Learn
How to build a weighted decision matrix that tells the truth
A decision matrix is a table with options down one side and criteria across the top. Each criterion carries a weight, each option earns a score on each criterion, and the weighted scores add up to a total. The arithmetic takes minutes. The care lies in what goes into the cells.
Where the method comes from
Engineers have used weighted tables to choose between designs for decades. Stuart Pugh, a British professor of engineering design, popularized a version in which each concept is marked better, worse, or the same against a reference design. The weighted sum, in which scores are multiplied by weights and added, is the form most teams now use for vendors, systems, and sites.
Building one, step by step
Choose four to seven criteria that genuinely separate the options; a criterion every option meets equally adds nothing. Divide 100 points of weight among them before looking at any scores. Define in words what a 1, a 5, and a 10 mean on each criterion, so that two people would score alike. Score each option, multiply by the weight, and add.
Then take the step most people skip. Move the largest weight ten points in each direction and see whether the winner changes. If it does, the decision turns on that weight, and the team should settle it openly.
Reverse-engineering the winner
This is the mistake that empties a matrix of meaning. A team that already prefers one option adds a criterion it happens to win, or nudges a weight until the totals agree with the feeling, and the table confirms a choice instead of testing it. Freezing the criteria and weights before scoring is the simplest protection; writing down any later change, with its reason, is the second.
What a matrix cannot show
A matrix treats every score as equally certain. A score built on a signed contract and a score built on a sales demonstration look identical in the cell. The table also judges only the options listed, and it says nothing about what would happen if a key assumption proved false.
The matrix inside a Decision Room
Argumont keeps the matrix and adds what it leaves out. Criteria carry weights that total 100 before options are scored, and the engine works the totals exactly. Each option has a Champion, the Challenger presses on the cells the leader depends on, and the Inspector marks whether the claim behind each score is supported, partly supported, unverified, or contradicted. The Verdict names the leader and the weight or score that would change it, and the Human Decision is recorded separately.
Two software vendors, scored
The criteria and weights are cost 30, fit to process 25, implementation time 20, support 15, and exit terms 10. Vendor A scores 6, 9, 5, 8, and 4; Vendor B scores 9, 6, 8, 6, and 8. Vendor A totals 180 plus 225 plus 100 plus 120 plus 40, or 665 out of a possible 1,000. Vendor B totals 270 plus 150 plus 160 plus 90 plus 80, or 750. B leads by 85 points. Move ten points of weight from cost to fit, and A gains 30 while B loses 30, closing the gap to 25. B still leads, so the choice is steady, and the team can sign with a clear account of why.
Questions
- How many criteria should a decision matrix have?
- Usually four to seven. Fewer hides real trade-offs; many more spreads the weight so thinly that no single criterion can move the result.
- Should matrix scores use decimals?
- Whole numbers on a clearly described scale are enough. Decimals suggest a precision that the underlying judgments rarely have.
Bring Argumont a Decision worth challenging.
Start with a paid 60 day pilot on your own real Decisions.