Process Analysis and Performance Measurement

Introduction

Operations and supply chain performance depends on how work moves through a system. A customer order, for example, may pass through demand entry, credit approval, picking, packing, transportation, delivery, and invoicing. Each activity consumes time and resources, and delays or variability at one stage can affect the entire system. Process analysis provides a structured way to understand this flow, identify waste and constraints, and improve performance using evidence rather than intuition.

This chapter develops the foundations of process analysis for business analytics. It explains how to represent processes with flowcharts and value-stream maps; how to calculate throughput, cycle time, capacity, utilization, bottlenecks, productivity, and efficiency; how to apply Little’s Law; and how to design operational KPI dashboards. The emphasis is on interpretation: a metric is useful only when its definition, time window, data source, and managerial purpose are clear.

Learning Objectives

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

1. Map a process using flowcharts and value-stream maps.
2. Distinguish processing time, waiting time, lead time, cycle time, and throughput.
3. Calculate capacity, utilization, bottlenecks, productivity, and efficiency.
4. Apply Little’s Law under appropriate steady-state conditions.
5. Interpret process performance and identify improvement priorities.
6. Design a balanced operational KPI dashboard.
7. Evaluate data-quality, ethical, and implementation issues in process analytics.

1. Process Thinking and Process Boundaries

A process is a linked set of activities that transforms inputs into outputs for a customer or downstream user. Inputs may include materials, information, labor, equipment, and capital. Outputs may be products, services, decisions, or information.

A process has a boundary. For example, an order-fulfillment study might begin when an order is released to the warehouse and end when the parcel is handed to the carrier. If the boundary is not specified, analysts may mix incompatible measures—for example, comparing warehouse processing time with end-to-end customer lead time.

Important process characteristics include:

• Flow unit: the item being followed, such as an order, patient, invoice, or pallet.
• Activity: work performed on the flow unit.
• Resource: a person, machine, system, or space used by an activity.
• Queue: flow units waiting for service.
• Handoff: transfer of responsibility between people, departments, or systems.
• Decision point: a point at which different routes are selected.
• Rework: additional work caused by an error or failure to meet requirements.
• Scope: the defined start and end points of the analysis.

A process may be serial, with activities performed in sequence; parallel, with multiple resources working simultaneously; or hybrid, with both structures. A process can also contain loops, such as inspection followed by rework and reinspection.

2. Process Mapping

Flowcharts

A flowchart represents activities and decisions in sequence. Standard conventions include rectangles for activities, diamonds for decisions, arrows for flow, and rounded shapes for start or end points. A good flowchart identifies the responsible role or system and makes exceptions visible.

A useful mapping procedure is:

1. Define the flow unit and process boundaries.
2. Observe the actual process, not merely the documented procedure.
3. List activities in the order they occur.
4. Add decisions, loops, queues, handoffs, and information systems.
5. Record time, defects, volume, and responsible resources where available.
6. Validate the map with process participants.
7. Distinguish the current state from a proposed future state.

The analyst should avoid mapping only the ideal path. Exception paths—missing information, stockouts, failed payments, damaged goods, and rework—often explain most delays and cost.

Value-Stream Mapping

A value-stream map (VSM) shows the flow of materials and information across a broader process, often from supplier to customer. It distinguishes value-adding work from waiting, transportation, inspection, storage, and other forms of non-value-adding time. Typical VSM fields include customer demand, available work time, inventory or work-in-process, processing time, waiting time, information flows, and production-control signals.

A customer generally regards an activity as value-adding when it changes the product or service in a way the customer requires and is willing to pay for, and when it is performed correctly the first time. Regulatory inspection may be necessary even though it does not directly transform the product; it should therefore be classified carefully as necessary but non-value-adding rather than simply wasteful.

Figure: A simplified process map showing activities, a decision, and elapsed-time components.

A VSM is especially useful when local efficiency hides system-level delay. A department may be highly utilized while the customer waits because work accumulates before or after that department.

3. Core Process Performance Measures

Throughput

Throughput is the rate at which completed flow units leave a process.

