A plant manager should track 8 decision-grade maintenance KPIs across reliability, work execution and cost, using defined formulas, named owners and decision thresholds rather than reporting activity alone [S5].
The dashboard should mix leading indicators such as preventive-maintenance compliance with lagging indicators such as MTBF, MTTR and availability, with weekly operating reviews and monthly reliability reviews [S4][S5].
Define the dashboard by the decisions it must support
A maintenance KPI combines a measured metric with a benchmark, so “increase uptime by 15% over one year” is a KPI while uptime itself is a metric [S4]. A useful dashboard therefore needs a formula, data source, review frequency, owner and action threshold for every KPI, not just a chart of completed work orders.
Maintain two distinct views. The plant-manager view should cover availability, asset reliability, maintenance cost and work backlog. The maintenance-supervisor view should cover PM compliance, schedule attainment, wrench time, overdue work and estimate-versus-actual labor [S1].
Define a metric before setting a target. MTBF is operating hours divided by failure count, MTTR is total repair time divided by repair count, and availability is MTBF divided by the sum of MTBF and MTTR [S5]. Preserve the equipment class, time window and failure definition in every calculation; aggregating unrelated assets can conceal a deteriorating critical machine.
Compare the core KPI options by decision value
Use a criteria-based comparison rather than selecting the largest number of metrics. Reliability measures tell management whether assets are available, work-execution measures expose constraints in planning and labor, and financial measures reveal whether the maintenance program is consuming an unreasonable share of asset value [S4][S5].
These thresholds are reference points, not universal acceptance criteria. A plant with older assets or stricter production requirements may set different targets after establishing its own baseline, while a KPI should remain Specific, Measurable, Achievable, Realistic and Timely [S4].
Use leading indicators to find the cause of falling reliability

When availability drops below 90%, treat the symptom as a reliability failure, not as proof that technicians need more overtime. The root cause may sit in PM compliance, work planning, spare-parts availability or diagnosis time, and the corrective action depends on which upstream metric changed first [S5].
Use condition or predictive signals to create planned CMMS work rather than alerts without ownership. The model in Integrate Predictive Alerts With CMMS Work Orders addresses that handoff. Replace or escalate an automatic trigger when its threshold lacks a verified operating basis, because an unsupported threshold creates more notifications without better maintenance decisions.
Correct wrench-time losses without misreading labor utilization
Do not use total labor hours, overtime hours or closed work orders as standalone performance measures. These are activity or input measures, while MTTR and availability connect labor performance to asset outcomes [S5]. When work-order response time appears, interpret it together with MTTR because response time alone omits the repair duration.
Reject a dashboard design that rewards closing many small work orders. Such behavior can improve the closed-work-order count while leaving critical failures unresolved. Require each work-order KPI to include planned duration, actual labor, asset criticality and failure status so the manager can distinguish throughput from effective maintenance [S5].
Control backlog and maintenance cost without hiding the asset base

Calculate backlog hours with the same capacity denominator each period, and distinguish safety-critical, regulatory, corrective and discretionary work.
Do not manage this ratio alone: changes in replacement value, inventory capitalization and work-order classification can move the percentage without changing the physical maintenance burden.
Separate committed corrective work from discretionary improvement projects. A backlog review should identify whether hours are waiting for parts, permits, access, engineering decisions or technician capacity. If a work item cannot be executed within the next planning cycle, reduce scope or close it with an auditable disposition rather than leaving it indefinitely open.
Apply asset-specific limits for high-consequence equipment
Plant-wide averages are inadequate for assets with distinct failure physics. A biogas example uses runtime hours, starts per day, spark-plug resistance trends, oil-analysis scores and combustion efficiency for CHP engines, with a reported action trigger when valve-clearance drift exceeds 15% [S3]. That 15% value is specific to that asset program, not a general plant benchmark.
For gas-treatment equipment, the cited biogas dashboard flags planned H₂S scrubber media changes as outlet H₂S approaches 100 ppm, ahead of a 200 ppm alarm point, and tracks media-bed pressure drop and replacement-cycle compliance [S3]. These are asset-specific operating values and must be validated against the manufacturer’s limits, site alarm philosophy and applicable safety requirements before implementation.
Compare asset classes rather than forcing one threshold across the plant. CHP engines, digesters, scrubbers and gas-handling equipment produce different failure signals, so each dashboard tile should identify the asset class, measurement basis, control limit and corrective action [S3]. If a threshold cannot be traced to an approved procedure or verified operating condition, label it for engineering review rather than treating it as an alarm setpoint.
Build the data pipeline and review discipline

A useful dashboard should connect CMMS work orders, asset hierarchy, production downtime and, where available, condition-monitoring telemetry. The research describes weekly spreadsheets and manual aggregation as an operating weakness, while a CMMS can calculate indicators from live work-order data and support baselines [S1][S3][S4].
Implement four controls before publishing a KPI: define data ownership, validate event timestamps, lock calculation rules and identify excluded records. A reported MTBF, for example, becomes unreliable if failure count, operating hours and production-stoppage codes come from systems with different boundaries. A 5-week deployment timeline appears in one biogas case, but it follows a data audit and CMMS integration rather than a universal installation period [S3].
Run a short acceptance test on one asset class. Recalculate three KPIs manually from CMMS records, compare the result with the dashboard, and trace one alert from measurement to work-order creation. Accept the dashboard only when the formula, period, owner, threshold and corrective action are visible on the same view; otherwise the system can produce accurate-looking numbers without operational control.
Set review rules that force action
Every dashboard review should begin with threshold breaches, then move to causal analysis. For availability below 90%, review the dominant failure modes; for PM compliance below 80%, address capacity or scope; for schedule attainment below 70%, correct planning; and for wrench time below 35%, review the execution workflow [S5]. Each review should end with an owner, action and due date.
Do not repair a metric definition when the plant has an equipment or execution problem. If the data is wrong, correct the data pipeline. If the data is right and performance is inadequate, correct the maintenance plan or execution. If a failure has safety, environmental or major-production consequences, escalate it under site procedures rather than waiting for the monthly KPI meeting.
Measure whether the dashboard changes decisions by reviewing action closure, recurrence of the failure mode and movement in MTBF, MTTR or availability. Most maintenance dashboards track 20+ metrics, while eight KPIs drive actual operational change.
The next node is a one-page KPI specification listing formula, data source, threshold, owner, review cadence and corrective action for each of the 8 metrics, followed by a manual reconciliation test on one critical asset class.