Every turnaround ends with a lessons-learned deck, and every deck has a KPI slide: schedule adherence, scope growth percentage, rework rate, inspection completion. The numbers are presented to two decimal places, and almost nobody in the room can say where they came from.
Usually they came from a planner's spreadsheet, assembled in the week after startup from memory, daily reports, and negotiation. That does not make the numbers wrong, but it makes them unfalsifiable, and unfalsifiable numbers cannot improve the next turnaround, because nobody can drill into them to find out what actually happened.
The test: can you click through?
A useful discipline for any turnaround metric: for every number on the slide, someone should be able to ask "show me," and the answer should be a list of records, not an explanation of methodology.
- Scope growth of 14 percent should decompose into the added work orders, each linked to the finding that justified it, each finding carrying its photo and its disposition.
- Inspection completion of 98 percent should be a count of planned inspection tasks with results attached, over planned tasks, and the missing 2 percent should be nameable: these tasks, these assets, this reason.
- Rework rate should trace to repair records that failed their acceptance and were reworked, not to an end-of-job survey of foremen.
- Closeout duration should run from mechanical completion to the date the last record carried its final sign-off, timestamps the system recorded, not dates someone remembered.
If a metric cannot survive the click-through, it belongs in the narrative section, honestly labeled as an estimate.
Provable KPIs are a capture problem, not a reporting problem
The reason most turnaround KPIs are hand-assembled is not that reporting tools are missing. It is that the underlying events were never captured as records. You cannot count findings that lived in daily-report prose, or trace scope additions that arrived by radio and left no trail, or timestamp sign-offs that happened on paper in a trailer.
Which means the KPI question is settled months before the closeout deck, in how execution is recorded:
- Findings are records with assets, timestamps, and evidence, so discovery metrics are queries.
- Scope additions link to findings, so growth decomposes into causes instead of arriving as a lump.
- Sign-offs are captured events, so completion and closeout dates are facts with timestamps rather than recollections.
- Vendor results import as data, so inspection completion counts actual results received, not reports promised.
Do those four things during execution and the KPI slide becomes an export. Skip them and no reporting tool can help, because the raw material does not exist.
Fewer numbers, harder ones
A side effect of the click-through standard is that the KPI list gets shorter, and that is an improvement. Three numbers that decompose into records beat twelve that decompose into meetings. The ones worth keeping are the ones that drive next-outage decisions: where discovery concentrated, which crews' work required rework and why, how long each system took from mechanical completion to signed closeout.
The last one deserves particular attention, because closeout duration is the metric that best predicts whether your equipment records can be trusted for the run cycle. A closeout that took one week produced records assembled while memories were fresh. A closeout that took four months produced reconstructions.
Credibility compounds
The first turnaround that reports provable numbers changes the conversation with management. When the scope-growth figure can be walked, finding by finding, the discussion moves from defending the number to acting on it. That credibility is what funds the next improvement.
intelliPROJECTS's execution records, findings, linked scope, and timestamped sign-offs, are the raw material these metrics query against, so the KPI slide is derived rather than assembled. Whatever system you use, apply the click-through test to your last closeout deck. The gaps it exposes are next outage's capture plan.