Here, TH is throughput, Q is the number of completed units, and T is the observation period. Units may be orders per hour, patients per day, or tonnes per week. Throughput should be based on completed, acceptable output unless the organization explicitly reports total output including defects.

Cycle Time and Lead Time

Processing time is the time during which resources actively work on a flow unit. Waiting time is time spent in a queue or idle between activities. Transport time, information delay, inspection, and other elapsed components may also occur.

Cycle time is used in two common ways: the elapsed time between successive completed units, and the time a particular unit spends moving through a process. Analysts must state the definition used. In this chapter, unit-level elapsed time through a process is called cycle time, while the interval between completions is called takt or inter-departure time when appropriate.

Lead time is the total elapsed time from the customer’s initiating event to completion or delivery. Thus, for a defined process boundary:

Value-stream analysis often reveals that processing time is a small fraction of lead time. Reducing processing time alone may therefore have little effect unless queues and handoffs are also addressed.

Capacity

Capacity is the maximum sustainable output rate under specified conditions of resources, demand, product mix, quality standards, and operating hours. For one resource with effective available time A per period and service time s per unit:

If a resource has m identical parallel servers, the idealized capacity is:

The formula assumes comparable servers, stable work content, no blocking or starving, and that A is effective productive time rather than scheduled time. Breaks, maintenance, setup, absenteeism, and downtime should be deducted or modeled separately.

Utilization

Utilization measures the proportion of available capacity used by actual workload:

It is often expressed as a percentage. A resource operating at 90 units per hour with capacity of 100 units per hour has 90% utilization. Utilization above 100% indicates that the assumed capacity, time window, or data definition is inconsistent; it is not evidence of superior performance.

High utilization can increase queues when arrivals are variable. A resource operating continuously near its theoretical maximum has little protection against demand surges, breakdowns, or rework.

Bottlenecks

A bottleneck is a resource or policy constraint that limits the throughput of the entire process. In a serial process, the bottleneck is generally the activity with the lowest effective capacity, but this conclusion must account for product mix, downtime, setups, quality losses, and batch policies.

For serial stations with capacities C₁, C₂, ..., C_n, under compatible flow and no inventory accumulation at the boundary:

The approximation matters: blocking, starving, parallel routing, rework, and finite buffers can make the actual system capacity different. Improving a non-bottleneck may create more work-in-process without increasing throughput. Bottleneck improvement should begin with eliminating avoidable downtime, reducing setup or changeover time, ensuring quality at the source, and aligning upstream release with bottleneck capacity.

Productivity and Efficiency

Productivity relates output to input:

Examples include orders per labor-hour, revenue per employee, or units per machine-hour. Productivity may be labor, machine, material, energy, or multifactor productivity. The input and output units must be stated.

Efficiency compares actual performance with a standard or theoretically attainable benchmark:

If producing 100 units should require 20 labor-hours but actually requires 25 labor-hours, efficiency is 20 ÷ 25 = 80%. Efficiency is not identical to utilization: a worker can be fully utilized while using more labor than the standard because of rework or poor methods.

Quality and First-Pass Yield

Process measures should include quality. First-pass yield (FPY) is the proportion of units that complete a process correctly without rework:

For sequential independent steps with first-pass yields y₁, y₂, ..., y_n, the rolled throughput yield is:

The independence assumption is a simplification; correlated defects may make the product estimate inaccurate. Quality losses reduce effective capacity and increase lead time.

4. Little’s Law

Little’s Law connects average work-in-process, throughput, and average time in the system:

L is average number of flow units in the system, λ is average throughput rate, and W is average time a flow unit spends in the system. Units must be consistent: orders = orders/day × days.

The law applies to a stable system over a suitable observation period: average inflow and outflow are approximately equal, the system is not accumulating unlimited inventory, and the averages refer to the same boundary and time window. It does not require a particular probability distribution or service discipline.

