PPC Performance Analytics and Decision Making

Introduction

Production planning and control (PPC) converts demand, materials, labour, equipment and information into coordinated production activity. Its effectiveness cannot be judged only by whether a schedule was issued. A professional PPC system must show whether the organisation is producing the right quantity, at the required time, with acceptable quality, resource utilisation, cost and inventory exposure.

Performance analytics provides this evidence. It links operational data to managerial decisions through key performance indicators (KPIs), visual dashboards, statistical analysis, root-cause investigation and scenario evaluation. The central challenge is to avoid managing isolated measures. High equipment utilisation, for example, may be achieved by producing excess inventory; high throughput may conceal poor quality or excessive overtime. Effective decision making therefore requires a balanced view of flow, reliability, responsiveness, cost and customer service.

This chapter examines the main PPC measures and shows how they can be combined into an integrated performance-management system.

Learning objectives

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

1. Define and calculate major PPC KPIs.
2. Interpret relationships and trade-offs among throughput, OEE, inventory, service and cost measures.
3. Construct a useful PPC dashboard with appropriate data governance.
4. Apply Pareto analysis, root-cause methods and structured problem solving.
5. Evaluate alternatives using scenario analysis and basic capacity relationships.
6. Recognise ethical, organisational and technological risks in performance analytics.

1. Performance measurement as a PPC system

A KPI is a selected measure that indicates progress toward an operational or strategic objective. A metric becomes a useful KPI only when it has a defined purpose, owner, calculation rule, data source, reporting frequency and target. For example, “on-time delivery” is incomplete unless the organisation defines the promised date, shipment completion, treatment of partial deliveries and treatment of customer-requested changes.

PPC performance should be viewed as a chain:

plan → schedule → release → produce → inspect → deliver → learn

Measures should therefore cover four dimensions:

• Flow: throughput, lead time and work-in-process.
• Resource effectiveness: capacity utilisation and OEE.
• Reliability: schedule adherence and service level.
• Economic performance: inventory turns, cost variance and cash impact.

The balanced-scorecard principle is relevant: no single KPI adequately represents operational health. Measures should be balanced between short-term output and long-term capability, internal efficiency and customer value, and leading indicators and lagging indicators.

Leading and lagging indicators

A lagging indicator records an outcome, such as monthly service level or cost variance. A leading indicator signals conditions likely to affect future outcomes, such as schedule instability, overdue preventive maintenance, material shortages or rising queue length. Managers should use both. Waiting for service-level failure before correcting a shortage is an example of poor analytical control.

2. Core PPC performance indicators

2.1 Throughput

Throughput is the rate at which the system produces acceptable output. In a stable period:

Throughput may be expressed as units per hour, tonnes per day or orders per week. “Good units” excludes scrap and units requiring rework if the purpose is to measure saleable output. The time basis must be stated clearly; calendar time, scheduled time and actual operating time produce different results.

Throughput is not identical to production volume. Producing defective units or building unwanted stock increases gross output but may not increase economically useful throughput. In a constrained system, the bottleneck usually limits the sustainable throughput of the entire process. Local improvements outside the constraint may therefore have little effect on total output.

2.2 Overall equipment effectiveness (OEE)

OEE combines availability, performance and quality:

where

Availability excludes planned non-production time, such as scheduled shutdowns, but includes breakdowns and setup losses within planned production time. Performance compares actual speed with the ideal cycle time. Quality excludes scrap and, depending on organisational convention, may exclude rework only when the unit is ultimately acceptable; the convention must be documented.

The multiplication is valid when the three components are measured over the same period and use compatible definitions. Each component is dimensionless, so OEE is also dimensionless and commonly reported as a percentage.

Worked example. A machine has 450 minutes of planned production time. It operates for 390 minutes, produces 1,200 units, and has an ideal cycle time of 0.25 minutes per unit. Of the output, 1,140 units are good.

Availability = 390 ÷ 450 = 0.8667, or 86.67%.

Performance = (0.25 × 1,200) ÷ 390 = 300 ÷ 390 = 0.7692, or 76.92%.

Quality = 1,140 ÷ 1,200 = 0.95, or 95.00%.

Therefore:

OEE = 0.8667 × 0.7692 × 0.95 ≈ 0.6333, or 63.33%.

