Performance Measurement, Cost Control and Decision Analytics

Introduction

Industrial safety and maintenance management require more than recording incidents, closing work orders, or comparing expenditure with budget. Managers must determine whether assets are available, whether work is performed safely and effectively, whether resources are used productively, and whether present decisions reduce future operational risk. Performance measurement converts operational activity into evidence; cost control ensures that resources are used economically; and decision analytics connects evidence to action.

A mature measurement system combines leading indicators, which signal conditions that may produce future outcomes, with lagging indicators, which record outcomes that have already occurred. Maintenance indicators such as overall equipment effectiveness (OEE), mean time between failures (MTBF), schedule compliance, wrench time, backlog, and maintenance cost ratios should not be interpreted in isolation. A high schedule-compliance percentage may conceal poor planning, excessive emergency work, or unsafe shortcuts. Similarly, low maintenance cost may indicate efficiency—or dangerous deferral of essential work.

This chapter develops an integrated approach to dashboard design, indicator selection, data quality, trend and Pareto analysis, cost-risk prioritization, budgeting, and management decisions.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Distinguish leading, lagging, preventive, predictive, and outcome indicators.
2. Calculate and interpret safety, reliability, OEE, downtime, schedule-compliance, wrench-time, and maintenance-cost measures.
3. Design a balanced safety and maintenance dashboard.
4. Apply trend analysis, Pareto analysis, and cost-risk analysis to operational problems.
5. Evaluate maintenance budgets and compare alternative interventions.
6. Recognize limitations, data-governance risks, and ethical issues in performance reporting.

1. Performance Measurement as a Management System

A performance measure is a defined method for comparing actual results with a reference such as a target, standard, baseline, or legal requirement. A metric is the numerical expression of the measure; an indicator is a metric selected because it provides useful evidence about performance or risk; and a key performance indicator (KPI) is an indicator judged important enough to influence management attention or decisions.

A sound measurement system follows a causal logic:

conditions and actions → equipment and human performance → operational outcomes → business consequences

For example, competency assessments, toolbox talks, and overdue safety-critical inspections are leading indicators. Near misses and unsafe-condition reports are intermediate indicators. Lost-time injuries are lagging indicators. Production loss, compensation, and reputational damage are consequences. The objective is not to maximize the number of reported indicators, but to identify a small set that supports control of important risks.

1.1 Leading and lagging indicators

Leading indicators should be valid, controllable, timely, and resistant to gaming. Counting safety observations alone can encourage superficial reports; the quality and closure of observations are more informative. Lagging indicators remain essential because they test whether the control system actually produces safe and reliable outcomes.

1.2 The balanced measurement principle

A balanced dashboard normally includes five perspectives:

• Safety: exposure, control effectiveness, incidents, and learning actions.
• Reliability: failure frequency, restoration time, availability, and repeat failures.
• Work management: planning, scheduling, backlog, and execution quality.
• Asset and production performance: OEE, throughput, quality losses, and downtime.
• Financial performance: labour, materials, contractors, energy, inventory, and risk-reduction value.

Metrics must have an owner, a definition, a data source, a reporting frequency, a target or control limit, and a specified response when performance deviates.

2. Core Safety and Maintenance Indicators

2.1 Incident frequency and severity

A commonly used total recordable incident rate is:

where N is the number of recordable injuries during the period and H is total hours worked. The constant 200,000 represents 100 full-time employees working 2,000 hours per year. Some jurisdictions and organizations use 1,000,000 hours instead; the chosen base must always be stated.

A lost-time injury frequency rate may be expressed as:

where L is the number of lost-time injuries, K is the selected exposure base, and H is hours worked. These rates are useful for comparison only when definitions, exposure bases, reporting rules, and workforce composition are comparable. A rate is not a probability that a particular worker will be injured.

Severity can be measured as lost days per hours worked:

where D is lost or restricted-work days. Incident rates should be accompanied by narrative investigation, because a low-frequency event can have catastrophic consequences.

