Every operations manager has a story like this: the exchanger that needs scaffolding for every cleaning, the valve requiring a crane because nobody drew an access platform, the pump with no isolation for online maintenance. Each one is a change that would have cost a line on a drawing during FEED — and instead costs a shutdown, every year, forever.

Why it keeps happening

FEED is run by project teams measured on capital cost and schedule. Operability and maintainability are lifecycle costs — real, large, but invisible in the metrics that govern the phase where they are decided. Unless the owner deliberately injects the operations perspective into FEED, nobody in the room is paid to care about year five.

The cost of the same change, by project phase FEED1× Detail design3–5× Construction10–20× Operations50–100×
The same change, priced at four project stages. Influence falls as cost rises.

The owner's checklist

Before FEED closes, an owner should be able to answer yes to each of these: Has a RAM study tested the configuration against the availability target — including realistic logistics delays? Have maintainability reviews walked every major maintenance task on the 3D model: access, laydown, lifting, isolation? Are sparing and redundancy decisions backed by production-loss analysis rather than habit? Do safety-critical systems have performance standards defined now, so verification is designed in rather than retrofitted? Is the data handover (tagging, documentation, CMMS readiness) specified in the contract, per CFIHOS or equivalent?

The economics

Industry experience is consistent: a change that costs 1× in FEED costs several times more in detail design, an order of magnitude more during construction, and up to two orders of magnitude once the plant is operating — when it also brings production loss with it. An owner's engineer reviewing FEED for operability is the highest-return engineering purchase in the project.

Key takeaways

  • Lifecycle cost is decided in FEED, by people not measured on lifecycle cost — unless the owner intervenes.
  • Run RAM, maintainability and access reviews before FEED closes, not after.
  • Define SCE performance standards in design so verification is built in.
  • Specify data handover (CFIHOS) contractually — day-one operations depends on it.