Set the weights before you look at the options
The order of operations is the only part of this method that is load-bearing. Weight the criteria first, in the abstract, before you have any option in front of you. Then score. If you score first, you will find yourself nudging weights until the option you already wanted comes out on top, and the exercise becomes an elaborate way of writing down a conclusion you reached in the first thirty seconds.
Weighting first also survives the argument better. Three people can usually agree that accuracy matters more than resale value long before they agree on which saw is more accurate. Fix the part you agree on, and the disagreement narrows to specific score cells you can go and check.
Score relatively, and invert anything where less is better
The 0-10 scale is not an absolute quality rating. It is the position of these options relative to each other on this criterion. If three saws cost $700, $1,400 and $2,800, the price row might read 10, 7, 2 — the spacing should reflect how far apart they actually are, not just their order. If all three are within $50 of each other, score them 7, 7, 7 and move on. That is not a failure of the method; it is the method correctly reporting that price does not separate these options.
Criteria where a lower number is better — price, footprint, lead time, noise — have to be scored inverted, with the best option getting the high score. Everything is summed, so a criterion scored in the wrong direction quietly pushes the ranking backwards.
Reading the "drop one criterion" table
This is the part of the output worth the most attention. It re-runs the ranking with each criterion removed in turn and reports any case where the winner changes. A criterion that appears there is single-handedly deciding the outcome. Sometimes that is correct and you meant it. Often it means you gave a 9 or a 10 to something you have not actually verified, and the whole result rests on one guess.
| What the table shows | What it usually means |
|---|---|
| Nothing flips | The result is broad-based. Reasonably safe to act on. |
| One criterion flips it | Verify the scores on that row before committing. |
| Several flip it | The options are close enough that the ranking is noise. Decide on something outside the grid. |
When a matrix is the wrong tool
Two situations defeat it. The first is a hard constraint: over budget, will not fit through the shop door, cannot be delivered before the job starts. A weighted sum will happily average a disqualifying flaw away against a pile of small advantages. Filter on constraints first, then score whatever survives. The second is missing information. If you keep hitting cells where the honest entry is "I don't know", the matrix is telling you to go and find out rather than to decide. Fill it in with guesses and you will get a confident-looking number built on them.
Neither of those is a reason to skip the exercise. It is a reason to notice what the empty cells are telling you before you read the total.
Questions people ask
What happens if I give every criterion the same weight?
You get a plain score total, which is a legitimate thing to want but is not what this tool is for. The whole value of the method is in recording that some things matter more than others. If you are stuck picking weights, anchor the ends first: give a 10 to the criterion you would sacrifice the most for, a 2 to the one you would drop first, then place everything else between them. Relative placement is easier than absolute rating.
Can I put real numbers in — dollars, inches, decibels — instead of 0-10 scores?
No, and it will break the ranking if you try. Raw values live on wildly different scales, so a criterion measured in dollars would swamp one measured in inches purely because its numbers are bigger. Convert everything to the 0-10 scale first, remembering to invert the criteria where a smaller value is better. That conversion is a judgement call, and it is one of the places your priorities enter the result.
The winner is not what I want. Now what?
Treat the mismatch as data rather than as an error. It almost always means one of two things: the weights you wrote down are more diplomatic than the ones you actually hold, or there is a criterion driving your gut that never made it onto the list. Find out which, fix it, and re-run. If it still disagrees after an honest edit, you are free to ignore it — the arithmetic has no standing that your judgement does not give it.
How many options and criteria should I use?
Two to five options and three to six criteria. Beyond about eight criteria the weights get diluted to the point that everything scores near the middle and the ranking stops separating; beyond five options you are doing screening rather than deciding, and screening is better done with a hard filter. The tool caps at five options and eight criteria for that reason.