H-01 · Methodologyarchive-recorded · snapshot 2026-09-17
Scoring Methodology: Five Dimensions & Source Tiers
How AiRecMark records what it publishes: every archive carries a five-dimension score (dimsVersion v2-5dim) plus per-field source tiers (T0–T3). This page documents the recorded scale, the tier census over all source fields, and the exact aggregate rules used across the insights batch. T1 benchmarks recorded: 3 / 448.
Key numbers
Archives
448
data/tools · N=448
Recorded dimensions
5
score.dims (v2-5dim) · N=448
Overall median
79.5
score.overall · N=448
Overall range
70.9–90.9
score.overall · N=448
Benchmark entries
3
benchmarks[] (T1) · N=448
Recorded dimension medians
score.dims as recorded; median and observed range over ${tools.length} archives
| Dimension | Field | Median | Observed range |
|---|---|---|---|
| Output Quality | quality | 80 | 64–96 |
| Feature Depth | features | 78 | 62–94 |
| Usability | usability | 82 | 62–96 |
| Performance | performance | 80 | 68–90 |
| Value for Money | value | 76 | 58–98 |
Source-tier census
count of recorded src.level values across pricing.src, specs[].src, features[].src and benchmarks[].src
| Tier | Recorded fields |
|---|---|
| T0 | 2801 |
| T1 | 3 |
| T2 | 0 |
| T3 | 0 |
Source-tier distribution
Recorded src.level occurrences across all sourced fields in the library.
Method
- Scores: score.overall and score.dims (quality, features, usability, performance, value) are quoted exactly as recorded in each archive; this library never re-scores, re-weights or blends them.
- Aggregation rules used across the insights batch: standard median (middle element, or mean of the two middle values rounded to 1 decimal), observed min/max, arithmetic means rounded to 2 decimals, and Pearson r rounded to 2 decimals — each published with its N.
- Source tiers are the archive's own src.level values: T0 (official vendor documentation), T1 (independent measured benchmark), T2/T3 (editorial or secondary). The census above counts every recorded src.level across pricing, specs, features and benchmarks fields.
- T1 benchmarks recorded: 3 / 448: only 3 archives (Aider, Claude Code, GitHub Copilot) carry benchmark entries, so no benchmark-based claims are made anywhere in this batch.
- Archives whose sources note flags pending re-verification (14) are cited as archive-recorded, never as independently confirmed.