2.2 Maintenance work and backlog

Planned work has a defined scope, labour estimate, materials, procedures, and scheduling requirements before execution. Scheduled work has been assigned to a specific time period. Emergency work is work requiring immediate action to protect people, the environment, production, or the asset.

Useful indicators include:

• planned work percentage;
• emergency-work percentage;
• preventive-maintenance compliance;
• percentage of work orders with a failure code and cause code;
• overdue safety-critical tasks;
• backlog by age and criticality;
• repeat-work percentage; and
• percentage of work completed with a verified quality check.

Backlog should be expressed in labour hours, not merely work-order count. A backlog of 100 minor inspections is not equivalent to 100 major overhaul tasks. It should also be stratified by risk and age.

2.3 Schedule compliance

Schedule compliance measures completion of work planned for a defined period:

The organization must define whether partially completed jobs, approved deferrals, and emergent work are included. Removing difficult jobs from the schedule can artificially improve the result. A useful companion metric is schedule loading, which compares scheduled hours with available craft hours. Excessive loading creates chronic non-compliance, while very low loading may indicate weak planning.

2.4 Wrench time

Wrench time is the proportion of paid or available craft time spent directly executing maintenance work at the asset. A basic form is:

The numerator should exclude travel, waiting for permits, tools, materials, information, access, and supervision. Definitions vary, so comparisons between sites require a common work-sampling method. Low wrench time is not automatically evidence of poor worker effort; it often reveals poor stores control, isolation delays, access problems, or inadequate planning. Increasing wrench time without improving safety or job quality can be harmful.

2.5 Reliability, maintainability, and availability

For repairable equipment, mean time between failures is commonly calculated as:

Mean time to repair is:

The simple steady-state inherent availability relationship is:

This assumes repeated cycles, stable average failure and repair behaviour, and that the times represent the relevant operating and repair definitions. It excludes or treats separately logistics, administrative, and preventive-maintenance downtime. Operational availability may include these additional delays and therefore is not interchangeable with inherent availability.

2.6 Downtime

Downtime should be classified by cause and responsibility. Typical categories include equipment failure, planned maintenance, changeover, material shortage, quality hold, staffing, utilities, and external interruption. A downtime percentage can be calculated as:

The denominator must be defined. If the plant is not scheduled to operate during weekends, those hours should not be treated as avoidable downtime. Lost production hours should be reconciled with production records and maintenance work orders.

2.7 OEE

Overall equipment effectiveness decomposes productive performance into availability, performance, and quality:

where each component is expressed as a fraction between 0 and 1.

Thus, equivalently:

The formulas assume consistent units, such as minutes per part and total minutes. Planned stops are excluded from planned production time, while failures, setup, and adjustments are normally included in losses within the OEE framework. OEE is a focused production effectiveness measure, not a complete measure of asset health, safety, cost, or customer service. It can be distorted by changing the ideal cycle time, redefining planned stops, or producing excess inventory.

Worked example. A line has 480 minutes of planned production time. It operates for 420 minutes, produces 800 units, and has an ideal cycle time of 0.5 minutes per unit. There are 760 good units.

Availability = 420 ÷ 480 = 0.875 or 87.5%.

Performance = (0.5 × 800) ÷ 420 = 0.9524 or 95.24%.

Quality = 760 ÷ 800 = 0.95 or 95%.

OEE = 0.875 × 0.9524 × 0.95 ≈ 0.7917, or 79.17%. The loss tree indicates that availability is the largest of the three percentage losses, but the financial priority still depends on the cost and controllability of each loss.

3. Maintenance Cost Control

Maintenance cost control is not the same as indiscriminate cost reduction. Cost control seeks the required level of safety, reliability, and production performance at the lowest justified life-cycle cost.

