Q 531. What is the difference between Control Limits and Specification Limits?

 

🔹Best Answer:

Click here to jump to the best answer

Difference between control and specification limit is as follows:

Control Limits defined via UCL and LCL	Specification Limits defined via USL and LSL
Derived from process (calculated from past measurements) – Voice of process Calculated typically as +- 3sigma from mean	Received from customer/ market – Voice of customer. It shows the expected performance
Factor in calculating process stability	Factor in calculating process capability
Appears in control charts and understand variations the process is liable to produce	Appears in histograms, box plots and probability plot and helps understand acceptable deviation from the performance target
Helps understand process variation	Helps understand gaps / defects in process performance vs. expected performance
Process control limits can be shifted or changed over time	Not much influence to change specification limits
For sub groups	For item or individual measurements
Helps identify and reduce producer’s rejects	Helps identify and reduce customer’s rejects

 

If a process is called out as in statistical control, then any variation in the process performance would be attributed to common causes only. This process would be consistent over time. If any process performance is noted beyond the control limits, it would be attributed to special causes and predicting performance for such a process would be difficult.

Therefore control limits indicates that majority of the variation (+- 3 standard deviations from the process performance mean) in the process is within these limits and only a small portion of the variation is outside which would be attributable to special causes.

 

Whereas the individual process performance measurements that fall out of customer provided specification limits would be said to not meeting the customer’s expectations. While measuring process capability, we generally use Cp, and Cpk when process is called out as in statistical control. We would use Pp and PPk to measure process performance , generally when process is too new to determine if its in statistical control.

 

Interaction between the control and specification limits:

a. if control limits fall within the specification limits

This is a good scenario as all variations due to common causes will still meet customer expectation. Process therefore has higher capability to meet customer expectations. Also, in future if the control limits change, process capability may not be as severely affected

 

b. if specification limits fall within the control limits:

This indicates that there are some measurements or process performances which are not meeting customer expectations even though they may have been caused by common causes or in other words there are larger number of defects.

 

c. if control and specification limits coincide:

While all variations caused due to common causes will meet customer expectations, any variations caused by special causes will not meet customer expectations.

 

d. No overlap between control and specification limits:

This then indicates that entire process performance or all the measurements are not meeting customer expectations

 

 

Difference between Control Limit and Specification Limit-

 

 

s.no	Type	Specification Limit	Control Limit
1.	Nature	External to the process	Internal to the process
2.	Reflect	Requirement to satisfy a CTQ characteristics	Expected Range of Variation for that process
3	Value defined	Individual Values	Sample average value
4	Determined by	Customer	Calculated as three standard deviation away from the mean or either side of the mean
5	If points outside the limit	Points outside the specification limit is considered as Defect	Point outside the control limit shows special cause
6	Upper and lower limit called	Upper limit – USL Lower Limit- LSL	Upper limit – UCL Lower Limit- LCL

 

Starting with the basics,

Control limits are the process's voice (what the process does) and

Specification limits represent the customer's voice (what we want the process to do).

 

Lower and upper control limits are LCL and UCL, respectively.

Lower and upper specification limits are LSL and USL.

 

In general, these limits represent our process variation and help highlight when our process is out of control.

x̄ - center line for the data is the sum of all the input data divided by the total number of data points

LCL – is 3 process standard deviation below the average and

UCL – is typically 3 process standard deviation above the average

 

If the data point falls within ±3SD of its average, it is considered as “EXPECTED” behavior for the process and thus is a common cause variation. Something special happened to those data pointers outside these control limits and can be special cause variation.

 

Let’s consider real-time data from the 2022 Japanese Grand Prix. The qualification session usually determines the starting order and the pole position of the race. To occupy pole position, the fastest driver must ensure that their performance is the quickest, that is, with the lowest lap timing.

Below are the lap time/s (actual performance) for the qualifying session.

LCL 01:29.304

UCL 01:31.511

Five of the slowest cars are eliminated.

Now comes the specification limit, which is set based on the 107% rule.

For instance, if the fastest lap time was 100 seconds, each driver who is eliminated in the session must complete at least one lap within 107 seconds to guarantee a race start, which is the USL. Often, only one specification limit is used as in this example. 

 

Control limits are applied to summary statistics, whereas specification limits are applied to individual measurements. Control and specification limits are extensively used in control charts and can give us an early warning if a process is showing irregularities, giving us the opportunity to take remedial steps before the situation becomes a problem.

 

Let’s consider another example from the customer contact center: if the average handling time is 4 minutes with a standard deviation of 1 minute, then the control limits are UCL = 7 minutes and LCL = 1 minute, respectively. Specification limits are the targets for the process and defined by the customer or based on the performance of the market. It is desired that those control limits be within the specification limits so that, in case of special causes, the customer will not be impacted.

