Decision Frameworks & Evaluation

What You’ll Learn

• How to use Claude to structure complex decisions with multiple competing options and criteria
• How to build weighted decision matrices that surface the best choice based on what actually matters
• How to use Claude as a “devil’s advocate” to stress-test decisions before you commit

The Use Case

Most business decisions that go wrong don’t fail because of bad luck — they fail because of incomplete analysis. Options were compared on the wrong criteria, important trade-offs were glossed over, or the person making the decision was already committed to one answer and evaluated the others unfairly.

Claude excels at structured decision analysis precisely because it doesn’t have a preference. It will systematically apply criteria to options without the cognitive biases that make human decision-making inconsistent — anchoring to the first option presented, overweighting vivid information, or avoiding the discomfort of a conclusion that challenges existing assumptions.

This workflow is valuable for vendor selection, technology stack choices, strategic pivots, pricing decisions, hiring choices, market entry decisions, and any situation where you have multiple viable options and need a defensible way to choose between them. It’s also a powerful tool for making your reasoning transparent to stakeholders — a well-documented decision matrix is far easier to present and defend than “we thought about it and this felt right.”

Step-by-Step Guide

Step 1: Define the decision and the constraints

Start by telling Claude exactly what you’re deciding, what the constraints are, and who will be affected by the outcome. Constraints narrow the option space and prevent Claude from suggesting technically valid options that are practically impossible.

Prompt:

“I need to make a decision about [describe the decision]. The constraints are: budget of [X], timeline of [Y], team size of [Z], and we must [any non-negotiable requirements]. The decision affects [who]. The goal of this decision is to [desired outcome]. Help me structure an evaluation framework.”

Claude will confirm the decision space and ask clarifying questions if the problem is underspecified. This step forces you to articulate the decision clearly, which is often the most valuable part of the process.

Step 2: Generate and enumerate the options

Before evaluating, make sure you have a complete and diverse set of options. One common decision-making failure is entering the evaluation with an incomplete option set — typically because the options were generated by the same team that will be evaluating them, creating confirmation bias from the start.

Prompt:

“I’ve identified these options for my decision: [list your options]. Before I evaluate them, can you suggest any options I might have overlooked? Also, are there any options I’ve listed that are likely to be eliminated immediately by my constraints, so I can simplify the comparison?”

Claude often surfaces options like “do nothing / keep status quo,” “outsource rather than build,” “phased rollout,” or “hybrid approaches” that teams overlook because they’re focused on comparing a short list of obvious choices.

Step 3: Define and weight the evaluation criteria

This is the most important step in structured decision-making. The criteria you choose and how you weight them determines the outcome. Bad decisions often result from implicitly applying criteria that don’t match the stated goal — for example, choosing the cheapest vendor when reliability is actually the most important factor.

Prompt:

“For my decision about [X], help me define 5-8 evaluation criteria. For each criterion, suggest a weight (percentage) based on my stated goal and constraints. My most important priority is [state it explicitly]. After you suggest weights, I’ll review and adjust them. Then we’ll use these criteria to evaluate each option.”

Have this conversation out loud (metaphorically) — the weights you assign are themselves a decision about what matters. If you and a key stakeholder assign very different weights to the same criteria, you have a misalignment that should be resolved before the decision, not after.

Step 4: Build the decision matrix

Now score each option against each criterion. Claude can help you score options based on the information you provide, and flag where you’re making assumptions.

Prompt:

“Build a decision matrix for my [decision]. Options: [list options]. Criteria with weights: [list criteria and weights]. For each option, assign a score from 1-5 on each criterion. Show your reasoning for each score in a separate column. Calculate the weighted total score for each option. Format as a markdown table.”

After Claude builds the matrix, review the scores critically. Where do you disagree with Claude’s scoring? Those disagreements are important — they reflect either new information you have that Claude doesn’t, or a bias you’re carrying that Claude’s objective scoring is revealing.

Step 5: Run a devil’s advocate analysis

A decision matrix tells you which option scores highest under your assumptions. But the most dangerous decisions are ones where the “right” answer depends entirely on assumptions that haven’t been challenged. Ask Claude to interrogate the decision.

Prompt:

“My decision matrix shows that [Option A] is the best choice. Now argue against it. What are the strongest reasons this is the wrong decision? What assumptions in our analysis could be wrong? If Option A fails in 12 months, what will the most likely cause be? What would have to be true for [Option B] to actually be the better choice?”

This devil’s advocate exercise consistently surfaces the 1-2 risks that weren’t adequately weighted in the initial analysis. It doesn’t have to change your decision — but it should inform your contingency planning and make you more confident in the choice you do make.

Prompt Template

I need to make a business decision. Here's the context:

Decision to make: [Describe what you're deciding]
Options I'm considering:
1. [Option A]
2. [Option B]
3. [Option C]
[Add more as needed]

Constraints:
- Budget: [X]
- Timeline: [Y]
- Non-negotiables: [List must-haves]
- Team capacity: [Relevant info]

Goal: [What outcome does the right decision produce?]
Who is affected: [Stakeholders and their concerns]

Please:
1. Identify any options I may have overlooked given my constraints
2. Suggest 5-7 evaluation criteria with recommended weights (must sum to 100%)
3. Build a weighted decision matrix scoring each option 1-5 on each criterion, with brief reasoning for each score
4. Calculate weighted totals and identify the top choice
5. Run a devil's advocate analysis on the top choice: what are the strongest arguments against it?
6. Summarize the recommendation in 3 sentences I could use to explain the decision to a stakeholder

Note any assumptions you're making that I should verify with my own data.

Tips & Best Practices

1. Separate option generation from option evaluation — Ask Claude to generate options in one step, and evaluate them in a separate step. When you mix the two, you unconsciously generate options that support your preferred answer. Keeping them separate produces more honest analysis.

2. Have stakeholders set the weights before they see the scores — If you’re using this framework to align a team, have each stakeholder independently assign weights to criteria before anyone sees how the options score. Comparing weights reveals real disagreements that would have stayed hidden and caused conflict later.

3. Build a sensitivity analysis — Ask Claude: “If I change the weight of [most important criterion] by plus or minus 10%, does it change the top-ranked option?” If the winner changes with small weight adjustments, the decision is fragile and you need more certainty about what truly matters most.

4. Use Claude to write the decision memo — After making your decision, ask Claude to help you write a one-page decision memo: what we decided, the options we considered, why we chose what we chose, what we’re trading off, and what would cause us to revisit this decision. This becomes invaluable documentation six months later.

5. Apply this to personal career decisions too — The same framework works for “should I take this job offer,” “should I leave this company,” “should I go back to school.” Claude is a useful thinking partner for high-stakes personal decisions where you want to think past your emotions to what actually makes sense.

Try It Yourself

Pick a decision you’re currently procrastinating on — a vendor to choose, a tool to adopt, a strategic direction to pursue. Write it down as a decision with 2-4 options. Give it to Claude with the prompt template above. When Claude builds the matrix, look at the criterion where your preferred option scores lowest. Ask Claude: “What would have to be true about [weakest criterion] for [my preferred option] to still be the right choice?” If the answer requires a lot of things to go right, you may be rationalizing rather than deciding. That insight alone is worth the exercise.