HKSM

Statistical sampling is a quality technique that examines a representative subset of outputs to make inferences about the entire lot or process. It reduces inspection cost and time while maintaining an acceptable level of confidence in quality decisions.

Key Points

• Used to estimate quality or conformance by inspecting a subset instead of every item.
• Two common types: attribute sampling (pass/fail) and variable sampling (measured values).
• Sample size depends on confidence level, margin of error, population size, and expected variability.
• Selection methods include random, systematic, and stratified sampling to reduce bias.
• Acceptance criteria and decision rules (for example, acceptance number or defect threshold) must be defined upfront.
• Results inform actions such as accept, reject, rework, adjust process controls, or increase sampling.

Quality Objective

Ensure reliable, cost-effective quality decisions by evaluating a representative sample that reflects the population, while controlling risk through clear confidence and precision targets.

Method Steps

• Define purpose: conformance verification, process control, or defect estimation.
• Identify the population and lot boundaries, and confirm it is produced under similar conditions.
• Choose sampling type: attribute (pass/fail) or variable (measured values).
• Set statistical parameters: confidence level, margin of error (precision), and expected defect rate or variability.
• Calculate sample size based on parameters and constraints (cost, time, risk tolerance).
• Select the sampling method: random, systematic (every k-th item with random start), or stratified (by subgroups).
• Define acceptance criteria and decision rules before sampling (acceptance number, control limits, or tolerance ranges).
• Prepare a sampling plan and checklist; train inspectors and calibrate measurement tools.
• Collect and record sample data consistently; protect traceability and data integrity.
• Analyze results, decide accept/reject/rework, and document findings and lessons learned.

Inputs Needed

• Quality management plan, quality metrics, and acceptance criteria.
• Population definition, lot size, and production records.
• Risk tolerance, confidence level, and desired precision.
• Historical defect rates or process capability data.
• Sampling plan template, checklists, and measurement procedures.
• Calibrated measurement tools and data collection forms or systems.

Outputs Produced

• Completed sampling plan with sample size and selection method.
• Inspection results and defect summaries (attribute counts or measured values).
• Estimated defect rate or performance metrics with confidence information.
• Acceptance or rejection decision and recommended actions (rework, adjustment, resampling).
• Nonconformance reports, change requests, and updates to quality records and metrics.
• Lessons learned and process improvement recommendations.

Acceptance/Control Rules

• Acceptance criteria and decision thresholds are defined before sampling begins.
• Sample selection must be unbiased and follow the documented method.
• Measurement system must be capable, calibrated, and consistent.
• If defects exceed the acceptance threshold, the lot is rejected or corrective action is triggered.
• Re-sampling is allowed only under predefined conditions (for example, after corrective action).
• Data integrity and traceability are maintained for audit and trend analysis.

Example

A team needs to verify conformity of 5,000 deliverables with limited time. They set 95% confidence and a 3% margin of error, calculate the sample size, and use systematic sampling (every 50th item after a random start). They apply attribute checks against acceptance criteria, find the defect rate within threshold, and accept the lot while documenting minor issues for process improvement.

Pitfalls

• Using a sample that is too small to support reliable decisions.
• Introducing bias by choosing convenient items instead of random or stratified selection.
• Mixing different conditions in one lot, leading to non-representative results.
• Applying attribute sampling when variable data is needed to assess tolerances.
• Poor measurement system capability causing false passes or fails.
• Ignoring confidence and precision, leading to over- or under-inspection.

PMP Example Question

A project must verify the quality of 10,000 units quickly while controlling inspection costs. Which approach best applies statistical sampling?

1. Inspect every 100th unit after determining the required sample size and acceptance criteria.
2. Inspect 100% of the units to eliminate uncertainty.
3. Select units only from the first hour of production to reduce downtime.
4. Rely on supplier certification and skip incoming inspection.

Correct Answer: A — Inspect every 100th unit after determining the required sample size and acceptance criteria.

Explanation: Statistical sampling uses a defined method and sample size to make inferences about the lot. Full inspection, biased selection, or skipping inspection do not meet the technique's intent.