Monitoring & operations

What should an industrial dashboard show?

Make roles, freshness, alarm context and uncertainty visible.

A good industrial dashboard helps a user answer a specific operating question with dependable information. More charts, animated counters and live badges are not evidence of quality. The user should understand what a value represents, when it was measured, whether it is valid and which decision it supports.

Start with roles and tasks

An operator needs current conditions, maintenance needs event context, and management needs comparable periods. Prioritise these tasks instead of placing every signal on one dense screen. Keep the most important information visible and provide a clear route to detail.

Also decide whether the screen is for monitoring or control. A monitoring view must not acquire command permission by accident. Apply authorisation in the API, and limit users to the equipment they are allowed to access. Visual simplicity does not replace access control.

Information beside the value

Display element Context it provides
Unit and equipment Physical meaning
Source timestamp Freshness
Quality state Whether the value is valid
Selected period What the chart actually covers
Alarm state Required assessment or action

Colour must not carry the whole message. Put understandable text beside a fault or missing-data indicator. Separate poor connectivity, a sensor error and a physical alarm. Users should be able to distinguish normal, stale and unknown conditions without interpreting decorative status dots.

Illustrative maintenance flow

A user opens an active alarm, then sees the relevant equipment trend, connection history and events around its start. The time zone is visible and the period can be widened. Acknowledgement and physical recovery appear as separate events.

This flow provides more context than a single metric card, but it still does not prove the root cause. If operator or maintenance confirmation is needed, show that limitation. A graph should not make an uncertain explanation look definite.

Do not silently join a normal line across missing data. If values are aggregated, explain the resolution or provide access to the underlying detail. A long-range average can conceal a short excursion. The user should understand what the chart’s level of detail can and cannot reveal.

Keep units and scales clear when comparing equipment. Dual axes can make an apparent relationship look stronger than the evidence supports. Presentation choices should help technical interpretation, not create a conclusion that the measurements cannot establish.

Test with actual tasks

Ask users to find the last valid measurement, investigate a stop, notice a data gap and change a report period. Check narrow screens, keyboard access, visible focus and readable type. Wide tables should scroll inside their own region without making the whole page unusable.

A screenshot alone is not acceptance. Observe whether the user can reach the correct answer and whether the evidence is sufficient. Remove unnecessary elements, clarify ambiguous labels and give failure states the same attention as the normal view. The result should help someone act with appropriate confidence, including recognising when the data cannot yet support a decision.

A concrete review task is better than a visual preference

For an illustrative usability exercise, ask a maintenance user to identify the last valid reading before a fault and explain whether the data was continuous. Provide a period containing a genuine gap and a late recovered record. Observe whether the user notices both. If a smooth line leads to a confident but incorrect conclusion, the chart needs a clearer representation of coverage, regardless of how attractive it looks.

Then ask the same person to compare two machines that use different units or sampling policies. The screen should make those differences visible before inviting a direct comparison. If aggregation changes when the time range widens, identify that change. A daily average and a current measurement answer different questions; presenting both as equivalent large numbers hides the distinction.

Use a keyboard and a narrow screen for the same tasks. Filter controls need meaningful labels, active selections need an understandable state and focus must remain visible. A tooltip that appears only on hover may conceal essential information from keyboard or touch users. Put the minimum context required for interpretation in a reliably accessible part of the interface.

Keep changing data understandable

Live updates should not repeatedly move the item a person is trying to inspect. Define how a historical investigation behaves when new events arrive: a fixed selected period may be preferable to an automatically advancing view. If the user is looking at historical records, label that mode clearly so it is not mistaken for the current physical condition.

Loading, partial data, permission denial and service failure need distinct messages. An empty chart could mean no matching observations, a restricted equipment selection or a failed request. A single generic no-data label hides the next useful action. Preserve user-selected filters when a retry is possible, and explain whether displayed values are retained historical context or refreshed results.

At handover, demonstrate a normal task and a failure task for each main role. The reviewer should reach the relevant evidence without needing to know the database schema or API structure. Keep a short record of unresolved interpretation problems and improve labels or workflow before adding more metrics. Bring the recurring questions of operators, maintenance and management to the dashboard discussion. A clear design follows those decisions and makes the limits of the data visible at the point where people use it.

Which data do you need to see? Which process could work better?

Tell us about your equipment and your requirements. Let's explore a suitable approach together.

Let's talk about your project