Methodology

This section analyses 135,551 Metal Archives reviews spanning 2002–2026, provided as a one-time admin export. Unlike the production data (band counts by country and genre), this dataset captures reception — what the community chose to write about, when, and how they rated it.

Selection bias: reviews are voluntary and self-selected. The dataset reflects what motivates someone to write a review, not a neutral sample of musical quality. Treat all ratings as community sentiment, not objective scores.

Grade inflation: the MA 0–100 scale clusters heavily in the 75–95 range. The reviewer bucket chart decomposes this by reviewer experience level to separate site-wide drift from casual-reviewer bias.

Split releases: reviews for split releases involving bands from multiple countries are counted once per country. This matches the policy used in the production data.

Release date precision: 80% of albums have full dates; 13% have year only (format: 1964-00-00); 7% have year and month. Canon formation uses the year component only, so precision loss is minimal.

End-of-data drop-off: recent releases always appear under-reviewed because reviews accumulate over time. This is a measurement artefact, not a sign of declining interest.