Review Network Analysis

Analyze the team's review relationships to surface collaboration patterns, detect silos, and track improvement toward distributed code review.

When to Trigger

• "review network", "who reviews whom", "collaboration graph"
• "review dynamics", "review distribution", "are we siloed?"
• As part of a team health report

Data Sources

• data/YYYY-MM/{login}-reviews.json — reviews given by each person
• data/YYYY-MM/{login}-prs.json — PRs authored (to identify inbound reviews)
• team.json — team roster for member list

Process

Step 1: Load review data for the requested period

Parse the date range (default: last quarter). For each team member, load their {login}-reviews.json files for the relevant months. Extract:

• Who authored the PR being reviewed
• Filter to team-internal reviews only (both author and reviewer are in team.json members)

Step 2: Build the review matrix

Construct an N×N matrix where:

• Rows = reviewer
• Columns = PR author
• Values = review count

| Reviewer | alice | bob | carol | dave | Total Given |
|----------|-------|-----|-------|------|-------------|
| alice    |   ·   |  12 |   8   |   3  |     23      |
| bob      |   5   |  ·  |  15   |   7  |     27      |
| carol    |   9   |  18 |   ·   |  11  |     38      |
| dave     |   2   |   4 |   6   |   ·  |     12      |

Step 3: Calculate network metrics

Network density (0-1 scale):

• Count of non-zero cells / total possible cells (N×(N-1))
• 1.0 = everyone reviews everyone; low values = siloed

Collaboration score (0-1 scale):

• Based on how evenly distributed reviews are across the matrix
• Uses normalized entropy: H / H_max where H = Shannon entropy of the flattened matrix
• 1.0 = perfectly uniform; 0.0 = all reviews flow through one person

Gini coefficient (0-1 scale):

• Inequality of review distribution
• 0.0 = perfectly equal; 1.0 = one person does all reviews
• Track this over time; decreasing Gini = healthier team

Bottleneck ratio:

• Max reviews received by one person / average reviews received

• 2.0 = someone is a bottleneck

Step 4: Reciprocity analysis

For each pair (A, B), calculate:

• A→B reviews vs B→A reviews
• Ratio (>1 means A reviews B more than reverse)
• Flag highly asymmetric pairs (ratio > 3:1)

Step 5: Time-series comparison

Compare current period to previous period(s). Show:

• Network density trend (is it improving?)
• Gini trend (is distribution getting more equal?)
• New review relationships (pairs that didn't exist before)
• Disappeared relationships (pairs that stopped reviewing each other)

Step 6: Visualization

Produce an ASCII network diagram showing connection strength:

Review Network (Jan-Mar 2026)
Density: 0.86 | Collab: 0.71 | Gini: 0.32

  alice ═══════ bob        (12/5 reviews, reciprocity 2.4:1)
    │  ╲         │
    │    ╲       │
    8      3    15
    │        ╲   │
    ▼          ╲ ▼
  carol ═══════ dave       (6/11 reviews, reciprocity 1:1.8)

Use line thickness/characters to indicate volume:

• ═══ heavy (10+ reviews)
• ─── moderate (5-9)
• ··· light (1-4)
• (blank) no relationship

Output Format

# Review Network: {period}

## Network Health
| Metric | Current | Previous | Trend |
|--------|---------|----------|-------|
| Density | 0.86 | 0.71 | 🟢 +21% |
| Collaboration | 0.71 | 0.52 | 🟢 +37% |
| Gini | 0.32 | 0.49 | 🟢 -35% |
| Bottleneck | 1.4 | 2.3 | 🟢 -39% |

## Review Matrix
[matrix table]

## Reciprocity (flagged pairs)
[asymmetric pairs with ratio > 3:1]

## Network Changes
- New connections: [pairs]
- Strengthened: [pairs with >50% increase]
- Weakened: [pairs with >50% decrease]

## Network Diagram
[ASCII visualization]

Interpretation Guide

• Improving network: density increasing, Gini decreasing, fewer bottleneck flags
• Siloed team: density < 0.5, Gini > 0.5, one person with bottleneck > 3.0
• Healthy team: density > 0.7, Gini < 0.35, all pairs have some bidirectional flow