Every site has a capital process. Most of them have exactly one, written for the largest project anyone there remembers, and applied to everything. The result is predictable: the big projects climb the ladder because they have the staff to, and the small ones, which are most of the projects and a good share of the money, find a way around it. They get done as maintenance. They get done as three purchase orders. They get done under a threshold that was set once and never revisited. The process is not being ignored. It is being evaded, rationally, by people who have a pump to replace and no intention of writing a basis of design for it.
The evasion is where the surprises come from. Not from the major project that had a steering committee and a gate calendar, but from the eleven small ones that never had a scope freeze, a tie-in register, or a pre-startup safety review anyone recorded.
The fix is not a shorter ladder
The instinctive fix is to write a lighter process for small projects. That produces two processes, then three, and eventually a debate about which one applies that is itself a way of evading all of them. The better answer is one work process with tier-scaled rigor: the same phases, the same gates, the same document family, deployed with fewer activities as the tier falls. A step that carries no dates for a tier simply does not deploy at it. That one mechanism is the whole difference between a major project’s hundred and fifty activities and a replacement-in-kind’s sixteen, and it means a small project and a major one can be read against the same standard, at different depths.
The industry’s own research supports the instinct. The public front-end planning practice for industrial projects exists in a small-project variant for exactly this reason: small projects need their own instrument, not a scaled-down copy of the big one.
Four tiers, one calculator
Four tiers are a good starting hypothesis, and the bands are calibrated to each site’s population rather than fixed:
- Major. Full front-end rigor, every gate, the full document set, independent review mandatory at sanction. Above roughly twenty million, with a board-level threshold higher still.
- Standard. The full gate structure with condensed deliverables. Roughly five to twenty million.
- Small. Combined definition, weeks of front end rather than months, a desk review ahead of sanction, startup readiness absorbed into the pre-startup safety review. Under roughly five million.
- Replacement in kind. Any value, provided the like-for-like test holds: form, fit, function, and materials. One combined definition-and-funding gate, standard templates, a pre-approved engineering basis, and an expedited change screening. The plan-for-plan is a checklist with owners and dates, not a countdown.
What decides the tier matters as much as the tiers. Cost is the primary driver, but a five-million-dollar project with a novel process step, six live tie-ins, and a turnaround window is not a Small project, and a calculator that only reads the budget will call it one. Seven factors, each scored one to five: total installed cost, engineering novelty, tie-ins, coupling to a turnaround window, regulatory exposure, contractor interfaces, and schedule criticality. The score assigns the tier. Two people sizing the same project land in the same place, and a factor that fires on its own, a novelty of four, say, adds its own requirement regardless of tier.
The light track has to be genuinely light
The tier that makes or breaks the whole design is the bottom one. A replacement-in-kind track that is secretly a Small project with the gates renamed will be evaded like everything else. It has to be light enough that doing it is easier than routing around it: one gate, a template, a screening record, and the two things that must never be skipped at any tier, the management-of-change screening and the pre-startup safety review.
That last point is the reason to bother. The tiers are not about paperwork. They are about making sure the safeguards that matter reach the projects that would otherwise never see them, by attaching them to a track those projects will actually use. A like-for-like pump swap does not need a basis of design. It does need someone to have confirmed it is actually like for like, and someone to have signed the pre-startup review. The light track is how those two signatures happen.
The tiering standard, the calculator, and the replacement-in-kind screening are GV-06 and FE-08 in the Capital Project Management System, and the platform deploys them: score the project, get the tier, get the ladder. The count of activities that deploys at each tier is computed from your own work process, not from a generic diagram, so the “why does this one have a hundred and fifty steps and that one sixteen” question has a real answer.