Rearrangements are useful:

Example. A repair center averages 240 open jobs and completes 30 jobs per day. Assuming stable long-run conditions, average time in the center is 240 ÷ 30 = 8 days. If management wants average time reduced to 5 days while throughput remains 30 jobs per day, target work-in-process is 30 × 5 = 150 jobs. The reduction requires removing or preventing approximately 90 jobs of excess work-in-process; it cannot be achieved merely by changing the dashboard.

Little’s Law is descriptive, not automatically causal. Reducing inventory can reduce time, but only if throughput is maintained and the process is not starved. Likewise, increasing throughput without controlling arrivals may increase work-in-process and waiting.

5. Worked Process Analysis

Consider an illustrative online retailer. A warehouse operates 8 hours per day. Orders pass through picking, packing, and dispatch. Effective available time per worker is 7.5 hours because of breaks and routine nonproductive time.

• Picking: 6 workers, 5 minutes per order.
• Packing: 2 workers, 8 minutes per order.
• Dispatch: 1 worker, 3 minutes per order.
• Historical completed output: 90 orders per hour.

Capacity calculations use minutes per hour. Picking capacity is 6 × 60 ÷ 5 = 72 orders/hour. Packing capacity is 2 × 60 ÷ 8 = 15 orders/hour if each worker performs one order at a time—clearly inconsistent with the reported output, suggesting that the stated packing time, staffing, or process boundary needs verification. Dispatch capacity is 60 ÷ 3 = 20 orders/hour. The apparent bottleneck is packing, but the 90-orders-per-hour report cannot be true for this configuration.

This is an important analytical result: calculations are also data-quality tests. Possible explanations include packing time being 8 seconds rather than 8 minutes, additional packing workers omitted from the data, orders being packed in batches, or output being measured at a different process boundary. An analyst should not force the numbers into a conclusion. The next step is observation and reconciliation of timestamps, staffing rosters, and definitions.

Suppose validation establishes that packing time is 8 minutes per batch of 10 orders, with two workers. Packing capacity is then 2 × 60 × 10 ÷ 8 = 150 orders/hour, while dispatch remains 20 orders/hour. Picking capacity is 72 orders/hour, so picking is the bottleneck and process capacity is approximately 72 orders/hour, subject to quality and downtime. At an output of 60 orders/hour, picking utilization is 60 ÷ 72 = 83.3%, and dispatch utilization is 60 ÷ 20 = 300%, again indicating that dispatch data or the assumed work content is inconsistent. This illustrates why every metric requires a clear unit of analysis and validation.

6. Designing Operational KPI Dashboards

A dashboard should support decisions, not merely display large quantities of data. KPI design begins with the operational question: “What decision will change if this measure moves?” A balanced process dashboard commonly includes the following dimensions:

A dashboard should distinguish leading indicators from lagging indicators. Queue size, schedule adherence, and downtime can signal future service problems; completed throughput and customer complaints describe outcomes after they occur.

Best practices include:

• Define the numerator, denominator, unit, owner, and time window for every KPI.
• Display actual performance against a meaningful target or control limit.
• Segment by product, shift, location, customer class, and exception type when appropriate.
• Show trends, not isolated values.
• Use medians and percentiles when averages hide skewed delays.
• Prevent double counting across systems.
• Record data latency and timestamp definitions.
• Pair speed measures with quality, cost, safety, and employee-impact measures.
• Establish escalation rules and a review cadence.

A target can distort behavior. For example, rewarding throughput alone may encourage premature completion, poor quality, or unsafe work. A balanced scorecard reduces this risk, but it does not eliminate the need for managerial judgment.

7. Applied Case: Reducing Order Lead Time

Background

The following is an illustrative case. A regional distributor promises delivery within two business days. Its order-to-ship process includes order entry, credit review, picking, packing, and carrier handoff. Customers report late deliveries even though each department claims to meet its local target.

Problem and Analysis

Process mapping shows that actual processing time totals 42 minutes per order, while average elapsed order-to-ship lead time is 19 hours. Orders spend most of the time waiting for a twice-daily credit-review batch and in a queue before picking. Average work-in-process is 1,140 orders, and average throughput is 60 orders per hour during operating periods.