A maintenance cost structure may include:

• internal labour and overtime;
• spare parts and consumables;
• external contractors;
• inspection and condition-monitoring services;
• tools, software, and training;
• inventory holding and obsolescence;
• energy and utilities associated with inefficient assets; and
• production losses caused by maintenance or failure.

A basic maintenance cost ratio is:

The base may be production value, sales, operating cost, replacement asset value, or units produced. Because these bases answer different questions, the chosen denominator must be reported. A cost per unit is:

A useful life-cycle view is:

where LCC is life-cycle cost, Cₐ acquisition cost, Cᵢ installation and commissioning cost, Cₒ operating cost, Cₘ maintenance cost, C_d disposal and decommissioning cost, and S residual or salvage value. If costs occur at different times, a discounted cash-flow model should be used rather than simple addition.

Budget variance is:

A positive value indicates overspending when cost is the measure. Variance analysis should separate price, quantity, timing, scope, volume, and failure-related effects. A favourable variance may be false economy if safety-critical work has been deferred.

3.1 Budgeting approaches

Historical budgeting adjusts the previous budget and is simple but may preserve waste. Zero-based budgeting requires justification of activities and resources, but can be administratively demanding. Risk-based budgeting allocates resources according to consequence, likelihood, uncertainty, and control effectiveness. Activity-based budgeting links cost to work packages, failure modes, inspections, and production plans.

A robust budget contains:

1. statutory and safety-critical work;
2. preventive and predictive work based on failure behaviour;
3. planned corrective work;
4. realistic shutdown and turnaround scope;
5. spare-parts and contractor assumptions;
6. contingency for credible risks; and
7. explicit consequences of deferral.

4. Dashboard Design and Data Governance

A dashboard should support a decision, not merely display data. Each KPI should have a documented specification containing its name, purpose, formula, numerator, denominator, exclusions, source system, owner, frequency, target, tolerance, and escalation rule.

A practical dashboard architecture has three levels:

• Strategic: safety risk, asset health, life-cycle cost, major trends, and investment decisions.
• Tactical: backlog, schedule compliance, OEE losses, contractor performance, and budget variance.
• Operational: today’s permits, overdue tasks, active failures, parts availability, and safety observations.

Use traffic-light colours cautiously. A green result should mean that defined conditions are satisfied, not merely that the number is aesthetically pleasing. Every red indicator should lead to an assigned action, owner, due date, and verification method.

Data quality dimensions include accuracy, completeness, timeliness, consistency, validity, and traceability. Common problems include duplicate work orders, missing failure codes, inconsistent downtime clocks, changing denominator definitions, and retrospective editing to meet targets. Master data—asset hierarchy, functional locations, criticality, job plans, and failure taxonomy—must be governed centrally.

5. Trend, Pareto, and Cost-Risk Analysis

5.1 Trend analysis

A trend chart plots a metric over time against a target, baseline, or statistical control limits. A simple period-to-period percentage change is:

This is descriptive, not proof of causation. Seasonality, production volume, asset mix, and reporting changes must be considered. A moving average can reduce short-term noise:

where xₜ is the current observation and k is the number of periods. Moving averages improve visibility but introduce delay and should not replace investigation.

5.2 Pareto analysis

Pareto analysis ranks categories by contribution to a selected loss, such as downtime hours, maintenance cost, failures, or safety events. The cumulative percentage for category j is:

where categories are ordered from largest to smallest contribution and cᵢ is the value of category i. The familiar “80/20” pattern is a heuristic, not a law. Pareto analysis identifies concentration; it does not explain root cause. The highest-cost category may be difficult to control, while a smaller category may offer a safer and cheaper improvement.

5.3 Cost-risk analysis

A qualitative risk priority can use likelihood and consequence ratings, but multiplying ordinal scores does not produce a true probability or monetary expectation. Where credible data exist, expected annual loss is:

where pᵢ is the annual probability of scenario i and Cᵢ is its consequence cost. Consequences may include injury, environmental damage, production loss, repair, legal exposure, and reputation. Uncertainty should be shown through ranges or scenarios rather than hidden inside a single precise number.