The result does not by itself identify the intervention. Performance is the weakest component, but the organisation should investigate whether slow running is caused by minor stops, material variation, operator practices or an unrealistic ideal cycle time.

2.3 Schedule adherence

Schedule adherence measures whether production occurred when planned. One common order-based definition is:

A stricter unit-based definition is:

These definitions are not interchangeable. An order-based measure can hide a large quantity shortfall on one order, while a unit-based measure can overemphasise large orders. A company must define the tolerance window, treatment of partial completion, and whether schedule changes authorised before the frozen period are excluded.

Adherence is different from schedule attainment. Attainment often asks whether the planned quantity was eventually produced, whereas adherence considers timing. Both timing and quantity matter in flow environments.

2.4 Inventory turns

Inventory turns indicate how many times average inventory is consumed or sold during a period. For manufacturing analysis, cost of goods sold (COGS) is generally preferable to sales revenue because both numerator and denominator are measured at cost:

If beginning and ending inventory are the only observations:

The ratio is dimensionless. Its reciprocal gives approximate inventory days when the period is one year:

The approximation assumes relatively stable consumption and representative average inventory. Seasonal businesses should use monthly or weekly averages rather than only beginning and ending balances. High turns may indicate efficient flow, but may also indicate under-stocking, shortages or lost sales.

2.5 Service level

Service level measures the extent to which customer demand is satisfied according to an agreed criterion. A cycle or order service-level definition is:

A fill rate instead measures the proportion of demand supplied immediately:

Service level and fill rate answer different questions. An order may be counted as failed if one item is missing, while a fill-rate calculation may show that 98% of total units were supplied. Safety-stock decisions require clarity about which measure is being optimised.

2.6 Lead time

Lead time is elapsed time between two defined events. Manufacturing lead time may be measured from order release to completion; customer lead time may be measured from order receipt to delivery. A generic decomposition is:

The terms may overlap in particular systems, so the organisation must specify whether the decomposition is additive. In many factories, queue time is the largest component. Reducing processing time alone may therefore produce little customer benefit if work continues to wait between operations.

2.7 Cost variance

Cost variance compares actual cost with a standard, budget or baseline:

A positive value may mean overspending, while a negative value may mean underspending; the sign convention must be stated. Percentage variance is:

Cost variance can be decomposed into material, labour, overhead, overtime, freight, scrap and expedite components. An apparently favourable labour variance may result from understaffing that increases late deliveries and future overtime. Variance analysis should therefore be integrated with service, quality and capacity measures.

3. Relationships and trade-offs among KPIs

KPIs are interdependent. Increasing batch size can reduce setup frequency and improve OEE performance, but it may increase finished-goods inventory and lead time. Reducing inventory can improve turns and working capital, but may reduce service level if replenishment variability is ignored. Maximising utilisation can create queues when demand approaches effective capacity.

Little’s Law provides a useful relationship for a stable flow system:

where WIP is average work-in-process, throughput is average completion rate, and flow time is average time in the system. Units must be consistent—for example, units, units per day and days. The relationship assumes a reasonably stable system over the measurement period and that averages are taken over the same boundary and period. It is descriptive, not a scheduling rule. If throughput is 240 units per day and average flow time is 2.5 days, average WIP is 600 units.

The relationship shows why releasing more work does not necessarily increase output. If the bottleneck is unchanged, WIP and waiting time may rise while throughput remains constant.

4. Pareto analysis and root-cause methods

Pareto analysis ranks categories by their contribution to a selected effect. It is based on the practical observation that a small number of categories often account for a large share of the total, not on a universal law that must produce an exact 80/20 split.

A Pareto analysis should:

1. Define the problem and measurement period.
2. Select a meaningful unit, such as lost hours, late orders or monetary loss.
3. Categorise causes consistently.
4. Rank categories from largest to smallest.
5. Calculate cumulative totals and percentages.
6. Investigate the vital few, while monitoring the remaining categories.

For example, if downtime losses are setup 42 hours, material shortage 28 hours, breakdown 18 hours and quality holds 12 hours, total loss is 100 hours. Setup contributes 42%, material shortage 28%, and the first two categories cumulatively account for 70%. A bar chart with a cumulative line makes this visible, but the numerical logic remains essential.

