It does not validate the fictional scores.
[asq-decision-matrix] American Society for Quality:What is a decision matrix?
Checked August 1, 2026. Supports: Weighted criteria, a consistent rating scale, using data when available, and treating the highest score as an input rather than necessarily the final choice.
[json-schema-2020-12] JSON Schema:JSON Schema Draft 2020-12
Checked August 1, 2026. Supports: The published JSON Schema dialect used for the pack structural contract. Cross-record arithmetic, authority, privacy, and chronology remain validator responsibilities.
[owasp-csv-injection] OWASP Foundation:CSV Injection
Checked August 1, 2026. Supports: The risk that spreadsheet applications interpret formula-prefixed CSV cells. The pack rejects those prefixes but does not claim a universal cross-application mitigation.
[rfc2606] Internet Engineering Task Force:RFC 2606: Reserved Top Level DNS Names
Checked August 1, 2026. Supports: Reserved example domain names for documentation and testing. The fictional pack restricts evidence and decision references to reserved .example.test hosts.
[wcag22] World Wide Web Consortium:Web Content Accessibility Guidelines 2.2
Checked August 1, 2026. Supports: The separate accessibility reference boundary. A matrix may track review readiness, but a weighted score cannot establish conformance.
Decision matrix template questions
How do I create a weighted decision matrix?
Freeze one question and option set, define criteria and evidence expectations, assign positive integer weights totaling 10,000 basis points, and score every option on the same fixed scale. Cite option-specific evidence, keep unknowns unknown, calculate once after summing, evaluate separate gates, and test whether local weight changes alter the eligible order.
Can I use this as a priority matrix template?
Yes, for one bounded set of initiatives or proposals. Replace the fictional criteria, evidence expectations, and weights with reviewed local ones. Keep missing evidence, mandatory gates, ties, and sensitivity visible. The matrix should not become an Eisenhower task tool, automatic backlog ranker, owner assignment, roadmap commitment, or universal priority formula.
What happens when evidence is missing?
The affected score stays unknown, so the option receives no total or rank. Do not substitute zero, the midpoint, an average, or another default. Record the missing evidence, confidence, owner, and next review instead. This distinguishes an unsupported judgment from a genuinely weak but supported score.
Should mandatory requirements be included in the weighted score?
Keep non-negotiable requirements as separate pass, fail, or unknown gates. A complete option may retain its weighted total for traceability, but a failed or unknown gate blocks eligibility and rank. This prevents a high score elsewhere from compensating for a condition the review explicitly declared mandatory.
Does the highest decision matrix score identify the best option?
No. The highest eligible total is a result of the declared criteria, weights, scores, evidence, and rounding. Review gaps, ties, near ties, sensitivity, excluded factors, and expired evidence before any decision. The pack intentionally contains no recommendation, selected-option, approval, or authorization field.
How should I handle a tie or near tie?
Show the tied options and the exact local threshold, then test relevant weight changes and evidence assumptions. Do not invent a hidden tie-break. A human decision authority may request more evidence, revise the model in a new version, or make a documented decision outside the matrix.
Can I open the CSV files in Excel or Google Sheets?
Yes. The CSV files are transparent projections of the canonical fictional JSON and can be imported into spreadsheet applications. The pack rejects formula-like prefixes and validates parity, but downstream applications differ. Review import settings, permissions, formulas, locale behavior, and adapted data before operational use.
Can Playcode turn this template into a review workflow?
Playcode can help build a bounded workflow around options, criteria, evidence, gates, calculations, reviews, sensitivity, and decision references. The real tool still needs authenticated identity, authorization, source verification, concurrent-write handling, audit integrity, retention, recovery, monitoring, and explicit human decision rights.
BUILD THE REVIEWED DECISION WORKFLOW
Turn a reviewed matrix into a bounded internal tool
Give Playcode the accepted record model, score scale, evidence rules, gates, sensitivity cases, review cadence, and authority boundary. Verify access, source truth, concurrency, audit integrity, recovery, and target behavior before operational use.
Build the decision workflowThis ordinary informational article does not grant signup AI credits. No option, priority, approval, compliance result, or business outcome is guaranteed.