A decision can be assessed using net benefit:

For competing projects, benefit-cost ratio is:

Discounting is appropriate when cash flows occur over multiple periods. Non-financial constraints—legal compliance, intolerable safety risk, and ethical duties—may make a project mandatory even if its BCR is below one.

6. Worked Decision Example

An illustrative packaging plant experiences recurring failures in a conveyor drive. Annual failure probability is estimated at 0.30. Each failure costs $18,000 in lost production and repair, giving expected annual loss of:

0.30 × $18,000 = $5,400.

An improved condition-monitoring programme costs $2,000 per year and is expected to reduce failure probability to 0.10. The residual expected loss is:

0.10 × $18,000 = $1,800.

Avoided expected loss is $5,400 − $1,800 = $3,600. Net annual benefit is $3,600 − $2,000 = $1,600. The simple benefit-cost ratio is $3,600 ÷ $2,000 = 1.8.

The decision should not be based only on this calculation. The manager should test the probability estimate, consider whether monitoring detects the failure early enough, verify technician competence, evaluate false alarms, and include safety or environmental consequences. If failure could cause serious injury, legal or ethical requirements may justify the control independently of financial return.

7. Integrated Case: Northbridge Components (Illustrative)

Background

Northbridge Components operates a three-shift machining line. During six months, OEE fell from 84% to 77%, emergency work increased, and maintenance expenditure was 9% above budget. The plant manager initially proposed reducing preventive-maintenance frequency to recover the budget.

Problem and analysis

The team established common definitions and examined work orders, downtime logs, production counts, and safety records. The dashboard showed:

• availability declined from 91% to 84%;
• performance remained near 96%;
• quality declined slightly from 96% to 95%;
• schedule compliance fell from 88% to 69%;
• emergency work increased from 12% to 28% of maintenance hours;
• wrench time fell from 34% to 24%; and
• two safety-critical inspections were overdue.

Pareto analysis of downtime identified three dominant causes: hydraulic leaks, sensor faults, and delayed material access. Interviews showed that planners were releasing incomplete job packages. Stores had changed its issue process, and technicians frequently waited for seals, permits, and drawings. The apparent cost problem was therefore connected to poor planning and material availability rather than excessive preventive work alone.

Approach and solution

Management took four actions. First, it protected safety-critical inspections and corrected the overdue tasks. Second, it introduced a weekly frozen schedule with a defined planning horizon and a materials-ready gate. Third, it created failure codes and reviewed the top downtime causes through root-cause analysis. Fourth, it redesigned the dashboard to display emergency work, schedule compliance, wrench time, backlog age, OEE loss categories, and cost by asset criticality.

The plant did not use wrench time as an individual productivity target. It used the measure diagnostically at team level and investigated waiting causes. The preventive-maintenance programme was altered only where failure evidence supported a change.

Outcome

After three months, schedule compliance increased to 86%, emergency work fell to 17%, and wrench time rose to 31%. Availability improved to 89%, producing an OEE of approximately 84% when performance and quality remained near their prior levels. Maintenance cost returned close to budget, while overdue safety-critical work fell to zero. These results are illustrative and should not be generalized without validating the data and sustaining the controls.

Learning points

1. Cost overruns may be symptoms of weak work management.
2. A balanced dashboard prevents one-dimensional cost cutting.
3. Leading indicators can reveal deterioration before injury or major failure.
4. Pareto analysis prioritizes investigation but does not replace causal analysis.
5. Metrics should support system improvement, not blame individual workers.

8. Limitations, Ethics, and Contemporary Practice

Performance measures can produce dysfunctional behaviour. Employees may avoid reporting near misses, planners may remove difficult work from schedules, supervisors may reclassify downtime, and managers may defer essential maintenance to meet a quarterly budget. Such gaming is reduced by audit trails, independent assurance, balanced indicators, and leadership that rewards truthful reporting.

