Designing an operator display that answers a question

The standard process graphic is a schematic representation of the equipment: vessels drawn as vessels, pipes as pipes, valves as valve symbols, with live numbers placed next to the items they belong to. It is an accurate picture of the plant and it is the wrong artefact for most of what an operator does.

The operator's question is almost never what the plant looks like. It is whether the plant is behaving normally, and if it is not, which direction it is moving and how fast. A screen full of accurate numbers requires the operator to hold the normal range for each of them in their head and perform the comparison mentally, thirty or forty times, while under time pressure.

The shift from depiction to indication

The design approach that has emerged over the last fifteen years — reflected in the ISA-101 standard and in the high-performance HMI literature that preceded it — treats the display as an instrument rather than a picture.

Team communication is influenced by emotional cues as well as formal procedures and technical information. Further details are available via this resource.

Additional industry context is available in the ISA-101 HMI standards.

The core move is to show deviation rather than value. A level presented as a number requires interpretation. The same level presented as a bar with the normal operating range marked, and the current position within it, is interpreted at a glance without reading anything.

This does not mean removing numbers. It means that the primary encoding is positional and the number is available for when precision is needed.

Colour as a scarce signal

The single most consequential decision in display design is what colour means, and the traditional answer — colour identifies equipment type and pipe contents — spends the entire visual budget on information the operator already knows.

The alternative reserves saturated colour for abnormality. Background is a mid grey. Equipment is drawn in outline. Pipes are thin grey lines. Text is dark grey. Nothing is coloured until something is wrong, at which point the coloured item is the only coloured thing on the screen and is detected peripherally without searching.

Operators moving from a traditional scheme find this jarring for a week or two and then report that they can see problems without looking for them. That transition period is real and should be planned for rather than treated as evidence that the design is wrong.

Grey is not a stylistic choice

The mid-grey background is frequently the most contested element and it has a specific rationale. A white background produces glare in a control room lit for twelve-hour shifts and reduces the perceptual contrast available for genuine indications. A black background produces the opposite problem and makes light-coloured abnormality indications extremely conspicuous, which sounds good until several appear at once.

A mid grey provides headroom in both directions: something can be made conspicuous by being darker or lighter, and neither direction is fatiguing over a shift.

Analogue indication and the normal band

The most useful single element in high-performance display design is the deviation indicator: a small horizontal or vertical bar showing the measurement's position, with the normal operating band marked, and the alarm limits marked beyond it.

It carries four pieces of information in a glance — the value, whether it is normal, how close it is to abnormal, and which direction it is heading if a trailing indicator is included — in the space a number occupies.

Defining the normal band is where the work is. It requires someone to decide, for each significant measurement, what the acceptable operating range actually is, which is frequently a question nobody has answered explicitly. That exercise has value independent of the display.

Trends belong on the screen

A number tells you where a variable is. A short trend tells you where it is going, which is the question that determines whether an operator acts now or watches.

Embedding small trends — a few hours of history, no axes, positioned next to the value they belong to — in the main display converts a screen showing state into one showing behaviour. It is a substantial change in what the operator can perceive without navigating.

The common objection is screen space, and it is usually resolved by removing decoration. A display with realistic three-dimensional vessel renderings and gradient-filled pipes has plenty of room for trends once those are removed.

Layout that reflects the process

Items should be positioned to reflect the process flow and the operator's mental model, which is usually left to right or top to bottom in the direction of material movement.

Consistency across displays matters more than the specific convention. If the feed always enters from the left and the product always leaves at the right, an operator covering an unfamiliar unit has a starting point. If each display was drawn by a different engineer, they do not.

Fewer elements, deliberately

Display density is the most common failure. Because adding an element is easy and removing one requires justification, graphics accumulate until every available measurement is present.

The discipline is to ask, for each element, what decision it supports. Measurements that nobody uses operationally belong on a detail display rather than the main one. This is uncomfortable because someone once asked for each of them.

A useful test: ask an experienced operator to point at the three things they look at first on a given display. Everything else is a candidate for demotion.

Design for the tired and the unfamiliar

Displays are designed by people who know the plant intimately and evaluated during the day. They are used at four in the morning by someone covering a second console.

That argues for readable text sizes, generous spacing, unambiguous labelling in process language rather than tag numbers, and no reliance on the operator remembering which of two similar-looking pumps is which.

The migration trap

Control system replacements almost always recreate the existing graphics as faithfully as possible, because that is the lowest-risk option and because operator familiarity has genuine value.

The cost is that a twenty-year-old design is carried forward for another twenty. A migration is the only realistic opportunity to change the display philosophy, and passing it up means the next opportunity is a generation away.

The workable compromise is a phased approach: recreate faithfully for the migration itself, then redesign one display at a time afterwards with operator involvement, so that the change is incremental and reversible rather than a single disruptive event.

Validate with operators, under load

A display reviewed by engineers in a meeting room is validated against their understanding. The test that matters is whether an operator can use it during an upset.

A structured evaluation — operators working through simulated scenarios on the proposed display while someone records every hesitation and every navigation step — finds problems that design review does not. The metric is not opinion but the number of actions required to answer a given question.

Density and the cost of an extra element

Every element on a display consumes two resources: screen area, and a fraction of the operator's scanning capacity. The second is the constrained one, and it is invisible during design because a designer examining a screen has unlimited time.

The practical consequence is that displays should be built by subtraction. Start from the decisions the operator makes on this screen, include what those decisions require, and then justify anything else individually. A display constructed by adding everything available and removing nothing will contain three or four times what is needed.

A useful discipline during design review is to ask, for each element, which decision it informs and how often. Elements that inform no decision, or a decision made twice a year, belong on a detail display.

Where the numbers should still be numbers

The argument for analogue indication does not mean removing numeric values. Some tasks require precision: entering a setpoint, comparing against a specification limit, recording a reading, communicating a value to someone on the plant.

The workable arrangement is both — the deviation indicator carrying the at-a-glance interpretation, with the numeric value alongside it in smaller type. The operator uses whichever the task requires without navigating.

Where numbers are shown, the number of decimal places should reflect the measurement's actual resolution. A temperature displayed to two decimal places from an instrument accurate to half a degree communicates false precision and makes the display harder to scan.

Units, labels and the person covering another console

Engineering units should appear on every value, in a consistent position and a consistent form. Mixed conventions across displays — some tags in bar, some in kilopascals, some unlabelled — impose a translation task and generate errors under pressure.

Labels should use process language rather than tag numbers. An operator who normally works another unit cannot interpret a screen labelled entirely by tag, and covering unfamiliar consoles is precisely the situation where the display has to carry more of the load.

The display is part of the safeguarding

Where a hazard study credits operator response as a protective layer, the display is the mechanism by which that response is triggered. A layer of protection analysis that assumes the operator will notice a condition and act is making an assumption about display design, usually without anyone having examined it.

That assumption deserves testing. Can the condition actually be detected from the display the operator will be looking at? Is it distinguishable from normal? Does detection require the operator to be looking at a specific screen, and are they?

Where the answer is unsatisfactory, the options are to improve the display or to reduce the credit taken for operator response. Both are legitimate; continuing to credit a response that the display does not support is not.