The method

List your criteria, assign each a weight, score every option on every criterion, then sum the products:

Score = Σ (weight × rating)

Weights should total 1.00 (or 100%), which forces the useful discipline that raising the importance of one thing must lower another. Ratings run on a fixed scale, usually 1 to 5.

A worked example: three job offers

Option B wins on 4.00, despite having the lowest compensation score of the three. That is the point of the exercise: it shows that once growth, flexibility and team are weighted as you claim to weight them, the highest salary does not win. Either you accept B, or you discover that your stated weights are not your real ones — which is equally valuable information.

Notice also that A and C tie at 3.60 by completely different routes. A is a spiky option, strong on pay and weak on commute; C is uniformly middling. A weighted sum treats those as equivalent, and you almost certainly do not.

Where weighted scoring genuinely misleads

This is the section most decision-matrix pages leave out, and it matters more than the formula.

• It manufactures precision. A total of 4.00 against 3.60 looks decisive. It rests on weights you invented and ratings you guessed. A difference under roughly 10% is noise — treat it as a tie and decide on other grounds rather than deferring to the decimal.
• People reverse-engineer the weights. The most common failure is deciding first, then adjusting weights until the matrix agrees. Guard against it by fixing the weights before you score anything, ideally writing them down and not revisiting them.
• Ordinal scores get treated as interval. A 4 is not twice a 2, and the gap between 4 and 5 is rarely the same size as between 1 and 2. Multiplying and adding them is mathematically sloppy but practically tolerable — provided you do not over-read small differences.
• Correlated criteria double-count. If you score both salary and total compensation, or both team and manager, you have silently given that dimension extra weight.
• Deal-breakers do not belong here. If an option fails a hard constraint — cannot relocate, cannot afford it, would not pass a background check — remove it entirely. A weighted sum lets a high score elsewhere compensate for an unacceptable value, which is exactly wrong.
• Missing criteria are invisible. The matrix is confident about what you put in it and silent about what you forgot, which is usually the thing that determines the outcome.

What it is actually good for

Not producing the answer — surfacing the structure of the decision. Three real uses:

Making implicit weights explicit. Most people cannot articulate how much they trade pay against flexibility until forced to put numbers on it, and the act of assigning weights is more clarifying than the resulting total.

Locating group disagreement. When a team scores the same options separately, the interesting output is where they diverge. Usually the disagreement is about weights, not facts, and identifying that turns an unproductive argument into a resolvable one.

Sensitivity testing. Change one weight and see whether the winner changes. If the answer holds across a wide range of plausible weightings, it is robust. If it flips when compensation moves from 0.30 to 0.35, you have learned that the decision genuinely hinges on that one judgement — which is far more useful than the original score.

Questions people actually ask

How many criteria should I use?

Four to seven. Fewer misses something important; more dilutes every weight to insignificance and invites overlap.

What if two options tie?

Take it as a real finding rather than a problem to break. A genuine tie means the criteria you listed do not distinguish them, so either the missing criterion matters most, or you can pick freely and lose little.

Should I score before or after gathering information?

Set criteria and weights first, gather information second, score last. Doing it in that order is the main defence against fitting the analysis to a conclusion you already reached.

Is my matrix stored anywhere?

No. Criteria, weights and scores stay in your browser and nothing is transmitted or saved.