Information Richness Rating (A/B/C) in AI Investment Analysis: The ai-berkshire Framework

The Information Richness Rating is a three-tier classification (A, B, C) used throughout the ai-berkshire repository to indicate the reliability and data quality backing AI-generated investment insights, with A representing high-confidence verified data and C indicating sparse, assumption-heavy analysis.

The ai-berkshire repository implements a rigorous transparency system to combat overconfidence in AI-generated financial research. At the core of this system lies the Information Richness Rating, a standardized A/B/C classification that communicates data quality to human reviewers. This rating appears in every automated research report produced by the framework, ensuring users understand the evidentiary foundation behind each investment recommendation.

Understanding the A/B/C Rating Tiers

Rating A: High-Confidence, Data-Rich Analysis

A-rated analyses draw from abundant, high-quality primary data sources including multiple audited filings and cross-validated metrics. The underlying numbers undergo verification through tools/financial_rigor.py, yielding conclusions with high confidence. Reports bearing this rating explicitly state that the analysis is "information-rich" and can be quoted with strong conviction.

Rating B: Moderate Data with Confidence Flags

B-rated insights rely on moderate data availability—figures may be older, limited to a single source, or contain estimations. While the AI still executes verification steps via financial_rigor.py, the system appends a confidence flag to relevant metrics. Users should treat B-rated insights as plausible but verify figures independently before making decisions.

Rating C: Sparse Data Requiring Manual Validation

C-rated analyses indicate very sparse data environments where only a few public data points exist or heavy assumptions are required. The AI tags these reports with a low-confidence disclaimer and explicitly recommends further manual validation. These statements function as exploratory thoughts rather than actionable investment advice.

Implementation in the ai-berkshire Codebase

The rating system is hardcoded into the repository's documentation and skill files to ensure consistent application across all generated reports.

In README_EN.md (line 93), the rating appears as part of the bias-awareness mechanism: "Information Richness Rating (A/B/C)". The skills/investment-research.md file (line 207) mandates that every report begin with the rating: "报告开头必须包含'信息丰富度评级'（A/B/C）". Similarly, codex-skills/investment-research/SKILL.md (line 222) enforces this requirement for Codex-compatible commands: "must contain '信息丰富度评级' (A/B/C)".

Determining Ratings with financial_rigor.py

The tools/financial_rigor.py verification engine provides the numeric foundation for rating assignments. The tool validates market capitalization, valuation metrics, and other financial data against reported figures to determine the appropriate classification.

Assign ratings based on these verification outcomes:

• A: All financial_rigor.py checks pass with less than 5% deviation
• B: Some fields rely on estimated or older data
• C: Many fields are missing or require heavy estimation

Practical Code Examples

Here is the standard report template incorporating the Information Richness Rating:

## Investment Research – {Company Name}

**Information-Richness Rating:** {{RATING}}   <!-- replace with A, B, or C -->

*AI-Generated Insight ({{RATING}} confidence). All key numbers have been verified with `financial_rigor.py`.*

Invoke the verification tool before assigning a rating using these commands:

# Verify market cap data

python3 tools/financial_rigor.py verify-market-cap \
    --price 45.12 --shares 2.5e9 --reported 1.13e11 --currency USD

# Verify valuation metrics  

python3 tools/financial_rigor.py verify-valuation \
    --price 45.12 --eps 3.24 --bvps 12.5

After running these checks, assign the rating based on data completeness and deviation thresholds.

Summary

• The Information Richness Rating uses A/B/C tiers to classify data reliability in AI investment analysis within the ai-berkshire framework.
• A-rated reports contain abundant, cross-validated primary data verified by financial_rigor.py with high confidence.
• B-rated analyses indicate moderate data quality requiring confidence flags and user verification.
• C-rated insights signal sparse data with low-confidence disclaimers, functioning as exploratory rather than definitive research.
• The rating system is enforced across README_EN.md, skills/investment-research.md, and codex-skills/investment-research/SKILL.md to ensure consistent transparency.

Frequently Asked Questions

What determines whether an analysis receives an A, B, or C rating?

The rating depends on data quality and verification results from tools/financial_rigor.py. An A rating requires multiple audited sources with less than 5% deviation in verification checks. A B rating applies when data is limited to single sources or contains estimations. A C rating indicates sparse data points requiring significant assumptions.

Where is the Information Richness Rating displayed in ai-berkshire reports?

According to skills/investment-research.md (line 207) and codex-skills/investment-research/SKILL.md (line 222), the rating must appear at the very beginning of every investment research report, immediately following the company name header.

How does financial_rigor.py support the rating system?

The tools/financial_rigor.py script validates financial metrics like market capitalization and valuation ratios against reported figures. It provides the quantitative verification necessary to distinguish between high-confidence A ratings and lower-confidence B or C classifications.

Can a C-rated analysis still be useful for investment decisions?

While C-rated analyses carry low-confidence disclaimers and require heavy assumptions, they serve as valuable exploratory tools for identifying potential opportunities in data-sparse environments. However, the ai-berkshire framework explicitly recommends manual validation before acting on C-rated insights.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}