A project health check for industrial capex

Most project health check templates were written for office and IT work. They ask whether stakeholders are engaged and whether the team feels aligned, and they will return a comfortable score on a plant project that is already four months behind on its tie-in window. On industrial capex the evidence sits somewhere else: in the native schedule file, in the scope definition package used at sanction, and at the interfaces between engineering centres.

What the office template misses

A generic health check measures process maturity. An industrial capital project is judged on physical commitments that cannot be rescheduled at will: long lead equipment slots, a shutdown window fixed a year in advance, permits tied to a season. Once one of those is missed, the recovery is not a matter of re-planning, it is a matter of paying for the slip somewhere else.

That cost is measurable. In the research overview of large capital projects by Bent Flyvbjerg at Oxford, nine out of ten projects run over budget, and a one year extension of the implementation phase correlates with an increase in cost overrun of 4.64 percent. Delay is not a schedule problem that turns into a cost problem later. It is priced the moment it happens, whether or not anyone has reported it yet.

Start with the schedule file, not with the team

Ask for the native schedule file, not the monthly PDF extract. Most of what a health check needs is visible in the logic, and the checks are mechanical. The US Defense Contract Management Agency assessment, documented with its thresholds, screens a schedule on points like these:

• more than 5 percent of incomplete activities missing a predecessor or a successor
• fewer than 90 percent finish-to-start relationships
• hard constraints on more than 5 percent of incomplete activities
• more than 5 percent of activities with total float above 44 working days
• any activity at all with negative float

Treat these as a screen, not a verdict. A brownfield tie-in schedule legitimately carries constraints, and a long float tail on site restoration work is not a defect. What matters is the combination. A schedule that fails the logic check and the float check at the same time is not modelling the project, and the critical path in the steering committee pack is then a drawing, not a calculation. The GAO Schedule Assessment Guide, which sets out ten best practices for reliable schedules, is the reference to hand the planner afterwards.

Then test whether the scope was ever defined

The second question is older than the schedule: was the scope defined well enough at sanction to be executed at all. The Construction Industry Institute reports that effective front end planning improves capital cost by about 10 percent, schedule by 7 percent and change orders by 5 percent, and that projects with better aligned teams during front end planning performed almost 10 percent better on cost and over 16 percent better on schedule.

In a review, the useful question is narrow. Take the five largest assumptions behind the sanction estimate and ask who has tested each one since, and when. If an assumption has no name attached to it, it is not an assumption, it is a hope, and it will surface as a change order. This is the part of a review that owners tend to skip because it feels like re-opening a closed decision. It is cheaper to re-open it now than at the first equipment delivery. Which of these questions carries the most weight depends on where the project sits in its lifecycle, and the stages are set out on our one page overview.

The interfaces nobody owns

Delay on multi-centre projects concentrates at the handovers: three engineering offices, a vendor package with its own document schedule, and a site that needs the drawings in a different order than the contract assumes. Count the open interface points that have a named owner on both sides plus an agreed date. The ones that do not are where the schedule will slip next quarter, and they are almost never on the risk register, because nobody owns them enough to raise them. Interface ownership sits inside EPC project recovery, execution strategy and governance, alongside the schedule and the scope basis.

What the result is for

A health check is only worth running if it changes something in the next two weeks. Keep it to that scale: the native schedule file, the sanction scope package, the interface register, and a short list of decisions that are overdue with the names attached to them. That is enough to tell the difference between a project that is late and a project that has lost control of its own basis. The first one needs a recovery schedule. The second one needs the basis rebuilt before any schedule is worth printing, and treating it as the first is how a project loses a second year.