Skip to main content
diversification.com
← Back to P Definitions

P chart

What Is a P chart?

A P chart is a type of control chart used in Statistical Process Control (SPC) to monitor the proportion of defective or non-conforming items within a process over time. It is particularly useful when dealing with attribute data, where items are classified into one of two categories, such as "pass" or "fail," "conforming" or "non-conforming," or "present" or "absent." The P chart helps organizations identify whether a process is stable and predictable or if it is experiencing unusual variation that requires investigation and process improvement. By plotting the proportion of defects from successive samples, the P chart provides a visual representation of process performance and helps in making data-driven decisions. It is a fundamental tool for quality control and continuous monitoring.

History and Origin

The concept of the control chart, from which the P chart derives, was developed by Walter A. Shewhart at Bell Telephone Laboratories in the 1920s. Shewhart's groundbreaking work laid the foundation for modern Statistical Process Control (SPC). On May 16, 1924, Shewhart created the first control chart, a tool designed to differentiate between common cause variation (inherent and random fluctuations within a stable process) and special cause variation (unpredictable variations caused by specific, identifiable factors)15. His aim was to help manufacturers reduce variability and improve product quality by identifying and eliminating these assignable causes14. The principles established by Shewhart were later popularized by W. Edwards Deming and became integral to various quality management frameworks, influencing industries globally and leading to the formation of professional societies like the American Society for Quality (ASQ)13.

Key Takeaways

  • A P chart monitors the proportion of defective units in a process, making it suitable for binary (pass/fail) data.
  • It helps distinguish between common cause and special cause variation, indicating whether a process is in a state of statistical control.
  • The chart plots sample proportions over time, along with a central line and upper and lower control limits.
  • Points falling outside the control limits or exhibiting non-random patterns signal that the process may be out of control.
  • P charts are valuable for continuous monitoring and identifying opportunities for process improvement.

Formula and Calculation

The P chart uses the proportion of non-conforming items in a sample. The key elements for its calculation include the center line (average proportion), the upper control limit (UCL), and the lower control limit (LCL). Since the sample sizes can vary, the control limits for a P chart will also vary from sample to sample12.

The average proportion of non-conforming items ($\bar{p}$) is calculated as:

pˉ=Total number of non-conforming unitsTotal number of units inspected\bar{p} = \frac{\text{Total number of non-conforming units}}{\text{Total number of units inspected}}

The standard deviation of the sample proportion ($\sigma_p$) for a given sample size (n) is estimated as:

σp=pˉ(1−pˉ)n\sigma_p = \sqrt{\frac{\bar{p}(1 - \bar{p})}{n}}

The control limits are then calculated as:

UCL=pˉ+3σp=pˉ+3pˉ(1−pˉ)nLCL=pˉ−3σp=pˉ−3pˉ(1−pˉ)nUCL = \bar{p} + 3\sigma_p = \bar{p} + 3\sqrt{\frac{\bar{p}(1 - \bar{p})}{n}} \\ LCL = \bar{p} - 3\sigma_p = \bar{p} - 3\sqrt{\frac{\bar{p}(1 - \bar{p})}{n}}

If the calculated LCL is a negative value, it is set to 0, as a proportion cannot be negative. The factor of 3 in the formula represents three standard deviations from the center line, which is a common practice in Statistical Process Control to define the control limits11.

Interpreting the P chart

Interpreting a P chart involves examining the plotted points in relation to the center line, the upper control limit (UCL), and the lower control limit (LCL). The goal is to determine if the process is stable and predictable or if it exhibits unusual variation. A process is considered "in control" if all plotted points fall between the UCL and LCL and show no discernible patterns or trends. This indicates that any observed variation is due to common causes inherent in the system. Statistical Process Control charts, including the P chart, help to visually represent data over time, enabling timely detection of deviations10.

Conversely, a process is considered "out of control" if any points fall outside the control limits. Such points signal the presence of special cause variation, which requires investigation and corrective action. For instance, a point above the UCL might indicate a sudden increase in defects, while a point below the LCL could suggest an unexpected improvement in quality. Other "out of control" signals include specific patterns, such as a run of several points on one side of the center line, or a trend of consecutive points steadily increasing or decreasing9. Analyzing these signals is crucial for effective data analysis and targeted interventions.

Hypothetical Example

Imagine a bank's loan processing department wants to monitor the proportion of loan applications that contain errors. They decide to use a P chart to track this attribute. Each week, they randomly sample 100 processed loan applications and count how many have errors.

Let's assume over the past 20 weeks, they have processed a total of 2,000 applications (20 weeks * 100 applications/week) and found 100 applications with errors.

First, calculate the average proportion of errors ($\bar{p}$):

pˉ=Total errorsTotal applications inspected=1002000=0.05\bar{p} = \frac{\text{Total errors}}{\text{Total applications inspected}} = \frac{100}{2000} = 0.05

Next, calculate the control limits for a sample size ((n)) of 100:

σp=0.05(1−0.05)100=0.05×0.95100=0.0475100=0.000475≈0.02179\sigma_p = \sqrt{\frac{0.05(1 - 0.05)}{100}} = \sqrt{\frac{0.05 \times 0.95}{100}} = \sqrt{\frac{0.0475}{100}} = \sqrt{0.000475} \approx 0.02179 UCL=0.05+3×0.02179≈0.05+0.06537=0.11537LCL=0.05−3×0.02179≈0.05−0.06537=−0.01537UCL = 0.05 + 3 \times 0.02179 \approx 0.05 + 0.06537 = 0.11537 \\ LCL = 0.05 - 3 \times 0.02179 \approx 0.05 - 0.06537 = -0.01537

Since the LCL is negative, it is set to 0.

