The procurement locked the final 10 percent retention to document upload volumes rather than demonstrated system behaviour across packages. Contractors deliver their portion complete, the dashboard flips green, and cash moves before anyone confirms the interfaces actually pass data in both directions under load.
You bought the system because the vendor showed clean status lights for each work package. What the demo never revealed is that those lights only track file presence, not whether the mechanical interlock listed in one contractor's drawings matches the electrical termination another contractor actually installed.
Commissioning teams then inherit the gap. They must reverse-engineer live tag lists against static PDFs while the asset is already under pressure to start revenue service. Every mismatch becomes an urgent variation that the original contracts never priced.
The measurement failure is simple. The PMO counts a package as commissioned the moment its own checklist is attached, regardless of whether adjacent packages have even powered up their side of the same junction. No single contractor owns the joint test, so no one runs it until after handover.
APAC projects repeat the pattern because alliance agreements still tie milestone approval to individual package evidence rather than joint performance criteria. The commercial team protects cash flow; operations later absorbs the reconciliation cost in downtime and rework.
Fix the mechanism at source. Tie every retention release to a witnessed interface test that includes at least two packages and one live control system read-back. Anything less just moves the discovery cost downstream where you no longer control the budget.