Where stated capacity is tested against operator interviews, effective utilization reliably runs below nameplate — and changeover time and maintenance windows explain more of the gap than demand does. The people who run the plant know the second number.
What does nameplate capacity actually mean?
The output a plant would produce running continuously at design specification with no changeovers, no maintenance and no quality losses. It is an engineering figure and a legitimate one — it describes what was built.
It is not a production forecast, and treating it as the base for one is where a lot of chemicals models start going wrong. The distance between nameplate and what the plant will actually make in a year is operational rather than commercial.
The figure is also produced for a purpose that has nothing to do with forecasting. Nameplate is what a plant was designed and permitted to do, which matters for engineering, for regulatory filings and for comparing assets on a common basis. Using it as a revenue input borrows a number from one context and applies it in another, and nothing in the number signals that.
How big is the real gap?
Large enough to move a valuation, and variable enough that it cannot be assumed from a sector average. Multi-product plants running frequent changeovers lose materially more than single-product continuous ones.
What matters more than the size is the cause. A gap driven by changeover scheduling is addressable with capital or planning; a gap driven by demand is a market problem the asset cannot solve, and the two get conflated in management presentations — as they are in the cost curve the asset is ranked on.
Who knows the true utilization?
Plant managers, shift supervisors and maintenance contractors. They know actual run rates, unplanned downtime and the quality losses that never appear in a capacity statement, because they are the people who deal with them — and a site visit puts you in front of them.
Commercial leadership is the wrong source here. They report capacity as specified because that is the number they were given, and the divergence is invisible from where they sit.
Access is the practical constraint rather than willingness. Current staff at a target are usually off-limits during a process, so the useful population is former operators and the contractors who serve the site, both of which take longer to reach than a management session. That argues for starting the operational workstream early rather than after the data room disappoints.
What should you ask them?
Not what utilization is, which invites the official number, but what a bad month looks like and why. Operators describe events readily and averages reluctantly, and the events are what you need.
Maintenance contractors are the most useful single source, because they see several plants and can tell you whether the pattern is normal for the technology or specific to this asset. Covering each layer of that chain is what a channel check is for.
What does this do to the model?
It moves the volume assumption, which usually flows straight to EBITDA. A model built on nameplate with a growth assumption layered on top compounds an error that was present in the first line, which is why commercial diligence tests the volume assumption before the price one.
It also changes the capex question. If the gap is changeover-driven, the investment case for debottlenecking is real; if it is demand-driven, the same capex buys nothing.
The correction also compounds through the model rather than sitting in one line. A volume assumption feeds revenue, unit cost absorption and working capital, so an error at nameplate arrives three times in the output. That is why establishing it early is worth more than refining anything downstream of it.