ISO 7870-4:2021 - Control charts - Part 4: Cumulative sum charts
Standard reference: ISO 7870-4:2021 (International Organization for Standardization)
Overview
ISO 7870-4:2021 specifies statistical procedures for designing and applying cumulative sum (CUSUM) charts for process and quality control. The standard covers both variables (measured) and attribute (discrete) data, and provides general‑purpose decision rules for online monitoring, control and retrospective analysis. It explains the CUSUM principle - subtracting a target/reference value from each observation and cumulatively summing the differences - and compares CUSUM performance with Shewhart and EWMA charts.
Key topics and technical requirements
- Fundamentals and construction: principal features, graphical construction steps, worked example (motor voltages).
- Decision rules and performance: need for rules, decision bases, and performance metrics including Average Run Length (ARL) calculations.
- CUSUM decision schemes:
- V‑mask configuration and application
- Fast‑Initial Response (FIR) CUSUM
- Tabular CUSUM (rationale and deployment)
- Monitoring approaches:
- CUSUMs for monitoring location (mean shifts) and variation (subgroup ranges and standard deviations)
- Selection of target/reference values (standard vs performance‑based)
- Special situations: large between‑subgroup variation, one‑at‑a‑time data, serial dependence, and outliers.
- Discrete data methods: CUSUM schemes for Poisson (event counts) and Binomial (two‑class) data.
- Annexes and tools: informative annex with tabular CUSUM example and estimation of change point when a step change occurs.
- Normative references: ISO 3534‑1 and ISO 3534‑2 (statistical vocabulary and symbols).
Practical applications
ISO 7870-4 is practical for organizations seeking sensitive, visual methods to detect process shifts earlier than standard Shewhart charts. Typical applications include:
- Manufacturing process monitoring (individual measurements and subgrouped data)
- Service metrics and business indicators (counts, rates, yes/no outcomes)
- Healthcare surveillance (event counts), reliability testing, and maintenance monitoring
- Quality improvement initiatives and retrospective root‑cause analysis
CUSUM charts are especially valuable when early detection of small to moderate shifts in process mean or event rates is critical.
Who should use this standard
- Quality engineers and process control specialists
- Statisticians implementing Statistical Process Control (SPC)
- Operations managers, reliability engineers, and regulatory compliance teams
- Software vendors and data scientists building SPC/CUSUM modules
Related standards
Keywords: CUSUM, cumulative sum charts, control charts, ISO 7870-4, process control, quality control, SPC, V‑mask, tabular CUSUM, Poisson CUSUM, binomial CUSUM.
