How to make a decision matrix with ChatGPT

Use the table to explain a choice, not hide it

A decision matrix is a table that helps you compare options using the same questions. You choose what matters, give each factor an importance, and score the options. ChatGPT can organize that work and calculate a first result. The table does not discover your priorities for you. If you choose poor criteria, the numbers can make a weak decision look careful.

Start with one decision stated plainly. “Choose a room for a Saturday workshop for twenty people” is more useful than “find the best venue.” Say who the choice is for, what the result must achieve, and what information you already have. Keep the options concrete enough to check.

A matrix is useful when several trade-offs matter. It is less useful when one rule settles the choice, such as a required license or a fixed maximum cost. In that case, filter the options first. Do not add unnecessary scores just to turn a simple decision into a complicated exercise.

Separate must-haves from preferences

A must-have is a condition an option cannot fail. Examples include a required date, an available size, compatibility with existing equipment, or access for the people attending. A preference is something you can trade against another benefit, such as a shorter journey or a nicer view. Put these in separate lists before assigning weights.

If an option fails a must-have, remove it or mark it as needing a confirmed solution. Do not let a high score elsewhere cancel the failure. A venue that cannot accommodate a participant should not win because its rent is cheap. If a missing feature can be added, include the cost and the confirmed arrangement before scoring.

Ask ChatGPT to show the filter result with evidence beside each option. If a required fact is unknown, label it unknown. Unknown is not the same as a pass or a fail. Decide whether you can get the information before making the decision. That small step often matters more than the final weighted total.

Choose a few criteria with clear meanings

Use criteria that describe different parts of the decision. If you include both purchase price and affordability with the same meaning, cost gets counted twice. If you include ease of use, define what would make an option easy for the actual user. A vague label lets the model score by general impressions.

Give each criterion a measurement or a scoring rule. Travel time might use minutes from the group’s starting point. Price might mean the full cost including required extras. A support score might depend on named services rather than a claim that a company is excellent. Tell the model where each value came from.

Keep the list short enough to understand. Four or five useful criteria can be better than fifteen overlapping ones. Ask whether removing a criterion would hide something important. If not, remove it. You want the table to reveal the trade-off, not bury it under a large number of small judgments.

Set weights before looking for a winner

A weight says how much a criterion matters relative to the others. You can use numbers adding to one hundred, such as travel time forty, price thirty-five, and room setup twenty-five. The total is easy to check. Equal weights are fine when you have no reason to value one factor more, but they are still a choice.

Discuss weights with the people affected before filling in scores. If the table already shows your favorite winning, it is tempting to adjust the priorities afterward. That can be a useful change when you discover a real need, but record why the weight changed. Do not quietly tune the table to justify a decision already made.

ChatGPT can suggest a starting set and explain the trade-offs. Treat those suggestions as questions to answer. For example, should a lower cost matter more than saving the team an hour of travel? The model cannot know that without your situation. You own the priority; it helps express it clearly.

A worked calculation with invented venue scores

This teaching example uses fictional venues and scores. All three have passed the required capacity, date, and access checks. Travel convenience has weight forty, full price has weight thirty-five, and room setup has weight twenty-five. Each option receives a score from one to five, where five is better. The same scoring rules apply to every venue.

Venue A scores four, three, and five. Multiply each score by its weight: four times forty is one hundred sixty, three times thirty-five is one hundred five, and five times twenty-five is one hundred twenty-five. Add them to get three hundred ninety. Divide by the total weight of one hundred to get 3.90 out of five.

Venue B scores three, five, and three, giving 3.70. Venue C scores five, two, and four, giving 3.70. A leads under these priorities. That does not establish it is the best venue in every situation. It tells you how the supplied judgments combine. Check the arithmetic with a calculator or spreadsheet and keep the facts that support the scores.

Weighted result = sum of (score × weight) / sum of weights
A = (4 × 40 + 3 × 35 + 5 × 25) / 100 = 3.90
B = (3 × 40 + 5 × 35 + 3 × 25) / 100 = 3.70
C = (5 × 40 + 2 × 35 + 4 × 25) / 100 = 3.70

Check the scoring rules and missing evidence

Every score needs a reason. Ask ChatGPT to return the source value, the scoring rule, and the assigned score in separate columns. For a price score, you might define ranges ahead of time. Do not allow a costly option to receive a high score just because the model describes it as good value. The score should follow the stated rule.

A missing value should remain missing. Replacing it with an average can create a winner from a guess. If you need to explore a decision before the value is known, show several possible scores and explain that they are assumptions. Keep that exploration separate from the evidence table.

Watch the direction of the scale. A lower cost may be better, while a larger capacity may be better up to the required level. Convert both into a consistent higher-is-better score. If extra capacity brings no benefit, do not keep rewarding it. A clear scale makes it possible for another person to understand and question the result.

See whether a small change reverses the result

A close result deserves a sensitivity check. That simply means changing a reasonable assumption to see whether the winner changes. In the venue example, raise the price weight to fifty and lower travel to twenty-five, leaving setup at twenty-five. A becomes 3.75, B becomes 4.00, and C becomes 3.50. B now leads because saving money has become more important.

This is not an error in the first table. It shows that the decision depends on your priorities. Ask whether the new weights describe the real situation better. You can also test an uncertain score. If A’s setup score falls from five to four under the original weights, its result becomes 3.65 and the lead changes.

When one uncertain fact can reverse the result, investigate it before committing. A brief call to confirm the room layout may be more valuable than refining every decimal in the table. If the result stays the same across reasonable changes, you have stronger support for that choice within the stated conditions.

Write the recommendation with its conditions

Ask ChatGPT to turn the result into a short recommendation that names the choice, the main reasons, the trade-off, and the remaining check. For example: “Choose A if travel convenience remains the top priority and the advertised room setup is confirmed. Choose B if the budget becomes the main constraint.” That is more useful than saying A is objectively the best.

Keep the raw evidence and the scoring sheet with the recommendation. If a price or requirement changes, update the affected fact and recalculate. Do not continue using the old conclusion simply because it was written neatly. A matrix is a record of a decision under certain facts and priorities.

For a consequential choice, use the table to support the responsible person’s judgment. It can explain options and expose assumptions, but it does not approve a purchase or replace expert advice where that is needed. You are ready to decide when the must-haves are checked, the scores have support, and you understand what would change the recommendation.

Use evidence about the actual setup before scoring a purchase. Check the product facts behind your scores →