U Chart (Defects per Unit)
Monitor the rate of defects per unit when sample sizes vary across time periods.
Chart Overview & Variable Sample Size Monitoring
Defect Rate Statistic & Control Limit Formula
Average defect rate (centerline)
= average defects per unit across all subgroups
U chart control limits
= sample size for the subgroup
= average defect rate (centerline)
If the calculated LCL is negative, it is set to 0.
Reading the u Chart
Example Chart & Out-of-Control Signal Interpretation
sixsigmacenter.com Implementation Guide
This platform provides comprehensive U chart analysis with automated control limit calculations, variable sample size handling, and detailed interpretation guidance for defect rate monitoring.
Automated Features:
- • Variable control limit calculations for each point
- • Automatic Poisson assumption verification
- • Out-of-control point identification
- • Trend and pattern recognition
- • Process capability metrics for defect rates
Common Applications:
- • Manufacturing: Surface defects per square meter
- • Software: Bugs per lines of code
- • Healthcare: Infections per patient day
- • Service: Complaints per customer transaction
- • Quality: Nonconformities per inspection unit
Best Practices & Guidelines
- • Ensure sample sizes are large enough for meaningful results
- • Consider using standardized control limits for very small samples
- • Investigate both increases and decreases in defect rates
- • Document the operational definition of a "defect" clearly
- • Use probability limits (2σ) for early warning systems
- • Consider transformations for highly variable defect rates
Example Data Layout for U Chart
Data for U Chart with variable inspection units. Calculate defects per unit (defect rate) to monitor quality with varying sample sizes. Sample Size Recommendation: Minimum 20-25 subgroups with adequate inspection units per period - larger samples provide more stable defect rate estimates for variable control limit calculations.
This is how your spreadsheet or CSV file should be structured before you upload it. Match the column names, keep one row per observation, and leave the analysis to the platform.