Pareto analysis identifies where to investigate; it does not prove why the loss occurred. Root-cause tools include:

• Five Whys: repeatedly asks why, while avoiding premature blame.
• Fishbone diagram: organises possible causes under categories such as methods, machines, materials, people, measurement and environment.
• Fault-tree analysis: works deductively from an undesirable event toward combinations of causes.
• Process mapping and time observation: distinguish value-adding work from waiting and rework.
• Statistical process control: identifies unusual variation using time-ordered data.

A sound investigation validates causes with evidence. “Operator error” is rarely an adequate root cause; the analysis should ask whether instructions, training, interface design, workload or maintenance conditions made the error likely.

5. Dashboards for operational decision making

A dashboard is a visual control system, not merely a collection of charts. It should support a defined decision cycle: detect, diagnose, decide, act and verify.

A practical PPC dashboard can contain:

Each KPI should display actual value, target, trend, variance, responsible owner and escalation threshold. Drill-down should move from plant to line, product, shift, order and cause. Colour coding should be accessible and should not be the sole means of communicating status.

Data governance

Reliable analytics requires a common data dictionary. Definitions should specify numerator, denominator, exclusions, time zone, calendar, revision policy and source system. Data should be reconciled among the enterprise resource planning system, manufacturing execution system, maintenance records, quality system and finance ledger. A dashboard with precise graphics but inconsistent data creates false confidence.

Managers should distinguish data latency from performance deterioration. A late transaction posting may create an apparent service failure. Timestamp synchronisation, master-data quality, audit trails and role-based access are therefore operational controls, not merely information-technology concerns.

6. Scenario analysis and decision models

Scenario analysis tests how performance may change under alternative assumptions. It is particularly useful when decisions involve uncertainty, such as overtime, subcontracting, extra shifts, inventory policies or preventive maintenance.

A simple capacity model is:

This is an analytical approximation. It assumes the factors are defined consistently and that their product represents usable output. It should not be applied when capacity losses interact in a way that causes double counting.

Worked scenario. A line has design capacity of 1,000 units per day. Current availability is 0.85, performance is 0.90 and quality is 0.98.

Effective capacity = 1,000 × 0.85 × 0.90 × 0.98 = 749.7, approximately 750 good units per day.

If a maintenance programme raises availability to 0.92 while other factors remain constant:

Effective capacity = 1,000 × 0.92 × 0.90 × 0.98 = 811.4, approximately 811 good units per day.

The improvement is approximately 61 good units per day, not 70, because the availability gain is multiplied by performance and quality. A complete decision would also compare maintenance cost, downtime during implementation and expected durability.

Scenario analysis should include sensitivity testing. If demand exceeds effective capacity, planners may compare overtime, subcontracting, inventory drawdown, product prioritisation and demand negotiation. Assumptions should be explicit, and results should be presented as ranges where forecasts are uncertain rather than as false precision.

7. Integrated case: improving delivery reliability

Illustrative case. Meridian Components manufactures small assemblies for industrial customers. Its monthly dashboard reports 96% equipment utilisation, 68% schedule adherence, 82% order service level and inventory turns of 5.2. Management initially proposes increasing utilisation by running larger batches.

Background and problem

Customers are experiencing late deliveries despite high utilisation. Production supervisors report frequent schedule changes, shortages of purchased components and long queues before final assembly.

Analysis

The PPC team establishes common definitions and examines eight weeks of data. Pareto analysis of late-order hours shows:

The first two categories account for 65% of delay. A value-stream review finds that purchasing uses a monthly forecast, while production schedules change daily. Component shortages are concentrated in a small number of volatile items. Changeovers are extended because tooling is not prepared before the machine stops.

The team also checks OEE. Availability is acceptable, but performance is reduced by short stops and changeover losses. The high utilisation figure had included time spent producing inventory that was not immediately required; it therefore did not demonstrate customer responsiveness.

Approach and solution

Meridian introduces a two-week frozen schedule for critical resources, exception-based material alerts, pre-kitting of tools, and a daily cross-functional review of shortage and quality constraints. It segments components by demand variability and value rather than applying one inventory rule to all items. The team uses a cause-and-effect review for quality holds and assigns corrective actions with owners and due dates.

