Weighted Decision Matrix

What this tool does

CM8-230 is a working weighted decision matrix — sometimes called a decision grid, prioritisation matrix or weighted scoring model. You list the options you are choosing between, agree the criteria that matter and how much each matters, then score every option against every criterion. The tool multiplies each score by its criterion's weight, totals the results per option, shows what drives each total, and — the part most spreadsheets skip — checks whether the winner survives having any single criterion removed.

Each record is one score: one decision, one option, one criterion. Fifteen rows make a complete matrix of three options against five criteria, and one file can hold several decisions side by side — name each decision consistently and the tool builds the matrix, charts and sensitivity check for the decision with the most scores in the current filter.

Everything runs inside this single file. There is no account, no upload and no network request of any kind, so a comparison of suppliers, machines, premises or candidates never leaves the computer you are using.

When to use a matrix

A decision matrix earns its keep in a specific situation: three or more options, several criteria that pull in different directions, and more than one person with a view. Choosing between two options on one criterion needs no matrix. Choosing between four machines where the cheap one is slow, the fast one is late and the reliable one is expensive is exactly what the method is for.

It is also the honest way to handle group disagreement. When two managers each champion a different option, arguing about the options goes in circles; arguing about criteria and weights — before anyone scores anything — surfaces what the disagreement is actually about. Half the value of the exercise is that conversation, not the arithmetic at the end of it.

Typical uses: selecting equipment or software, shortlisting suppliers, choosing premises, prioritising projects for a limited budget, and comparing job candidates against a defined role profile.

Choosing the criteria

The criteria decide the answer far more than the scores do, so spend the time here.

• Four to eight of them. Fewer than four and the matrix adds little over a conversation; more than eight and the weights flatten out until nothing is decisive. The tiles flag a count outside that range.
• Independent. "Purchase cost" and "value for money" overlap — an option good at one is scored twice for the same virtue. If two criteria would always move together, merge them.
• Assessable. Every criterion needs a question you can actually answer: "how long until it is running on our floor?" works; "strategic fit" does not, until you define what evidence would move the score.
• Differences, not table stakes. If every option must meet a requirement, it is a pass/fail gate applied before scoring, not a criterion. Only score the things on which the options genuinely differ.

Setting the weights

Weights run 1–10 and belong to the criterion, not the option — set the same weight for a criterion on every option's row, or the totals stop being comparable. (If rows disagree, the tool uses the highest weight it finds for that criterion when it draws the matrix and the weights chart.)

The common failure is that everything gets an 8 or a 9, because everything "matters". Equal weights are not wrong — that is the unweighted variant, covered below — but accidental equal weights mean nobody made a choice. A forced ranking beats a round of everything-is-a-9: put the criteria in order of importance first, then assign weights that respect the order, and make yourself use at least a spread of four or five points between the top and bottom of the list. The criterion-weights chart shows the result: a balanced profile, or one bar towering over the rest. A one-issue profile may be right, but it should be deliberate — if one criterion carries most of the weight, the matrix mostly restates that single priority.

Scoring discipline

Scores run 1–10 for how well the option meets the criterion. The single habit that most improves a matrix: define what a 10 means for each criterion before anyone scores anything. "Delivery lead time: 10 = running on our floor within a month; 1 = more than six months." Without anchors, one scorer's 7 is another's 4 and the totals compare optimism, not options.

Write the rationale for every score — one or two sentences on why. It forces the scorer to have a reason, it lets the group challenge a number rather than a feeling, and it is what makes the record worth keeping when the decision is questioned a year later.

Record a confidence level per score. A 9 based on a signed quotation and a 9 based on a salesperson's verbal promise are different animals; the tool counts the low-confidence scores on a tile so you can verify them before the decision is made, not after.

Reading the results

The totals bar answers "who wins"; the stacked chart answers the better question, "why". Each option's bar is built from its weighted score per criterion, so two options with similar totals show their different routes — one riding on price, another on reliability and service. If the room is surprised by which segments dominate, the weights or the scores need debating; that debate is the point.