Privacy must be protected when dashboards use access-control, wearable, location, or productivity data. Monitoring should be necessary, proportionate, transparent, and compliant with employment and data-protection law. Safety data must not be used to punish legitimate reporting or discourage stop-work authority.

Artificial intelligence and advanced analytics can classify failure text, detect abnormal vibration, forecast remaining useful life, optimize spares, and identify leading patterns across work orders. However, models require reliable labelled data, stable operating conditions, explainability, cybersecurity, and human oversight. A prediction is not a maintenance decision: it must be translated into an executable work order with suitable risk controls. False positives create unnecessary work, while false negatives can create dangerous confidence.

Contemporary best practice includes condition-based maintenance, reliability-centred maintenance, digital twins, mobile work management, integrated safety-risk registers, and carbon-aware maintenance planning. Technology should follow a clear business and risk problem; digitizing poor processes merely produces faster poor decisions.

Best-Practice Checklist

• Define every KPI, denominator, exclusion, owner, and response.
• Combine leading and lagging measures.
• Segment results by asset criticality, shift, product, and failure mode where relevant.
• Reconcile production, safety, finance, and maintenance data.
• Use Pareto analysis to focus investigation, not to claim causation.
• Protect statutory and safety-critical work from arbitrary budget cuts.
• Review trends with context, seasonality, exposure, and data changes.
• Use cost-risk analysis with uncertainty and non-financial constraints.
• Audit dashboards for gaming, privacy, bias, and data integrity.
• Convert every significant deviation into an owned, time-bound action.

Chapter Summary

Performance measurement links safety controls, maintenance execution, asset reliability, production results, and financial outcomes. Leading indicators provide early evidence about the strength of controls, while lagging indicators reveal realized outcomes. Essential maintenance measures include schedule compliance, backlog, wrench time, MTBF, MTTR, availability, downtime, OEE, and maintenance cost ratios. Each requires precise definitions and appropriate assumptions.

Dashboards are effective when they are decision-oriented, balanced, traceable, and governed. Trend analysis reveals direction, Pareto analysis concentrates attention on major contributors, and cost-risk analysis supports rational prioritization. Budgeting should protect compliance and safety-critical work while considering life-cycle cost rather than short-term expenditure alone. Ethical management requires truthful reporting, privacy protection, avoidance of metric gaming, and human oversight of analytical technologies.

Key Terms

Leading indicator: A measure of conditions or actions that may influence future outcomes.

Lagging indicator: A measure of an outcome that has already occurred.

KPI: A strategically important performance indicator used to guide management action.

TRIR: Total recordable incident rate based on a defined exposure-hour base.

Schedule compliance: The proportion of scheduled work completed within the defined period.

Backlog: Accumulated authorized maintenance work not yet completed.

Wrench time: Proportion of available craft time spent in direct hands-on maintenance.

MTBF: Mean operating time between functional failures.

MTTR: Mean time required to restore equipment through repair.

Availability: Proportion of relevant planned or calendar time that equipment is operational.

Downtime: Time during which an asset cannot perform its required function.

OEE: Product of availability, performance, and quality.

Planned production time: Time available for production after defined planned stops are excluded.

Pareto analysis: Ranking categories by contribution and displaying cumulative contribution.

Expected annual loss: Probability-weighted consequence across risk scenarios.

Life-cycle cost: Total cost of acquiring, operating, maintaining, and disposing of an asset.

Data governance: Policies and controls that ensure data quality, ownership, security, and traceability.

Metric gaming: Behaviour that improves a reported measure without improving actual performance.

Condition-based maintenance: Maintenance initiated by evidence of actual equipment condition.

Decision analytics: Use of data, models, and judgement to compare alternatives and guide action.

