ENGINE QUALITY
Timory samples three layers every week: whether strong evidence is truly strong, whether ordinary signals hide payment evidence, and whether no-signal content was missed entirely. Only conclusive human verdicts enter these rates — each n below is that rate's own denominator, and reviews marked unsure are listed separately instead of being counted either way.
STRONG EVIDENCE
Precision
Of conclusively judged current-spend and explicit-WTP labels, how many were accurate. Unsure reviews sit outside the denominator.
ORDINARY SIGNALS
Strong-label miss rate
How often a conclusively judged ordinary signal should have been labelled as payment evidence.
NO-SIGNAL CONTENT
Signal miss rate
How often conclusively judged content with no extracted signal actually contained one.
A rate is published only once that rate's own denominator reaches 5 conclusive verdicts. Below that the week still shows its count, because one or two verdicts move a percentage further than the underlying quality does.
2026-10-12
0 / 30 reviewed
2026-10-05
0 / 30 reviewed
2026-09-28
0 / 30 reviewed
2026-09-21
0 / 30 reviewed
2026-09-14
0 / 30 reviewed
2026-09-07
0 / 30 reviewed
2026-08-31
0 / 30 reviewed
2026-08-24
5 / 30 reviewed
2026-08-17
4 / 30 reviewed
2026-08-10
5 / 30 reviewed
2026-08-03
7 / 30 reviewed
2026-07-27
30 / 30 reviewed
These summaries disclose only source family and provenance changes. Reviewer notes, authors, excerpts and internal identifiers never appear here.
Follow evidence changes, including corrections
Weekly movement · source-linked dossiers · explicit limitations