Using Little’s Law, the implied average time in the defined system is:

The calculation agrees with observed lead time and provides a useful validation. A VSM identifies the batch credit-review policy as a major source of delay. Picking is the capacity constraint during the afternoon peak, with effective capacity of 66 orders per hour and demand release of 70 orders per hour. The queue therefore grows during the peak unless earlier work is reduced or capacity is added.

Approach and Solution

Management pilots three changes: continuous credit review for low-risk customers, release leveling to avoid a large afternoon wave, and cross-training one packing employee to support picking during peak periods. KPIs are defined as throughput, 90th-percentile order-to-ship time, picking utilization, first-pass yield, and on-time shipment rate. The dashboard reports both hourly and daily views so that short-term congestion is not hidden by daily averages.

Outcome

After the pilot, average work-in-process falls to 840 orders while throughput remains 60 orders per hour. Little’s Law predicts a new average time of 840 ÷ 60 = 14 hours, a five-hour reduction. The organization should verify this result with timestamped order-level data and check that quality and employee safety have not deteriorated. If first-pass yield falls, apparent speed may represent rework shifted downstream rather than genuine improvement.

Learning Points

The case demonstrates that local targets can coexist with poor end-to-end performance; queue and batch policies may dominate processing time; Little’s Law can validate observed averages; and a dashboard must combine flow, time, capacity, quality, and service measures.

8. Limitations, Governance, and Contemporary Analytics

Process metrics are sensitive to definitions. “Completed order” may mean picked, packed, shipped, or delivered. Different timestamps can produce contradictory lead times. Missing events, manual overrides, clock differences, and status changes can bias analysis. Analysts should maintain a metric dictionary, version process definitions, audit data lineage, and reconcile operational systems with samples of observed work.

Employee-level metrics require particular care. Monitoring may support coaching and safety, but excessive surveillance can undermine trust and encourage gaming. Data should be limited to legitimate purposes, access-controlled, retained appropriately, and evaluated for discriminatory effects. Automated scheduling or performance scoring should be explainable and subject to human review.

Process mining tools can reconstruct flows from event logs and identify variants, bottlenecks, and conformance failures. Machine-learning models can forecast queue growth or predict late orders. These tools are useful only when event logs are complete and process definitions are stable. Prediction does not establish causation: a model may identify correlation without showing which intervention will improve performance. Human process knowledge, controlled pilots, and post-implementation measurement remain essential.

Chapter Summary

Process analysis makes work visible by defining the flow unit, boundaries, activities, resources, queues, decisions, and exceptions. Flowcharts describe sequence and responsibility, while value-stream maps connect material and information flow and expose waiting and waste. Throughput measures completed output per time; capacity measures sustainable potential; utilization compares actual rate with capacity; bottlenecks constrain system output; productivity relates output to input; and efficiency compares actual input with a standard. Processing time is only one component of lead time, and high utilization may create queues under variability. Little’s Law, L = λW, links average work-in-process, throughput, and time when stable-system assumptions hold. Effective KPI dashboards use precise definitions, balanced measures, trends, segmentation, and appropriate governance. Sound process analytics combines calculation with observation, data validation, ethical controls, and tested operational change.

Key Terms

Process: Linked activities that transform inputs into outputs.

Flow unit: Item tracked through a process.

Queue: Flow units waiting for service.

Handoff: Transfer of responsibility between roles or systems.

Flowchart: Diagram showing activities, decisions, and sequence.

Value-stream map: Map of material and information flow across a process.

Processing time: Time during which active work is performed.

Waiting time: Time a flow unit spends idle or queued.

Lead time: Total elapsed time across a defined customer or process boundary.

Throughput: Completed output per unit of time.

Capacity: Maximum sustainable output rate under stated conditions.

Utilization: Actual output rate divided by available capacity rate.

Bottleneck: Constraint that limits system throughput.

Productivity: Output divided by input.

