Scoring System

Every ThinkReview analysis includes four metrics scored from 0 to 100. These scores provide a high-level health check for your code changes.

Metric	What it measures
Overall	A holistic assessment of the code’s health and readiness.
Code Quality	Clarity, maintainability, and structural integrity.
Security	Identification of potential vulnerabilities and risks.
Best Practices	Adherence to language idioms and industry standards.

How Scores are Calculated

Our AI analyzes your code based on standard software engineering principles, but you can also influence the focus of the review through your settings.

Security Score

The Security score is highly sensitive to potential risks:

• Starting Point: Every review starts at 100.
• Deductions: We deduct 15 points for every unique security issue identified (e.g., SQL injection, exposed secrets, insecure logic).
• Criticality: A security score below 70 should be considered a “Blocker” for merging.

Custom Rules calibration

In the portal under Custom Rules, you can select specific Security Checks and Best Practice Guidelines. See Custom Rules for setup.

When these are selected, ThinkReview:

1. Injects specific instructions into the AI’s review prompt.
2. Prioritizes those concerns in findings (scoring and severity layouts).
3. On scoring layout, applies stricter scoring penalties if those specific rules are violated.
4. On severity layout, treats matching violations as higher severity when warranted.

Understanding the Ranges

• 85 - 100 (Excellent): The code is solid and adheres to best practices. Ready for merge.
• 70 - 84 (Good): Minor improvements are suggested. Merging is usually safe but review the feedback.
• 50 - 69 (Needs Work): Contains noticeable issues that should be addressed before the PR is approved.
• 0 - 49 (Critical): Significant security risks or poor code quality detected. Action is required immediately.

Pro Tips for Better Scores

1. Fix Security First: A single security vulnerability can tank your overall score. Address these findings immediately.
2. Read the Feedback: Scores are just indicators. The real value is in the Suggestions and Best Practices lists provided with the review.
3. Calibrate: Use Custom Rules in the portal to tell ThinkReview what your team cares about most. This makes scores (and severity findings) more relevant to your project.
4. Iterate: After fixing issues, use the Refresh button or push a new commit to see your scores improve.

---

TL;DR: Scores (0-100) help you quickly judge code health. Security issues carry heavy penalties (-15 pts each). Calibrate what the AI looks for with Custom Rules so scoring-layout scores and severity-layout findings match your team’s bar.