The dashboard is redesigned to place service level, schedule adherence, shortage hours, quality-hold hours and constrained-resource OEE together. Inventory turns remain visible but are not treated as the sole objective.

Outcome

After three months, the illustrative results are: schedule adherence rises from 68% to 88%, order service level from 82% to 94%, and shortage-related delay falls by 55%. Inventory turns decline temporarily to 4.8 because critical components are deliberately protected during stabilisation. Management accepts the short-term inventory effect because customer reliability improves and aged finished-goods inventory falls.

Learning points

The case demonstrates that high utilisation can coexist with poor service. The most useful intervention was not a universal push for more output, but better synchronisation of materials, schedules, changeovers and quality decisions. It also shows why KPI targets must be interpreted as a system rather than pursued independently.

8. Technology, AI and governance

Modern PPC analytics can combine ERP transactions, machine sensors, barcode scans, maintenance records and customer-order data. Predictive maintenance models may estimate failure risk; anomaly detection may identify unusual cycle times; optimisation software may recommend sequencing or allocation decisions.

These tools do not eliminate managerial responsibility. AI outputs depend on historical data, and historical schedules may encode undesirable priorities or discrimination between customers. Important controls include:

• traceable data lineage and versioned models;
• validation against independent operational evidence;
• human approval for high-impact schedule or service decisions;
• monitoring for drift and systematic bias;
• cybersecurity and access control;
• protection of employee and customer information;
• clear accountability when an automated recommendation causes harm.

A model that predicts downtime but ignores maintenance capacity may be analytically accurate yet operationally useless. Technology should therefore be selected after defining the decision, not before.

9. Limitations and best practices

Performance analytics has several limitations. KPI definitions can be manipulated through denominator changes, delayed transactions or selective exclusions. Aggregated averages can conceal variation by product, shift or customer. Targets may encourage gaming, such as producing low-value inventory to improve throughput or postponing maintenance to protect short-term availability. Correlation does not establish root cause, and dashboards can create information overload.

Best practice is to:

1. Start with strategic and customer requirements.
2. Define each KPI mathematically and operationally.
3. Use a small balanced set of measures.
4. Separate signal from normal variation.
5. Analyse distributions and trends, not only averages.
6. Combine Pareto prioritisation with evidence-based root-cause work.
7. Test decisions through scenarios and sensitivity analysis.
8. Assign owners, actions and review dates.
9. Audit data quality and protect transparency.
10. Review whether measures are producing the intended behaviour.

Chapter Summary

PPC performance analytics links production plans to measurable operational outcomes. Throughput measures useful output; OEE decomposes equipment effectiveness into availability, performance and quality; schedule adherence measures timing reliability; inventory turns indicate stock efficiency; service level captures customer fulfilment; lead time reveals responsiveness; and cost variance connects operations with financial control. Each KPI requires explicit definitions, consistent units and appropriate assumptions.

Because KPIs interact, managers must avoid local optimisation. Little’s Law explains the relationship among WIP, throughput and flow time, while scenario analysis helps compare capacity, maintenance, overtime and inventory alternatives. Pareto analysis focuses attention on major loss categories, but root-cause methods and evidence are needed to explain them. Dashboards are most valuable when they support a decision cycle and are governed by reliable data. AI and advanced analytics can strengthen PPC, but only with validation, transparency, security and human accountability.

Key Terms

KPI: A selected measure used to monitor progress toward an operational objective.

Throughput: Rate of acceptable output completed by a process.

OEE: Product of availability, performance and quality.

Availability: Operating time divided by planned production time.

Performance: Actual production speed relative to ideal speed.

Quality: Good units divided by total units.

Schedule adherence: Degree to which production occurs at the planned time and quantity.

Inventory turns: COGS divided by average inventory at cost.

Service level: Proportion of eligible demand fulfilled according to a defined criterion.

Fill rate: Proportion of demanded quantity supplied immediately.

Lead time: Elapsed time between defined start and completion events.

Cost variance: Difference between actual and budgeted or standard cost.

Pareto analysis: Ranked analysis used to identify major contributing categories.

