What the tool is doing
A weighted average decision matrix turns a comparison into arithmetic. It has been standard practice in engineering and procurement for decades, under names like decision matrix, Pugh matrix and weighted scoring model. The idea is small enough to state in one line: not everything that matters to you matters equally, so multiply each score by how much the factor counts before adding anything up.
That single change is what a pro-and-con list cannot do. On a pro-and-con list, "free parking" and "I would never see my family" are both one line. Here the second one carries nine times the weight, so it moves the answer nine times as far.
Four numbers go in: your choices, the factors you care about, a weight from 1 to 10 for each factor, and a score from 1 to 5 for how well each choice does on it. Score times weight, summed down the column. The biggest total wins. Nothing is hidden, nothing is learned from other users, and no model picks the answer.
A worked example
Two job offers. Four things matter, and they don't matter equally. Weights are 1 to 10. Scores are 1 to 5, where 1 is barely and 5 is excellent.
| Factor | Weight | Offer A | Offer B |
|---|---|---|---|
| The work itself | 9 | 3 | 4 |
| Pay | 8 | 5 | 3 |
| Short commute | 6 | 2 | 5 |
| The team | 4 | 4 | 3 |
| Total | 95 | 102 |
Offer B wins by 7 points, and it is the offer that pays less. Pay scored higher for A, but pay is one factor out of four, and the two heaviest things on the list both point the other way. That result is the useful kind: it is not what the person expected, and the table says exactly which rows caused it.
Now argue with it. Suppose the commute barely registers, weight 1 instead of 6. Rerun the same scores and A wins 85 to 77. Nothing about the two jobs changed. One judgement about your own life did. That is the point of separating weights from scores, and it is why the answer is worth arguing with rather than obeying.
When it fits, and when it doesn't
It fits a choice between two or more real alternatives you could actually take: job offers, cities, apartments, schools, cars, laptops, software, supplier bids, which of three features to build first.
It does not fit a yes or no question. "Should I quit?" has one option in it, and a matrix needs something to compare against. Rewrite it as alternatives you could pick between: stay, quit and freelance, quit and job hunt. Then it works.
It also does not fit a decision with one disqualifying condition. If a flat over your budget is simply not happening, no weight makes that honest. Rule it out first, then compare what survives.
Three ways people get it wrong
Scoring to reach an answer you already picked. The commonest failure by far, and the reason this tool never shows a running total while you fill it in. If you can see who is winning, you will nudge the next score. Fill the whole thing in blind, then look.
Naming factors as negatives. Mixing "Cost" with "Salary" in one table breaks the arithmetic, because a high score means good on one row and bad on the other. Phrase every factor so higher is always better: affordability, not cost. Short commute, not long commute. Safety, not risk.
Too many factors. Fifteen rows feel thorough and behave like noise: each one gets a small weight, and the differences cancel. Five or six factors that genuinely differ between your options beat fifteen that mostly don't.
Keep reading
Worked examples covers real decisions with the factors filled in, and each one opens the wizard pre-loaded. About explains why the site exists and exactly where AI does and does not touch your decision.