 

Application of control limits in control chart:

The position and scattering of data points plotted on the control chart assist us in identifying process behavior. Process behavior includes that of identifying the stability and understanding the pattern of process variation from a special and common cause viewpoint.

 

Let’s consider one more example based on the below referred Shewhart chart of a manufacturing unit of Prod X.

 

 

If we could notice that none of the data points are outside the specification limits, and usually the production management team will be OK and glad about this type of run, it is worth noting that if the process were managed statistically, these patterns would assist the line engineer in adjusting and retargeting the process. However, the effectiveness of re-targeting will be determined by the usefulness of the process gain factors. 

 

Consider the following scenarios to better understand process performance:

Scenario 1: The specification limit is within the control limits.

Scenario 2: The specification limit is the same as the control limit.

Scenario 3: The specification limits exceed the control limits. 

 

 

In scenario 1, part of the process (natural variation) may function outside of customer-defined targets, leading to rejections and defects.

 

In scenario 2, the process may meet customer specifications, but it can produce defects when there is an uncommon source of variation.

 

In scenario 3, which I would call an optimal scenario, the production is stable, within its capability, produces no defects or rejections, and of course meets the customer specification.

 

To summarize,

• Control limits are calculated from the process data, so they are the voice of the process, and specification limits are defined by the customer, so they are the voice of the customer.
• Control limits emphasize location, spread, and width, whereas specification limits focus on meeting the requirements.
• Control limits are applied to subgroups, and specification limits are applied to items.
• Control limits help reduce internal rejections in the process, and specification limits help reduce customer rejections.
• Control limits are displayed in the control chart, and specification limits are displayed in the histogram and probability plots.

Quote

“We draw limits on the graphs to indicate the standards for evaluation. These lines will indicate the dispersion of the data on a statistical basis and let us know when an abnormal situation occurs in production.” - Dr. Ishikawa

 

Control Limits Vs Specification Limits

 

· Control limits represents the voice of the process whereas the Specifications Limits represent the voice of the customer and talks about SLA/tolerance and tells us acceptance/rejection criteria

· Control limits can be calculated from Data (Control Limits = X + 3S, X-3S) while Specifications Limits is defined by customer and it is always given (No formula to calculate Specifications limits)

· Control limits appear on control charts where control limits are for the sample averages whereas Specifications Limits appear on histograms, box plots, or probability plots. Specification limits are for individual values

· Control limits apply to subgroups, guide for process actions and tells us what the process is doing whereas Specification Limits apply to items, separate good items from bad and tells us what we want the process to do

A limit is the optimal value which is allowed. In order to achieve a stable process and capability in six sigma, control limits and specifications limits are widely used. Control Limit(CL) and Specification Limit(SL) are mathematical restraints unlike sky is the limit which know no boundary. However, both have different applicabilities and vary completely in terms of concepts and statistical perception.

 

Essentially, CL is the voiceprint of the process which act as an indicator to apply actions in a process. It is computed based on measured  data population which is presented on Control Charts. Usually an X-bar chart is used for representation. Such values are then attributed to subgroups to reveal the performance of the controlled process. These limits are set at the center of +/-3(sigma) such that the center and the sigma are enumerated according to the selected control chart.

 

It exist 2 types of CL being Upper Control Limit(UCL) and Lower Control Limit(LCL).

 

In simple terms, UCL is obtained by adding the mean with 3 times the standard deviation(s.d): (Mean + (3*s.d)). On the other side, LCL is obtained by deducting the mean with 3 times the s.d: (Mean - (3*s.d))

 

An illustrated X-bar where the specifications fits into the control limits is shown below.

 

SL is the voiceprint of the customer which acts as a law maker to produce what the process really have to materialized. It is designated by the customer with regards to the characteristics set out which is presented on Histograms. Probability and box plots are also used for representation. These values are attributed to items rather than subgroups in order to unfold the good items away from the bad one.

 

The Upper Specification Limit(USL) and Lower Specification Limit(LSL) are the 2 types for SL. USL and LSL are used to gauge the level of customer demand. If the item fits the range of the limits, then the item is considered to align with the needs of the customer. Else, if the items is out of range, the item does not comply with what the customer expected.

 

A vivid representation of SL using histogram is illustrated below.

 

To have a clear picture of CL and SL, let’s take a look upon the relationship between these two.

