F-UJI FAIR Trends
Monthly automated FAIR assessment of every registered dataset DOI — measured trends, by assessment date.
Average score by assessment run
Per-platform averages
Where the points are lost — per-metric pass rates
How this page measures
One point per assessment run. Every monthly run of the self-hosted F-UJI server appends a dated snapshot to fair-history.jsonl — per DOI: the total, the F/A/I/R split, and each metric’s earned score and verdict. This page reads the per-run aggregates. Trends here are by assessment date: the same DOIs, re-measured — so a moving average means the metadata changed, not the cohort.
Comparability. Scores are comparable only within one metric version (pinned v0.8, 17 metrics). v0.8 runs structurally low — several metrics fail on almost every real-world repository — so 35–60% is the normal band, and the interesting signal is movement, not level.
Averages weight each assessed DOI equally.
When a source fails. The history feed is the only thing this page reads. If it does not answer, a red banner says so and every figure renders as “—” with the reason attached. No number on this page is ever produced by a failed request — a missing measurement and a measured zero are different facts, and they are drawn differently.