The items that sink Agile programmes are rarely technical and almost always late. This is what to settle in each phase, who owns it, and which items block the phase after if they slip.
Across Agile programmes the same five things slip, and all five are decisions rather than build tasks. Each one blocks work downstream, and each is cheap to settle early and expensive to settle late.
First, what Page Two and Page Three fields mean. Nobody can map Text14 until somebody says what it holds, and the person who knows has often left. Second, the migrate-or-archive boundary, which determines volume and therefore the schedule. Third, retention obligations, which quality and legal have to state before the archive can be designed rather than after.
Fourth, item numbering. Whether numbers change in the target is a business decision with consequences for structures, AML, changes, attachments and every downstream system that references a part. Deciding it late invalidates work already done. Fifth, attachment scope, because vault volume sets the critical path and the answer is rarely all of it.
None of these is a build problem. All of them stall the build if they are open, which is why the checklist assigns each an owner and flags whether the next phase can start without it.
What has to be true before each phase can be called complete.
Agile version and configuration documented, object volumes measured, custom classes and flex fields inventoried, vault sized, data-quality register opened.
Class, lifecycle and value crosswalks drafted with every source value accounted for, flex fields resolved, exceptions decided and the mapping signed and versioned.
Read-only access in place, full baseline extracted and signed, checksums recorded, delta mechanism proven.
Full-volume run completed, dependency failures resolved, stage durations measured, attachment strategy proven.
Counts matching by object family, exception register trending down, business validation completed on representative products.
Freeze scope agreed, go and no-go criteria written, rollback tested, hypercare staffed, decommissioning plan agreed with quality and legal.
Each phase inherits the unfinished business of the one before it.
Establish who owns the flex-field meanings, who states retention obligations, and who can decide item numbering. Named people, not teams.
Profiling complete, so mapping works from measured attribute usage rather than from the data dictionary.
Mapping signed and versioned, target configuration confirmed, and the migrate-or-archive boundary drawn.
Full extract baselined and signed, so rehearsals compare against a fixed reference rather than a moving source.
Three full-volume rehearsals, reconciliation clean, exceptions closed or accepted in writing, rollback exercised.
Most checklist items that slip are decisions rather than tasks, and decisions slip when they belong to a team rather than a person. Assigning a name and a date is usually enough to unblock them.
These six ownerships are worth agreeing in the first fortnight, before anyone starts building.
Three common patterns, all avoidable.
Mapping stalls while the business works out what a field means, and the answer often needs someone who has left.
The vault turns out to set the critical path, after the schedule was built around item counts.
The migrate-or-archive boundary moves after mapping is signed, and the work already done has to be revisited.
Seven phases, each with entry and exit criteria and a named owner per item. Discovery documents version, configuration, volumes, custom classes, flex-field usage and vault size. Mapping produces signed crosswalks and resolves the attributes with no natural target. Extraction establishes a signed baseline and a proven delta mechanism. Trial migration runs at full volume and measures stage durations. Reconciliation closes the exception register. Cutover has written go and no-go criteria and a tested rollback. Decommissioning has retention obligations agreed with quality and legal.
The meaning of the Page Two and Page Three attributes. Every long-lived Agile instance carries business-critical data in flex fields named generically, and the documentation is usually the institutional memory of one or two people. Mapping cannot proceed without it, so the work of finding out belongs in discovery, with a named owner and a date, not in the mapping phase where it becomes a blocker.
In discovery, before the archive is designed and before the migrate-or-archive boundary is drawn, because both depend on the answer. Quality, legal and compliance need to state which regimes apply to which data and for how long. Leaving it until decommissioning is the classic failure: the archive is built, then someone discovers a regime requires a longer period or a residency constraint the design does not satisfy, and the work is redone.
Start from operational need rather than from what is technically possible. Product lines still in production, items with open changes, current structures and current approved sources have to be in the new PLM. Superseded revisions, closed changes, retired products and their documents generally do not. Attach volumes to each side before deciding, because the cost difference is substantial, and record the boundary so it is not reopened mid-build.
Engineering for item, structure and change semantics and for flex-field meanings. Quality for retention obligations and validation acceptance. Procurement for manufacturer and AML decisions. IT for access, environments and infrastructure. Legal or compliance for retention and any export-control constraints. The commitment is modest in hours but concentrated at the start, and programmes that defer it pay for the deferral later.
Allow six to nine months from first assessment to go-live for a single instance, of which the technical programme is twelve to twenty-four weeks. The extra time is for the decisions: flex-field meanings, retention obligations, scope boundary and numbering. Organisations planning around the December 2027 Premier Support change should work backwards from that with a comfortable margin, since the work is far easier while the people who know the system are still available.
We will run it as a short assessment, tell you which items are already settled, which are open, and which of the open ones are blocking the phase after.