Robbins Analytics.The Cascadia portfolio · Evidence inventory
What each build could check.
What the evidence can actually say.
Eleven independent portfolio builds. Fifteen test surfaces. A way to trace the controls each project carried, with failures and limits kept in view.
Independent projects, in first-commit order
The original fourteen, plus expected event accounting
Expected event accounting
The comparison
Compare all eleven builds, or narrow the rows by build and the columns by kind of check. Column numbers identify definitions, not scores.
Activate a filled cell to open its recorded artifact or verification summary directly. Column headings open definitions. On small screens, scroll the matrix; headers and build names stay visible. Artifact recorded is not a pass rating.
| Build / first commit | Data | Numbers | Presentation | Operation | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 01Medical Devices2026-06-11 | ● | − | − | ● | − | − | − | − | − | − | − | − | − | − | ? |
| 02Pharmacy2026-06-29 | ● | − | − | ● | − | − | − | − | − | − | − | − | − | − | ? |
| 03Staffing2026-07-16 | ● | − | − | ● | − | − | ● | − | − | − | − | − | − | − | ? |
| 04Finance2026-07-21 | ● | ● | − | − | ● | − | − | − | ● | ● | ● | − | − | − | ? |
| 05Deal Desk2026-08-01 | ● | ● | − | ● | ● | − | − | − | − | ● | ● | − | − | − | ? |
| 06Control Tower2026-08-08 | ● | ● | ● | ● | ● | ● | − | − | ● | ● | ● | ● | − | − | ? |
| 07Matter Ledger2026-08-26 | − | ● | ● | − | − | ● | ● | − | − | ● | ● | ● | ● | ● | ? |
| 08Fee Examiner2026-09-09 | − | ● | ● | − | − | − | ● | − | − | ● | ● | ● | ● | ● | ? |
| 09Revenue Assurance2026-09-11 | ● | ● | − | ● | − | ● | ● | ● | − | ● | ● | ● | − | ● | ? |
| 10Early Warning2026-10-06 | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ? |
| 11Claim Integrity2026-10-17 · New | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ● | ○ | ● | ● |
Showing 11 builds and 15 test surfaces.
Original ten builds: inventory frozen October 7, 2026, without a new outcome audit. Claim Integrity: repository evidence pinned to 9e21fc7; statuses come from the current scoped checklist and retained review evidence. A served snapshot can differ from a later repository artifact; the evidence register keeps both identities.
The latest addition
Claim Integrity added expected event accounting: checking which records should exist, including work that never arrived. It also introduced this inventory's first recorded explicit N/A assessment.
Expected event accounting starts from the triggering facts, rather than only the records that arrived. A missing claim or promised response remains an obligation.
The question: what was due, at which cutoff, and is it still missing?
Read the definition ↓All 98 applicable golden-fixture variants pass, with 649 of 649 assertions compared. The restricted four-seat simulated panel and chart review are complete, with six clarity fixes and no open actionable findings.
The dated prepublication checklist records 152 passes and six N/A requirements at its frozen-build scope and phase. Actual public readback later passed at publication.
Inspect the completed records ↗Scheduled-run evidence does not apply to Claim Integrity’s frozen module: it has no live data edge or continuing refresh claim. Applicability follows what a build does and claims. N/A is documented scope, not a passing test or missing evidence.
Expected event accounting is the additional inventory definition. Earlier builds have not been assessed against it. No first-ever capability claim or estate-wide standards change is implied.
A shared vocabulary
Open a definition for its purpose. The counting rules state which artifacts qualify and which do not. Limited evidence stays distinct from a completed requirement.
A documented command or ordered build sequence reproduces the published outputs.
Read the counting ruleInputs have a declared cutoff and an identifiable committed snapshot.
Read the counting ruleA named comparison checks population counts against an independent expectation.
Read the counting ruleA committed report records validation checks and their actual outcomes.
Read the counting ruleExecutable checks assert the definitions and invariants the build relies on.
Read the counting ruleA separately written derivation checks published values against the definitions.
Read the counting ruleSeparately specified cases provide known answers before implementation.
Read the counting ruleDeliberate defects show that a particular control can fire, with clean controls retained.
Read the counting ruleA recorded review checks charts against the written visualization standard.
Read the counting ruleA reading exercise records interpretations, exposure conditions and dispositions.
Read the counting rulePublished measures carry definitions, grains, denominators and a link to implementation.
Read the counting ruleA scheduled execution publishes its receipt and health, including failed runs.
Read the counting ruleComplete populations and residuals are reconciled and the differences published.
Read the counting ruleAn independent population of expected work retains missing records, due times and historical deadline outcomes.
Read the counting ruleHow far the comparison goes
Different problems need different controls. Artifact counts are not quality, maturity or business-impact scores. Older rows retain their original inventory meanings and have not received a new outcome audit.
Claim Integrity records 98/98 applicable fixture variants, 81/81 applicable public diagnostic injections and 22,173/22,173 numerical cells matching. Its chart review and four-seat restricted simulated panel are complete. The dated prepublication checklist records 152 passes and six N/A requirements at that scope and phase.
The frozen snapshot has no live edge, so scheduled operation is N/A. The panel uses simulated readers with a shared model service and workstation; certified blindness and human research are not claimed. Actual public readback later passed at publication; its prepublication N/A is not a continuing exemption.
Build-forward candidate
This candidate is outside the current test matrix for all eleven builds. Future work could use independently held faults and clean opportunities to measure misses, false alarms and detection timing across a declared roster.
Claim Integrity's exploratory unblinded attempt 2 met must-catch in 20 of 26 represented classes, with six missed targets and ten of 36 classes unrepresented. It used older controls and shared custody. No final-control effectiveness or independent evaluator custody is claimed. These records remain context for future work, not a completed current surface.
Read the candidate and retained disclosures ↗The original 10 × 14 inventory and standards history retain their October 7 freeze at 66b9de6. This is an additive October 19 extension, not a claim that all eleven repositories were audited on one date. The refreshed row reads commit 9e21fc7efa3978956e17a369cfd0c9f3755874f3. The earlier extension remains in Git history and the provenance record.
Provenance record · Original ten-build snapshot · Extended inventory · Evidence and outcomes · All counting rules
Dates not established by the original history remain unknown. This lane is historical and does not claim current chart-review completion.