The weakest-criteria chart looks at the leading option alone: where does it score badly? Choosing the winner does not make those weaknesses disappear — they become the risks to manage after the decision. If the winning machine's weak spot is delivery lead time, the follow-up action is a contractual delivery date, written down before the order is placed.

The matrix table is the classic print view — criteria down the side with their weights, options across the top, score × weight in each cell, totals along the foot — so anyone can check any cell by hand.

The sensitivity check

The sensitivity table recomputes every option's total with each criterion removed in turn and reports whether the leader changes. It is a deliberately simple, honest test of how robust the result is.

If the result holds with every criterion removed, the winner does not depend on any single judgement, and you can proceed with reasonable comfort.

If removing one criterion flips the winner, the whole decision hinges on that one criterion — its weight, and the scores given under it. That is not a fault in the method; it is the method telling you where the decision actually lives. The response is more discussion of that criterion — are the scores evidenced? is the weight really right? — not a bigger spreadsheet. A decision that flips on one criterion and was scored with low confidence under that same criterion is a decision not yet ready to make.

A small gap between the top two options is flagged on the Score spread tile using the close-call threshold in Settings. A close call means the matrix has done its job — it has told you the options are genuinely comparable, and the final choice can rest on factors outside the matrix without dishonesty.

Unweighted and group scoring

Unweighted matrix: give every criterion the same weight (say 5) and the totals reduce to a straight sum of scores. This is the classic simple decision matrix, and close to a Pugh-style comparison — useful for a first pass, or when the group cannot yet agree on weights. If the unweighted and weighted results agree, the weights were not decisive; if they disagree, the weights are doing real work and deserve scrutiny.

Group scoring: averaging everyone's scores feels fair but launders disagreement into a bland middle number. A better routine: agree the criteria, weights and score anchors together; score independently; then compare, and discuss only the scores that differ by three or more points. Those outliers are where someone knows something the others do not — or where the criterion was understood differently. Enter the scores the group settles on after that discussion, and note in the rationale where views diverged.

The formulas

Nothing else happens to the numbers. There is no normalisation and no hidden scaling, which is why every cell of the matrix table can be checked with mental arithmetic — and should be, at least once.

FAQ

Can I keep more than one decision in the file? Yes — the decision name on each row separates them. The charts, the matrix and the sensitivity check are built from the decision with the most scores in the current filter; the register and CSV export always contain everything.

What if an option has no score for some criterion? Its total simply misses that criterion's points, which silently punishes the option. Complete the grid — score every option on every criterion — before reading anything into the totals. The matrix table shows a dash where a cell is missing.

Two options are within a few points. Which wins? Neither, on this evidence. Totals within the close-call threshold mean the matrix cannot separate them; identical totals certainly do not mean identical options. Decide on something real outside the matrix — or sharpen the scores where confidence is low.

Should cost be a criterion or handled separately? Both approaches are defensible. As a criterion, price competes with everything else through its weight. The alternative is to score everything except cost, then weigh the benefit totals against the prices by eye. For large purchases the second often produces the clearer conversation.

Can I use it to rank many items rather than pick one? Yes — project prioritisation is the same arithmetic. The totals bar becomes the ranked list; read the close-call flag as "these two are effectively tied".

Saving your work

Scores, settings and the report header are written to this browser's local storage as you type, and the toolbar shows the time of the last save. That storage belongs to one browser on one computer: another browser, a private window, a second machine or a clean-up tool that clears site data will not have it.

Treat Export .json as the real save — one file containing everything, which Import .json restores anywhere. Export CSV gives you the flat score register for spreadsheet work. Reset asks twice, then erases everything this tool has stored. There is no undo.

Accuracy & disclaimer

The arithmetic is deliberately simple — weight times score, summed — and the tool does it faithfully. Everything that matters sits underneath it: whether the criteria are the right ones, whether the weights reflect a real ranking of priorities, and whether the scores rest on evidence or optimism. A weighted total is structured judgement, not measurement, and a two-point gap between options is noise unless the scores underneath it are solid.

This is a decision-structuring aid, not advice on the decision itself. It cannot tell you whether a criterion is missing, whether a weight flatters a favourite, or whether a score was generous — and any of the three can change the winner.