Efficiency: Standard input for actual output divided by actual input.

First-pass yield: Proportion completed correctly without rework.

Work-in-process: Flow units currently inside the process.

Little’s Law: Relationship L = λW between work-in-process, throughput, and time.

KPI: Key performance indicator used to monitor an important objective.

Process mining: Analysis of event logs to reconstruct and evaluate process flows.

Review Questions

Short Answer Questions

1. What is the difference between a process boundary and a flow unit?
2. Why can high utilization increase waiting time?
3. Define throughput and state its usual units.
4. What does first-pass yield measure?
5. State the assumptions required for applying Little’s Law reliably.

Long Answer Questions

1. Explain how a flowchart and a value-stream map differ in purpose and scope.
2. Discuss why improving a non-bottleneck may fail to increase total process throughput.
3. Explain the difference among processing time, waiting time, cycle time, and lead time.
4. Describe how a balanced operational KPI dashboard should be designed and governed.
5. Evaluate the risks of relying on employee-level productivity metrics for management decisions.

Analytical / Application Questions

1. A service center has average work-in-process of 360 cases and throughput of 45 cases per day. Calculate average time in the system and state the assumption that makes the result valid.
2. Three serial stations have capacities of 40, 55, and 48 units per hour. Identify the likely bottleneck and explain two reasons why actual process capacity might differ from the simple minimum-capacity estimate.
3. A warehouse reports 500 orders per day, capacity of 625 orders per day, and 4% rework. Calculate nominal utilization and explain why the dashboard should also report quality-adjusted performance.

Multiple Choice Questions with Answers

Question 1

Which measure is defined as completed output divided by an observation period?

A. Throughput

B. Utilization

C. Efficiency

D. First-pass yield

Correct Answer: A. Throughput

Question 2

A process has average work-in-process of 240 units and throughput of 30 units per day. Under stable conditions, what is the average time in the system?

A. 4 days

B. 8 days

C. 30 days

D. 270 days

Correct Answer: B. 8 days

Question 3

In a serial process with no relevant complications, which station is most likely the bottleneck?

A. The station with the highest utilization target

B. The station with the lowest effective capacity

C. The station with the shortest processing time

D. The station with the largest physical area

Correct Answer: B. The station with the lowest effective capacity

Question 4

Which expression correctly represents utilization?

A. Input divided by output

B. Waiting time divided by processing time

C. Actual output rate divided by available capacity rate

D. Standard input divided by actual output

Correct Answer: C. Actual output rate divided by available capacity rate

Question 5

A process requires 20 labor-hours by standard but uses 25 labor-hours for the actual output. What is its efficiency?

A. 20%

B. 25%

C. 80%

D. 125%

Correct Answer: C. 80%

Question 6

Which feature is characteristic of a value-stream map?

A. It shows only employee reporting relationships

B. It connects material and information flows across a broader process

C. It excludes waiting and inventory

D. It measures only machine speed

Correct Answer: B. It connects material and information flows across a broader process

Question 7

What does first-pass yield exclude from its numerator?

A. Units completed correctly without rework

B. Units entering the process

C. Units requiring rework before acceptance

D. Units measured during the observation period

Correct Answer: C. Units requiring rework before acceptance

Question 8

A dashboard reports throughput but not quality, safety, or service reliability. What is the principal design weakness?

A. It uses too many leading indicators

B. It is balanced across all operational dimensions

C. It may encourage speed while hiding harmful trade-offs

D. It cannot display a time trend

Correct Answer: C. It may encourage speed while hiding harmful trade-offs

Question 9

Which condition is important when applying Little’s Law to operational averages?

A. The process must have exactly one activity

B. The system should be approximately stable over the measurement period

C. All processing times must be identical

D. Demand must be zero during breaks

Correct Answer: B. The system should be approximately stable over the measurement period

Question 10

A resource has capacity of 100 units per hour and produces 90 acceptable units per hour. What is its utilization?

A. 9%

B. 10%

C. 90%

D. 190%

Correct Answer: C. 90%