Version history
Every change, and whether it broke comparability.
A longitudinal measure is worthless if the method quietly changes underneath it. Each entry states plainly whether scores either side of the change can still be compared. That single field is what a report covering two dates depends on.
| Version | Change | Comparable? |
|---|---|---|
| v1.4 current | Kernel composition: single canonical output shared by report, share link, email and PDF surfaces | Yes, no scoring change |
| v1.3 | changelog to be reconstructed from source | to confirm |
| v1.2 | changelog to be reconstructed from source | to confirm |
The gaps are real and stay visible until the history is reconstructed properly from the source. A changelog invented after the fact would defeat the purpose of publishing one.
When a peer benchmark gets published
Not until there’s a defensible cohort: provisionally at least 100 comparable observations per segment, where a segment is one business model at one maturity level, with the exact threshold set against the observed distribution on statistical advice. At 31 companies measured in total, no segment is close, so no percentile is shown anywhere in the product. We could display one today and it would be fiction.