How to build a decision matrix
- Write the question. One sentence with a constraint: "Which help desk tool for six agents, starting in December?" A vague question produces vague criteria.
- List the must-haves. Pass or fail. Anything that fails is out before scoring.
- Shortlist two to five options. More than five and the scoring takes longer than the decision is worth.
- Pick four to six criteria and write what a 1, 3 and 5 look like for each.
- Set weights that add up to 100, before anyone scores.
- Score each option, separately if more than one person is deciding, then compare and settle the big gaps.
- Multiply, add, divide by 100. Then run the sensitivity check before you announce a winner.
The sensitivity check
The total is the output of your weights, and the weights are guesses. The quickest test of whether the winner is real: move 10 points from the heaviest criterion to the one you argued about most, and recalculate.
- The winner stays the winner. The result is robust. Decide. The project example above is like this: team billing wins under both weightings.
- The winner changes. You do not have a decision yet. You have found the question that matters. In the job offer example, the answer depends on stability, and stability depends on one fact the matrix cannot supply (the startup's runway). Go and get that fact.
A lead of under about 0.2 on a 1 to 5 scale is close to a tie. Treat it as one.
Ties, gates and lower-is-better criteria
Ties. Rerun the totals with each decision-maker's weighting. An option that wins or ties under all of them is the safer pick, which is how the apartment example breaks its tie. If that does not separate them, choose the option you would be happier to have been wrong about.
Gates. A must-have is not a heavy criterion. If an option does not allow cats or cannot connect to your sign-in system, no score on other criteria should rescue it.
Lower is better. Cost, effort and risk run the wrong way. Flip them so that 5 always means good (cheapest, least effort, lowest risk) and say so in the criterion name. Mixing directions is the easiest arithmetic error to make, and it quietly rewards the most expensive option.
When the result feels wrong
Sometimes the matrix picks an option and your stomach says no. That is useful information, not a failure of the method. Ask which cell you would change to make the result match your instinct. If you can name it and defend the change, the matrix was missing something, so fix the score or add the criterion. If you cannot, your instinct may be reacting to something you are avoiding. Write that down in the decision note too.
Decision matrix vs pros and cons
A pros and cons list works for one option, yes or no. Give each item a weight from 1 to 3 and total each side. When the totals are close, as in the team lead example, stop counting and ask which cons can be undone. For three or more options, or when several people need to agree, the matrix is better because it forces every option onto the same criteria.
Common mistakes
- Adjusting weights after seeing the totals. Write them down first.
- Two criteria for the same concern. "Price" and "cost" double-count money.
- Scores without anchors. One person's 4 is another's 3.
- Two decimal places of false precision. The inputs are guesses. Report the lead and whether it survives the sensitivity check.
- No written record. Without the matrix and the reason, the debate restarts the first time the choice is inconvenient. Log it in a decision log.
Keep the scores with the decision they justified
Save this decision matrix template in Scribelet, link it to the project it decides, and log the outcome next to it. When someone asks in March why you picked Ticketbird, the weights are one search away.
Free plan, no credit card. See pricing