A weighted decision matrix compares several options against one shared set of criteria. What it produces is not the sentence “choose A” but a record that can be checked: which conditions are non-negotiable, what each option scored and on what basis, what change could reorder the ranking, and what trade-off was finally made by a person.

This tool separates must-have conditions from wanted criteria. A must-have has only three states — met, unmet, unchecked — and takes no part in the weighted sum; wanted criteria use a weight from 1 to 5 and a score. An option with incomplete scoring keeps the scores it has but is kept out of the complete-option ranking. PNG and SVG are for showing, CSV for checking the arithmetic, Markdown for keeping the report, and JSON for restoring the document to keep editing. The export is not a third-party certification, nor purchasing, investment or any other professional advice.

1. 01Write down the specific decision question, the comparison period and the feasible options. Consider at least one “keep things as they are” or low-cost trial option, so you are not only comparing several essentially identical choices.
2. 02List the must-have conditions first, then the wanted criteria. A must-have has to be checkable; writing “the better the better” as a must-have loses any explainable screening standard.
3. 03Fix the criterion weights before looking at the option scores, and agree what a 1 and a 5 each mean. Do not count the same benefit twice through price, value for money and budget saving.
4. 04Fill in the scores and the evidence behind each. Leave a blank where there is no information; unknown is not zero and does not mean the condition failed. After changing a must-have's state, re-check the related scores and evidence.
5. 05Examine the result for the complete options and the weight sensitivity. The tool nudges each wanted weight up and down by one step and shows which local changes would change the leader; it is not a robustness proof covering every uncertainty.
6. 06Write down the human review conclusion, the negative consequences, the open questions and the date of the next review, then export the report and the JSON. Long-term storage cannot rely on browser storage or a link in a chat app alone.

The decision analysis of Charles Kepner and Benjamin Tregoe splits objectives into MUSTs and WANTs: first screen by the minimum necessary conditions, then compare wanted performance, then examine adverse consequences. That separation matters more than “multiply everything by a weight and add it up”. The Kepner-Tregoe method description and the FHWA's practical application both distinguish clearly between non-compensable necessary conditions and comparable wanted objectives. [1][2]

The calculation on this page is total = Σ(wanted criterion weight × that option's score), with a maximum of 5 × Σweights. It is a workable additive comparison, not a calibrated utility measurement; the tool's 1–5 input range also does not represent the full original process of any institution. The multi-attribute evaluation literature provides a more rigorous theoretical background, but it cannot make every real preference automatically linear, independent or mutually compensable. [3]

Compared with a PMI list, the matrix requires several options to share their criteria; compared with a decision tree, it has no branch probabilities or sequential choices; compared with a risk matrix, it compares the value trade-offs between options rather than event risk. Its strength is exposing inconsistent standards and gaps in the evidence. Its limits come from the same table: missing a key criterion, double-weighting correlated factors, or back-fitting weights to a preferred answer can all produce a tidy but unconvincing ranking. When totals are close, look at the specific differences instead of adding decimals to manufacture certainty.

The built-in “Study laptop: fictional comparison data” has three options. The must-have is that the total cost stays within budget; the wanted criteria are carrying convenience, fit with coursework and battery convenience, weighted 4, 5 and 3.

A's arithmetic is 4×5 + 5×3 + 3×5 = 50. C's computed value can still be checked, but a high score cannot offset a failed must-have. Delete one of B's wanted scores and it becomes incomplete rather than taking a formal place with a zero. These numbers only demonstrate the algorithm; they are not a product review or a market price.

A second template compares an office, a shared community space and an online breakout meeting for a small-team workshop. It makes actual participant accessibility a must-have, reminding the organiser to confirm accessibility, connectivity and timing limits instead of letting a convenience total stand in for a necessary condition.

• Up to 40 options and 20 criteria; a table much larger than that may be exportable but unsuitable for a single discussion.

• The score evidence is user input and is never looked up or verified automatically.

• Weight sensitivity is only shown when every option satisfies the current complete-calculation conditions, and it only moves one weight by one step.

• Totals should not be compared directly across people, across time or across different criterion tables.

• PNG has pixel and size limits; a particularly large report should use SVG or a structured file.

• CSV is a table view, not a complete application archive; use JSON to restore settings, evidence and the human conclusion.

Treating an unchecked must-have as met skips the most important investigation. Filling an unknown wanted criterion with a middle score also disguises it as evidence. Another misuse is choosing the preferred option first and then adjusting the weights until it leads; keep the original criteria and the reason for any change. Serious consequences, irreversible commitments and professional compliance matters need a separate review and cannot be handed to a compensable additive total.

• Individuals can compare study devices, courses or reversible life arrangements; a project team can compare venues, ways of working or design options; an organisation can use it to record the criteria and evidence behind a procurement discussion.
• For high-impact decisions about people, such as hiring, check the legality, fairness, accessibility and evidence of the criteria separately, and do not treat this tool as an automatic screening system.

Editing, calculation, drawing and exporting happen in the browser, with no account or cloud sync. Local autosave may be unavailable because of private mode, quota or cleanup, and the interface says so and points to a JSON backup. Sharing is not encryption: anyone who has the link can read the document inside it. Do not put private comments, commercial quotations or personal identifying information into a public link; the reference links are optional external visits.

[1] Kepner-Tregoe: the MUST / WANT split in Decision Analysis. For the original method background see also Kepner and Tregoe, The New Rational Manager, 1981. （访问日期：2026-09-22）

[2] FHWA: Work Zone Road User Costs, Chapter 3. This is a source for how the method is applied; it does not mean this tool is suitable for formal traffic-safety analysis. （访问日期：2026-09-22）

[3] Ward Edwards: multiattribute utility measurement, 1977. Distinguishes theoretical measurement from the simplified subjective scoring on this page. （访问日期：2026-09-22）