Recall scope
When a material lot is found out of specification, the affected work orders, finished batches, shipments, and customers are identified and listed in full by level.
Knowledge graph analysis
Recall scope, root cause, and change impact, defined from one set of relationships
The decisive questions in regulated manufacturing cross multiple systems and tables. PhaseNext holds the relationships between material lots, work orders, finished batches, shipments, equipment, and SOP versions as governed data, so four classes of quality and change control questions are answered from one basis.
Book a demoFour questions
Four decisive questions in quality and change control
Root cause analysis
Across a period of out-of-specification batches, the shared equipment, SOP versions, and operators are identified and ranked by how widely they are shared.
Change impact assessment
Before an SOP is revised, the affected equipment qualifications, training records, open work orders, and change projects are listed as the basis for change control review.
Cross-system identity
Where a supplier is registered under different names in ERP and QMS, the records resolve to a single entity so a query covers the full scope.
Query and result
Results are presented by level and traced back to source records
A query starts from one record and expands by direction and level, with relation types available as a constraint. The result provides both the relationship view and the complete list, and each entry carries its level and a link back to the original record for verification.
Audit completeness
The list is the complete set
- The list is paginated in full and grouped by database, with the level at which each record was reached. Display limits apply to the relationship view alone.
- A record reached through several paths is counted once, so the total does not inflate where paths converge.
- Rework and other cyclic relations terminate safely.
Audited manufacturing requires the complete recall set. Silent truncation counts as a defect.
Governance
Identity resolution is a governed change
Candidate pairs reviewed individually
Candidates are scored across mapped fields and split into matched, ambiguous, and rejected bands, with each pair approved or returned by a reviewer. A merge writes a link and preserves the original records.
Two-person approval
A job advances through draft, preview, confirm, and apply. Apply is a manual action, and the confirming user differs from the creator.
Reproducible as of a date
Resolved identities carry versions and effective dates. A query run against an earlier date returns the membership frozen at that point, so the same recall returns the same result on any date.
Permission and AI boundaries
Permission limits are stated in the result
Queries apply workspace and database permissions at every level. Branches beyond a permission boundary stop at that line and the result is marked incomplete, with the identity, count, and structure on the far side withheld. A formal recall report is run by a user with full access. The AI entry point turns a question into query conditions, the deterministic engine produces the result, and every conclusion in a summary carries a relation path that can be checked.
Traceability demo