evidence-grounded-comparison

name	evidence-grounded-comparison
description	Compare 3-5 options using evidence found in provided source documents. Use this skill when the task is to define comparison dimensions, search source material for evidence on each option, build a comparison matrix with source citations, identify which option is strongest on each dimension, and write a summary recommendation based on the completed matrix. Focus on evidence-based comparison tables built from provided documents.
metadata	{"environment":"research-information-synthesis","skill_id":"E5-LS2","short-description":"Compare multiple options using evidence from source documents","version":"1.0"}

Evidence-Grounded Comparison

Use this workflow when 3-5 options need to be compared using source documents rather than impressions. The goal is to define clear comparison dimensions, collect evidence for each option, build a matrix with citations, and then recommend an option based on the completed comparison.

Scope: evidence-based comparison matrices from provided source documents.

When to Use

Use this skill when:

• several options need to be compared side by side
• the user wants an evidence-backed table or matrix
• the comparison should cite specific source support for each claim
• the final output should include both a matrix and a recommendation

Comparison Workflow

Follow these five steps.

Step 1 — Define the Comparison Dimensions

Start by choosing the dimensions that matter for the comparison.

Examples:

• price
• performance
• features
• support

Choose dimensions that:

• are relevant to all or most options
• are discussed in the provided sources
• can be compared clearly across the options

List them explicitly before collecting evidence.

Example:

Comparison dimensions:
1. Price
2. Performance
3. Feature set
4. Support

Step 2 — Search the Sources for Evidence

For each option and each dimension, search the sources for supporting evidence.

Look for:

• specific numbers
• direct statements
• concrete feature descriptions

Examples:

• $49/month
• 99.9% uptime
• includes 24/7 phone support

Capture the evidence together with its source.

Example:

Option: Vendor A
Dimension: Price
Evidence: $49/month
Source: Pricing Guide, page 2

If no evidence is available for a cell, record that explicitly rather than guessing.

Step 3 — Build the Comparison Matrix

Put the evidence into a matrix or table.

Example:

| Dimension   | Option A               | Option B               | Option C               |
|-------------|------------------------|------------------------|------------------------|
| Price       | $49/mo [S1]            | $39/mo [S2]            | No data                |
| Performance | 50ms [S1]              | 80ms [S2]              | 45ms [S3]              |
| Support     | Email only [S1]        | 24/7 phone [S2]        | Chat + email [S3]      |

Each cell should contain:

• the evidence itself
• a source citation

Use specific numbers and concrete quoted claims when available instead of vague wording.

Step 4 — Identify the Strongest Option on Each Dimension

After the matrix is complete, determine which option appears strongest for each row.

Examples:

• lowest price wins on price
• highest performance number may win if larger is better
• strongest feature set may win on features

Write a short note explaining why that option is strongest on the given dimension.

Example:

Price -> Option B
Reason: lowest listed monthly cost at $39/month

If a row is missing important evidence for one or more options, note that the comparison is incomplete for that dimension.

Step 5 — Write a Summary Recommendation

Finish with a short recommendation based on the overall matrix.

The recommendation should:

• refer back to the evidence in the table
• note which option is strongest across the comparison
• mention any important missing cells that limit certainty

Keep the recommendation grounded in the filled matrix rather than adding new criteria that were not compared.

Basic Implementation Pattern

def comparison_matrix(options, dimensions, evidence_lookup):
    matrix = {}

    for dimension in dimensions:
        matrix[dimension] = {}
        for option in options:
            matrix[dimension][option] = evidence_lookup(option, dimension)

    return matrix

Practical Rules

Use Tables for Clear Comparison

A matrix makes it easier to compare multiple options across the same dimensions.

Cite Specific Evidence

Specific numbers and direct source-backed statements are stronger than general descriptions.

Keep Dimensions Consistent Across Options

Each option should be evaluated on the same set of dimensions whenever the data exists.

Mark Missing Data Explicitly

If a cell has no support in the provided sources, say so directly.

Base the Recommendation on the Matrix

The summary recommendation should follow from the completed evidence table, not from unsupported preferences.

Common Pitfalls

• comparing options without first defining shared dimensions
• filling cells with vague claims instead of source-backed evidence
• forgetting to cite where each data point came from
• leaving missing cells blank instead of marking them as missing
• recommending an option for reasons not shown in the matrix

When NOT to Use

• the task is to summarize one option rather than compare several
• there are not enough sources to support a multi-option comparison
• the user wants a free-form opinion instead of an evidence-backed matrix

Quick Summary

1. Define the comparison dimensions
2. Search the sources for evidence on each option and dimension
3. Fill a comparison matrix with evidence and citations
4. Identify the strongest option on each dimension
5. Write a summary recommendation based on the matrix