Commissioning & Startup Readiness: A Practical Guide to Tracking Percent Complete

Every commissioning and startup program has one artifact everybody argues with: the readiness tracker. It decides what the 7 a.m. meeting talks about, what the owner hears on Friday, and whether the startup date is a plan or a hope. This guide covers how strong CSU teams structure that tracker — whatever tool it lives in — and where the common failure modes hide.

Why CSU tracking is its own discipline

A CSU program is not a generic project. Work arrives as systems and subsystems that must be walked down, reinstated, tested, and handed over in a defensible order; progress claims carry contractual weight; and the people doing the updating are in the plant, not at a desk. A tracker that ignores any of those three realities produces numbers nobody trusts — and an untrusted tracker gets abandoned for hallway conversations by week three.

The breakdown: Milestones → Activities → Tasks

The structure that survives contact with a real turnaround has three working levels under each phase (CSU, operational readiness, and so on):

• Milestones — the units the steering committee recognizes: mechanical completion of a system, loop checks complete, ready-for-startup review.
• Activities — the crew-sized bundles inside a milestone: piping reinstatement, motor solo runs, instrumentation loop checks.
• Tasks— the individual, assignable, verifiable actions: hydrotest the LP separator inlet, torque-check the flange joints, walk down P&ID sheet 14.

The rule that makes the whole thing computable: only tasks carry a manual percent complete. Every level above is calculated, never typed. The moment a milestone percent can be hand-edited, it will be — usually the night before a review.

Weighted rollups, not simple averages

Ten tasks are rarely ten equal pieces of work. Give every item a weight — its share of effort relative to its siblings — and roll up with a weighted average at each level:

activity %  = Σ(task.weight     × task.%)     / Σ(task.weight)
milestone % = Σ(activity.weight × activity.%) / Σ(activity.weight)
phase %     = Σ(milestone.weight × milestone.%) / Σ(milestone.weight)

Default every weight to 1 so an untuned tree behaves as a plain average, then differentiate only where it matters (“commissioning is 60% of this milestone”). The companion post on weighted percent complete works a full example, including the failure that makes unweighted trackers overstate readiness.

Status and percent are different questions

Percent complete answers “how much of this is done”; status answers “can this proceed”. Keep both, independently. A task can be blocked at 40% — parts on order, permit pending, weather. Blocked deserves to be first-class in a facility tracker precisely because so much field work waits on something, and the blocked list is what a good morning meeting actually works through.

Update at the workface, or the data is fiction

If progress can only be recorded at a laptop in the site office, Tuesday's numbers are typed in Thursday from memory. The tracker has to accept updates where the work happens — on a phone, with gloves, in a plant with dead zones — and reconcile them cleanly when connectivity returns. Append-only field notes (corrections are new entries, not edits) make that reconciliation boring, which is the goal.

Keep Excel in the loop — deliberately

Owners, contractors, and package vendors will not all adopt your tool, and pretending otherwise re-forks the truth within a month. Treat the spreadsheet as an interface: export in the layout your stakeholders already read, and import returned workbooks through a dry-run diff that shows every create, update, and conflict before anything commits. Round-trip beats rip-and-replace.

The habits that keep a tracker honest

• One live tracker; copies are read-only exports with a date.
• Computed rollups are never hand-editable, at any level.
• Every change carries who/what/when — an audit trail, not a mystery.
• Blocked items get a reason and a review cadence, not a color.
• Weights are reviewed when scope changes, not silently in the summary tab.

These habits are tool-independent — but they are exactly the properties we built Fieldcurve to enforce by construction rather than by discipline: computed rollups, first-class blocked status, field-capable updates, Excel round-trip, and an audit trail under everything.