Root cause: Fundamental condition whose removal prevents recurrence of a problem.

Dashboard: Visual system for monitoring performance and supporting decisions.

Scenario analysis: Evaluation of outcomes under alternative assumptions.

Little’s Law: Relationship among average WIP, throughput and flow time.

Data governance: Rules controlling data definitions, quality, access and accountability.

Bottleneck: Resource that constrains the throughput of a system.

Leading indicator: Measure that signals conditions likely to affect future performance.

Review Questions

Short Answer Questions

1. Define throughput and explain why gross production is not always equivalent to useful throughput.
2. Identify the three components of OEE.
3. Distinguish service level from fill rate.
4. What is the purpose of a Pareto analysis?
5. State the relationship expressed by Little’s Law.

Long Answer Questions

1. Explain how a PPC dashboard should combine flow, equipment, delivery, inventory and cost measures.
2. Discuss the assumptions and limitations of OEE as a performance indicator.
3. Explain why improving one KPI can sometimes worsen another, using manufacturing examples.
4. Describe a rigorous root-cause investigation from problem definition to corrective action.
5. Evaluate the benefits and governance risks of using AI in PPC decision making.

Analytical / Application Questions

1. A machine has 480 minutes of planned production, operates for 420 minutes, produces 1,000 units, has an ideal cycle time of 0.30 minutes, and produces 950 good units. Calculate availability, performance, quality and OEE.
2. A factory’s annual COGS is $12 million. Beginning inventory is $2.0 million and ending inventory is $2.8 million. Calculate average inventory and inventory turns, and interpret the result.
3. A line has design capacity of 1,500 units per day, availability of 0.80, performance of 0.92 and quality of 0.97. Calculate effective capacity and recommend what additional information is needed before approving overtime.

Multiple Choice Questions with Answers

Question 1

Which expression correctly represents OEE?

A. Availability + Performance + Quality

B. Availability × Performance × Quality

C. Throughput ÷ Inventory

D. Quality ÷ Availability

Correct Answer: B. Availability × Performance × Quality

Question 2

A process produces 900 good units in 30 operating hours. What is its throughput?

A. 3 units per hour

B. 30 units per hour

C. 270 units per hour

D. 930 units per hour

Correct Answer: B. 30 units per hour

Question 3

If annual COGS is $10 million and average inventory at cost is $2 million, inventory turns are:

A. 0.2

B. 2

C. 5

D. 20

Correct Answer: C. 5

Question 4

Which measure specifically evaluates whether production occurred at the planned time?

A. Schedule adherence

B. Inventory turns

C. Quality rate

D. Cost variance

Correct Answer: A. Schedule adherence

Question 5

In Little’s Law, if throughput is 100 units per day and flow time is 3 days, average WIP is:

A. 33.3 units

B. 97 units

C. 103 units

D. 300 units

Correct Answer: D. 300 units

Question 6

What is the primary purpose of Pareto analysis?

A. To calculate machine cycle time

B. To identify the largest contributing categories of a problem

C. To replace root-cause investigation

D. To determine accounting depreciation

Correct Answer: B. To identify the largest contributing categories of a problem

Question 7

A line has availability 0.90, performance 0.80 and quality 0.95. Its OEE is approximately:

A. 68.4%

B. 72.0%

C. 85.5%

D. 95.0%

Correct Answer: A. 68.4%

Question 8

Which is a leading indicator of possible future delivery failure?

A. Last month’s completed service level

B. Final annual cost variance

C. Current material shortages affecting scheduled orders

D. Prior-year inventory turns

Correct Answer: C. Current material shortages affecting scheduled orders

Question 9

Why must KPI definitions specify numerator, denominator and exclusions?

A. To make the dashboard visually attractive

B. To ensure consistent and comparable measurement

C. To guarantee that targets are always achieved

D. To eliminate the need for data validation

Correct Answer: B. To ensure consistent and comparable measurement

Question 10

A favourable labour-cost variance accompanied by increased late deliveries most likely illustrates that:

A. One KPI can hide adverse performance elsewhere

B. Cost variance is always incorrect

C. Service level and cost are unrelated

D. Late deliveries automatically improve inventory turns

Correct Answer: A. One KPI can hide adverse performance elsewhere