Delta · Q1 → Q2
What changed since the baseline.
3 finding metric(s) resolved or improved; 4 new or regressed. Metric direction is per-finding (lower is better for defect counts).
What changed: verified narratives
Every claim below is recomputed from the two corpora, not taken on faith from the generator's own ground truth.
Party-record clean-up ran, but new duplicates kept arriving
Between the snapshots, 120 duplicate organisation groups were merged (259 party records removed from the master). In the same period 50 NEW duplicate groups appeared (98 redundant records); clean-up without an intake-side control only rents cleanliness.
The stuck-matter backlog was worked down
75 matters that had been sitting for years in holding phases (Pending Close: 39, Re-Open: 36 at Q1) were closed out by Q2. Verified: 75 of 75 now show phase 'Closed'. This is the lifecycle lens's stuck-phase indicator improving for a real, checkable reason.
Taxonomy drift: the coarse type keeps absorbing matters
935 matters were re-typed between the snapshots INTO the already-bloated general dispute type (from: matter-property-third-party (217), matter-directors-officers (196), matter-cyber-incident-response (189), matter-property-first-party (172), matter-workplace-injury (161)). The coarse work type is not a one-off configuration accident; it is actively growing, which is exactly why the taxonomy drill matters.
Health score trend
Finding-level changes
| Status | Domain | Finding | Q1 | Q2 |
|---|---|---|---|---|
| regressed | Config Usage | 17 fields carry meaningful data (>= 50% fill rate). | 18 | 17 ↓ |
| regressed | Config Usage | 95 of 287 attribute fields are never populated (0% fill rate). | 80 | 95 ↑ |
| regressed | Lifecycle | Median open-to-close duration is 251 days across 19,852 closed items. | 250 | 251 ↑ |
| regressed | Usage Journey | Work is not spread evenly across roles: matter-owner accounts for 36,314 sittings against matter-partner's 333 sittings. | 36,298 | 36,314 ↑ |
| improved | Config Usage | 94 fields are populated in fewer than 5% of work items. | 100 | 94 ↓ |
| improved | Lifecycle | 1,831 of 19,852 closed items were later reopened (9.2%). | 0.093 | 0.092 ↓ |
| improved | Usage Journey | 52% of working sittings start the same way: open a matter. The next most common way in is close a matter, at 30%. | 52.4 | 52.3 ↓ |
| unchanged | Config Usage | 5 logical field names are stored under 10 distinct raw keys. | 5 | 5 |
| unchanged | Config Usage | 1 attribute stores values two different ways at once (a literal boolean AND a numeric option-set code). | 1 | 1 |
| unchanged | Lifecycle | Median close days computed for 10 work types. | n/a | n/a |
| unchanged | Lifecycle | Work items were opened across 12 distinct years. | 40,500 | 40,500 |
| unchanged | Lifecycle | 2 of 3 phases with open work have a median age at or above 365 days. | 2 | 2 |
| unchanged | Usage Journey | 4 captured requests came from the platform rather than a person (widget loads, permission probes, list re-sorts) and are held out of every journey measure. | 4 | 4 |
| unchanged | Usage Journey | 98% of sittings are a single action and out, against a median depth of 1 distinct actions. | 97.7 | 97.7 |