Now, let's track the proportion of errors for the next five weeks:

  • Week 21: 3 errors out of 100 applications = 0.03 (In control)
  • Week 22: 6 errors out of 100 applications = 0.06 (In control)
  • Week 23: 12 errors out of 100 applications = 0.12 (Out of control - above UCL!)
  • Week 24: 4 errors out of 100 applications = 0.04 (In control)
  • Week 25: 7 errors out of 100 applications = 0.07 (In control)

The P chart would show the point for Week 23 above the UCL, signaling that the loan processing procedure experienced a special cause of variance that week, which needs immediate investigation. This practical application of data visualization helps pinpoint unusual events.

Practical Applications

While originating in manufacturing, the P chart has broad applicability in various sectors beyond traditional production lines, including financial services and administrative processes. It is a key component of Statistical Process Control frameworks used for continuous performance measurement.

In financial services, P charts can be used to monitor:

  • Banking: The proportion of erroneous transactions, incomplete loan applications, or fraudulent claims processed daily or weekly. For example, a study analyzed yearly insurance claims data using a P chart to assess whether the company's claim process was within statistical control relative to pre-fixed margins8.
  • Insurance: The proportion of policy applications with missing information, incorrectly processed claims, or customer complaints.
  • Investment Operations: The percentage of failed trades, settlement errors, or compliance breaches within an investment strategy.
  • Credit Card Companies: The proportion of disputed charges or late payments within a specific portfolio segment.
  • Auditing and Compliance: The rate of detected audit findings or non-compliance instances, enabling risk management professionals to track and mitigate risks.

By continuously monitoring these proportions, organizations can identify shifts in their processes, address underlying issues promptly, and work towards greater efficiency and higher return on investment.

Limitations and Criticisms

While a powerful tool for Statistical Process Control, the P chart, like other control charts, has certain limitations. One significant drawback is its reliance on attribute data, which means it can only classify items into two categories (e.g., conforming or non-conforming)7. This simplification can sometimes lead to a loss of detailed information compared to control charts that use variable data (e.g., actual measurements of length, weight, or time).

Furthermore, setting up and maintaining P charts requires careful sampling and consistent data analysis. If the sample size is too small, the control limits may not be accurate, potentially leading to false alarms or missed signals of process changes6. Conversely, if sample sizes are excessively large, the chart might identify statistically significant changes that are not practically meaningful5. A common pitfall is misunderstanding that control limits represent the "voice of the process" (what the process is capable of), while specification limits reflect the "voice of the customer" (what the customer requires); confusing these can lead to misinterpretations4.

Moreover, P charts detect when a process is out of control, but they do not automatically explain why it is out of control or how many defective products have been produced until that point3. Identifying and addressing the root cause of special variation still requires further investigation and other quality tools. Implementing SPC and P charts without taking action on the insights gained will not yield improvements2.

P chart vs. np chart

The P chart and the np chart are both types of attribute control charts used in Statistical Process Control to monitor non-conforming units. The primary distinction lies in what they plot on the chart.

The P chart plots the proportion of non-conforming items in each sample, making it suitable when sample sizes vary between subgroups. This is because the proportion inherently adjusts for different sample sizes.

In contrast, the np chart plots the actual number of non-conforming items in each sample. For an np chart to be effectively used, the sample size (n) for each subgroup must remain constant. If the sample size varies, the np chart's control limits would fluctuate, making interpretation more complex and potentially misleading. Therefore, if data is collected in subgroups of varying sizes, the P chart is the more appropriate tool. If the sample size is fixed and consistent across all observations, either chart can be used, though the np chart might be simpler to interpret as it deals with raw counts rather than proportions. Both charts use the binomial distribution as their underlying statistical basis1.

FAQs

What kind of data is used with a P chart?

A P chart is used with attribute data, specifically when classifying items into two categories, such as "defective" or "non-defective," "yes" or "no," or "pass" or "fail." It monitors the proportion of items falling into one of these categories.

When should I use a P chart instead of other control charts?

Use a P chart when you need to monitor the proportion of non-conforming units and your sample sizes may vary. If your data consists of actual measurements (e.g., length, weight), you would typically use variable control charts like X-bar and R charts. If you are counting the number of defects per unit where multiple defects can occur on one item, a c-chart or u-chart might be more appropriate. If you are counting the number of defective units and your sample size is constant, an np chart can be used.

What do the control limits on a P chart represent?

The control limits on a P chart, typically set at three standard deviations from the center line, define the expected range of variation for a stable process. Points falling within these limits suggest that the process is operating consistently due to common causes. Points outside these limits, or specific non-random patterns, indicate the presence of special causes of variation, which warrant investigation and data analysis.

Can P charts be used outside of manufacturing?

Absolutely. While originating in manufacturing, P charts are highly versatile. They are applied in service industries, healthcare, administrative processes, and even financial analysis to monitor proportions of occurrences like customer complaints, billing errors, data entry mistakes, or incomplete forms. Their utility extends to any process where the proportion of binary outcomes needs to be tracked for quality control.