On 2026-05-04, my benchmark harness printed its verdict on two Serverless Framework caching plugins: mine scored 0.88; the community incumbent, 0.3025. Every digit of that gap is arithmetic — seven dimensions, each one measured, normalized, weighted, summed. The raw JSON actually stores my side as 0.8800000000000001, kept as-is: IEEE 754's reminder that a number can be very precise about being an opinion. Because the most consequential input in that file was never measured by any instrument: Lifecycle Correctness is worth 25% of the score because I said so.

A composite score is opinions wearing a number costume — the general argument, and the honesty requirements any composite owes its readers, live in Composite Scores and Weighting, and I won't re-derive them here. What I owe you instead is the costume's pattern, dimension by dimension — including whether the verdict survives when you attack the weights.

What are the seven dimensions — and where does each number come from?

Dimension

Weight