A decision matrix (also called a weighted scoring model or Pugh matrix) turns a messy “which one should we pick” into a transparent calculation. You list the options, name the criteria that matter, assign each criterion an importance weight, and score every option against every criterion. The weighted totals reveal the winner and — just as importantly — show how close the race was. This builder computes the result live and generates a prompt that makes an LLM explain the trade-offs in plain language.
How it works
You enter your options (the choices) and your criteria, giving each criterion an importance weight from 1 to 10. Then you score each option against each criterion on a consistent scale. The tool multiplies every score by its criterion weight, sums the products per option, and ranks them — updating instantly as you type. The display normalises the weights so you can see what share of the decision each criterion controls. Alongside the live result, it assembles a prompt containing your full matrix and asks the model to narrate which criteria decided the outcome, flag any near-ties, and recommend what to verify. Everything runs in your browser.
A worked example
Suppose you are choosing between three CRM platforms — call them Option A, Option B, and Option C — for a 50-person sales team. Your criteria and weights might be:
| Criterion | Weight |
|---|---|
| Integration with existing tools | 9 |
| Ease of onboarding for the sales team | 8 |
| Total cost over 3 years | 7 |
| Reporting and analytics depth | 6 |
| Vendor support quality | 5 |
After scoring each option 1-5 on each criterion, the tool multiplies score by weight per cell and sums the rows. For illustration: if Option A scores 4 on integrations (4×9=36), 3 on onboarding (3×8=24), 2 on cost (2×7=14), 5 on reporting (5×6=30), and 4 on support (4×5=20), its total is 124. The other options are scored the same way and ranked. The generated prompt then asks an LLM to narrate why Option A won or lost, which criteria were decisive, and what assumptions in the weights or scores should be re-examined.
How to build a good matrix
Start with criteria, not options. Decide what matters and how much it matters before you think about how the options score. Setting weights after you already know the scores makes them rationalise a choice rather than reveal it.
Keep the scoring scale consistent. Pick 1-5 or 1-10 and use it identically for every criterion-option cell. If “good” means 4 for one criterion and 3 for another, the numbers are not comparable.
Limit to five to eight criteria. More than this and each additional criterion has diminishing influence on the result while making the matrix harder to fill in honestly. If you have twelve candidate criteria, consolidate them into the five or six that genuinely differentiate the options.
Don’t let cost dominate by accident. Cost is a strong criterion but it is rarely the only thing that matters. If you assign it a weight of 10 and everything else a weight of 3, you have built a cost calculator, not a decision matrix.
Treat a near-tie as a signal. When the top two totals are within about five percent of each other, the matrix is telling you the decision is genuinely close and you should look harder for a differentiating factor — a reference call, a pilot, a deeper technical comparison — rather than calling the higher number the winner.
When to use the AI-narrated explanation
The numeric output tells you who won; the generated prompt tells you why and how confidently. The narration is most useful when you need to explain the decision to stakeholders, when you want the model to identify what would flip the result if one assumption changed, or when you want to document the reasoning for an audit trail. Paste the prompt into your LLM of choice to produce a ready-made decision write-up.
Tips and examples
Keep your scoring scale consistent — pick 1 to 5 or 1 to 10 and use it for every cell. Set weights before you score, so the importance of each criterion is decided independently of how the options happen to perform. When the top two totals are within a few percent, treat the matrix as inconclusive and add a tie-breaking criterion rather than forcing a call. Use the generated prompt to produce the write-up you share with stakeholders: it converts the bare numbers into a defensible argument and exposes the assumptions a reader should challenge.