1. If the CL fits into the specifications, then it is indeed the top notch design for an organization process. This is so because any variation emanating from known source will fit into the specifications range. In such situation, even if the CL vary over a period of time, it will not sever alarming consequences to render the customer specifications as required.
2. If the CL is equivalent to the SL, then all variations which are known from the process will definitely align with the specifications of the customer. The major stumbling block in this situation however will be to make the process undergo statistical control due to the appearance of any unknown variations will result into defects.
3. If SL fits between CL, there will a portion of the process that will cause dysfunctional operability. When the CL surpass the SL, some process will be operating beyond the specifications quotas. In other words, having CL larger than SL will without a doubt lead to the creation of defects.
4. If there exist no linkage between LSL and USL, it is clear enough to conclude that the whole process is unfeasible with regards to the specifications meaning that the process itself is defective. This can happen whenever the design of the process is by all means incapable to unite customer specifications. It can happen that the process is behaving correctly but the expectations from the customer is erroneous or the process has triggered highly deficient issues from the design being incapable to meet what the customer expected.

 

A simplified scenario presented below to better understand how CL and SL is being applied winning a pizza competition.

 

Panecuit is a known pizza maker in Paris who wants to compete for the international best pizza in Italy. He looks out for the best Chef in Italy to gave him a helping hand to compete against the other participants. The Chef gave him the records of the top performance statistics and provide him with valuable piece of information. The key to reach the pinnacle to win is to complete the pizza within the timeframe of 20 seconds(sec) to 25 seconds. It is within that range that the preceding competing participants won the competition. Without any further do, Panecuit get holds unto practicing and begin countdown his performance. Under his scrutinize monitoring of performance, his range of having a pizza into the oven is between 22sec to 30sec, considering that deviation of 3 sigma with respect to the mean performance. At that time, 20sec to 25sec was the objective to be achieved while 18sec to 26sec was the current performance.

 

One important factor to be taken into observation. CL is primarily used to analyze variations arising out of performance. Even Panecuit’s performance ranges from 18sec to 26sec, it still pose a matter of concern for the competition. The other participant, Cuitlente, might present a steady performance which will make him stand out as a winner. He is performing with a countdown ranging from 20sec to 24sec having an identical mean of performance as Panecuit with 24sec. What is to be noted here, is that Cuitlente clearly shows to be more consistent and display few variations and thus preference will be given to Cuitlente rather than Panecuit.

 

Let’s introduce Cuitvidé, with a consistent CL of 20sec to 22sec. Cuitvidé showing clear sign of constrict variations and more consistency within the timeframe, that is the specifications area. Hence, he will stood at a winning advantage as compared to the remaining.

·        Specification limits i.e., Upper Specification Limits (USL) and Lower Specification Limit (LSL) are the voice of customer/ requirements/ selection-rejection criteria specified by the customer. This can be based on the specific requirements or industry/ company benchmarks.

 

Control limits i.e., Upper Control Limit (UCL) and Lower Control Limit (LCL) are the calculated (statistical) values from process data.

UCL = x + 3s and LCL = x- 3s where x is the mean or the target value & s is the standard deviation

 

·        Specification limits are used to check adherence or conformance for accepting/ rejecting the products or services. Control limits are used to monitor and control the performance of the process, i.e., if the process is within the control limits it suggests that the assignable causes have been eliminated and variation is due to randomness. In this case, the process is considered as stable and following a normal distribution.

 

·        Specification limits are used to segregate good & bad products/ services while control limits are used to identify the actions on process of producing/ delivering the products/ services.

 

·        Specification limits are straight lines while control limits may not be straight lines for variable sample size between the sub-groups.

 

·        The ideal scenario is to have control limits within the specification limits. This suggests the process is efficient to deliver products/ services within the conformance criteria.

 

·        If the control and specification limits coincide, any variation due to common causes (assignable) will produce defects.

 

·        It the specification limits are within the control limits, some part of the process will always be outside the specification and the process will produce many defects.

 

·        If specification and control limits do not overlap at all, the process will never produce an acceptable product/ service. Either the process needs to be drastically optimized or the customer expectations are too far off.

Control Limits vs Specification Limits

 

Control Limits are derived from the process variations while specification limits are derived from the customer requirements.

First let’s look at the definition of both these terms

Control Limits: Control limits are drawn on control charts to represent the Upper and Lower Limits of the acceptable range for a process. A process is said to be out of control when the data goes beyond these control limits. Typically control limits are set at 3 times the standard deviation.

 

Specification Limits: Specification Limits are targets set for the product or process by the customer, a regulatory body or the end user. Specification limits can be quantified as the point where the losses due to variation in the product or process is equal to the benefit of the same. Generally, the spec limits are drawn on a histogram.

All the published answers are correct 
 

Dheeraj - Tabulated differences
Vijay Krishnan - Description with displayed images. 
Anupam - Provided a tabulated comparison. Explained all scenarios. 
Nunhuck - Explained all scenarios and provided a beautiful example of Pizza-making. 

Anuj - Provided very good description and explained different scenarios. 
Asif - Detailed explanation with illustrations (each illustration shows control limits as well as specification limits), description of scenarios, more examples and a nice summary.

 Mohd Asif is the winner and he takes his best answer tally to 37. He continues to lead in the all time tally.