Review Questions

Short Answer Questions

1. Distinguish between a leading indicator and a lagging indicator, giving one example of each.
2. State the three components of OEE.
3. Why must the denominator of a maintenance cost ratio be reported?
4. Define wrench time and identify two activities normally excluded from its numerator.
5. What is the purpose of Pareto analysis?

Long Answer Questions

1. Explain how a balanced safety and maintenance dashboard should be designed and governed.
2. Discuss the assumptions and limitations of MTBF, MTTR, and the simple availability relationship.
3. Explain how schedule compliance, emergency work, backlog, and wrench time should be interpreted together.
4. Compare historical, zero-based, activity-based, and risk-based maintenance budgeting.
5. Discuss ethical risks in performance measurement and propose controls to prevent metric gaming.

Analytical / Application Questions

1. A line has 600 minutes of planned production time, operates for 510 minutes, produces 900 units, has an ideal cycle time of 0.5 minutes per unit, and produces 855 good units. Calculate availability, performance, quality, and OEE.
2. A failure has an annual probability of 0.25 and a consequence of $40,000. A monitoring programme costs $3,000 annually and reduces probability to 0.08. Calculate the expected annual loss before and after, the net annual benefit, and the simple benefit-cost ratio. State two non-financial factors that should also be reviewed.
3. A maintenance department reports improving cost variance but worsening emergency work, overdue safety inspections, and schedule compliance. Diagnose at least three possible explanations and recommend a dashboard-based management response.

Multiple Choice Questions with Answers

Question 1

Which measure is primarily a leading indicator?

A. Lost-time injury rate

B. Number of completed safety-critical inspections

C. Total injury days

D. Number of equipment failures

Correct Answer: B. Number of completed safety-critical inspections

Question 2

A line has 480 minutes of planned production time and operates for 420 minutes. Its availability is:

A. 79.2%

B. 84.0%

C. 87.5%

D. 95.2%

Correct Answer: C. 87.5%

Question 3

Which expression correctly represents OEE?

A. Availability + Performance + Quality

B. Availability × Performance × Quality

C. Availability ÷ Performance × Quality

D. Performance − Availability − Quality

Correct Answer: B. Availability × Performance × Quality

Question 4

In the wrench-time formula, the numerator represents:

A. All paid hours including waiting

B. Direct hands-on maintenance hours

C. Total scheduled production hours

D. Emergency work orders only

Correct Answer: B. Direct hands-on maintenance hours

Question 5

Pareto analysis is mainly used to:

A. Prove the root cause of every failure

B. Rank contributors so investigation can be focused

C. Replace risk assessment

D. Calculate MTTR

Correct Answer: B. Rank contributors so investigation can be focused

Question 6

If MTBF is 90 hours and MTTR is 10 hours, simple inherent availability is:

A. 10%

B. 50%

C. 90%

D. 100%

Correct Answer: C. 90%

Question 7

A maintenance cost ratio is meaningful only when:

A. Its denominator is omitted

B. The denominator and scope are clearly defined

C. It is always based on labour cost

D. It excludes all contractor costs

Correct Answer: B. The denominator and scope are clearly defined

Question 8

Expected annual loss is calculated by:

A. Adding probability and consequence

B. Multiplying consequence by the relevant probability and summing scenarios

C. Dividing consequence by maintenance hours

D. Subtracting budget from actual cost

Correct Answer: B. Multiplying consequence by the relevant probability and summing scenarios

Question 9

Which practice is most likely to create misleading schedule compliance?

A. Defining exclusions consistently

B. Auditing work-order status

C. Removing difficult jobs from the schedule after release

D. Separating emergency work from planned work

Correct Answer: C. Removing difficult jobs from the schedule after release

Question 10

A prediction from an AI maintenance model should generally be:

A. Accepted without human review

B. Converted into work only after technical and risk validation

C. Used to eliminate safety procedures

D. Treated as proof of failure

Correct Answer: B. Converted into work only after technical and risk validation