The Needs of the Customer

Service versus Industry

A customer is anyone who is affected by the products or processes. Three categories of customers exist: (1) external customer (2) internal customer and (3) suppliers who should be viewed as extensions of internal customer departments such as manufacturing. In identifying customers an important tool used is the flow diagram. A cross functional team creates the flow diagram as no one individual or department can describe the total process.

Four terms that are commonly associated with customer behavior are needs, expectations, satisfaction and perception. Customer needs describe the basic physiological and psychological requirements for survival and well-being. Customer expectations are the anticipated characteristics and performance of the goods or service. Customer satisfaction denotes the degree to which the customer believes that the expectations are met by the benefits received. Customer perception is the impression made by the product. Perception occurs after a customer selects, organizes and interprets information on the product.

Needs related to product features: To start, attributes that customers say are important in their purchasing decision must be identified. The important next step is to compare how the product compares with that of the competitor’s. This can be accomplished by performing a multi-attribute study. Market research in the field can provide access to realities that cannot be discovered in a laboratory. Though the laboratory provides a simulation that helps in making decisions, it cannot fully disclose the fitness for use under actual conditions of use.

Needs related to deficiencies: Understanding customer needs also includes the deficiencies side of quality definition. The emphasis must be on defect prevention. Every organization must have a system of collecting and analyzing information on customer complaints. A Pareto analysis of field failures, complaints and product returns identifies the vital few quality problems that are to be addressed in both current and future product development. Data on defects found internally must be analyzed on a proper basis to prioritize prevention efforts which must be addressed during product development.

Customer satisfaction and quality are abstract terms when measuring the level of satisfaction of customers toward a product. Attributes of a product that collectively define satisfaction must be identified. Customer satisfaction measurement should also include asking customers the relative importance of product attributes. Data on customer satisfaction can be collected by methods such as market surveys, post transaction interviews, mystery shoppers and by establishing call centers for complaints and other product related comments. Customer satisfaction data must be linked to customer loyalty analysis.

Designing for Quality

 Statistical Tools in Quality

  1. Histogram: A histogram is a vertical bar chart that depicts frequencies of numerical data. The purpose of a histogram is to provide a pictorial or graphical summary of data. The following histogram is a plot of gas mileage versus the number of cars. The horizontal axis consists of numerical intervals.

Figure 2: Histogram of Gas mileage data

In a histogram the height of the vertical bar tells how many data points were observed for that particular interval.

  1. Scatter diagram: A scatter diagram is used to represent bivariate data and bring out the relationship between them. The purpose of a scatter plot is to visually study the relationship between two variables. The possible patterns observed in a scatter plot can give an idea of the relationship between two variables:

Figure 3: Scatterplot

Application of advanced statistical tools will bring out the perfect relationship between two variables but a scatter diagram is the first step in visualizing and quantifying that relationship.

  1. Run Chart: The run chart is a plot of individual data points versus time. While a histogram summarizes data according to size it misses out on the important dimension of time. Time is generally represented on the horizontal (x) axis and the property under observation on the vertical (y) axis. Run charts are analyzed to find anomalies in data that suggest shifts in a process over time or special factors that may be influencing the variability of a process. A typical run chart is shown below:

Figure 4: Run chart

Design of Experiments

Experiment is defined as the act of observing. It is a sequence of trials or tests performed under controlled conditions which produces measurable outcomes. Prior to performing an experiment, a list of factors that affect the desired outcome must be identified. Experiments are done to (1) optimize the average of a process or product characteristics (2) minimize the variability in a product or process characteristic and (3) minimize the effects of uncontrollable variability on product or process characteristics. An important step to be conducted in the pre experiment analysis is to flowchart the process. It helps everyone understand the process, where it starts and ends and who the suppliers and customers of the process are. A cause and effect diagram must be prepared after flowcharting the process is done but before choosing the experimental factors. Experimental factors must be developed on the basis of flow charts, cause and effect diagrams and the expert knowledge of people familiar with the process. The following classes of experiments are possible:

  1. Experiments with only one factor: This is the simplest of all experiments which involves varying one factor while holding all other factors fixed. Several trials are done at each level of the factor which is being varied. This is aimed at providing the experimenter with the data needed to estimate variation in the response variable for each of the test levels of the factor.
  2. Multifactor two level experiments: Each factor is tested at two different levels. Though restrictive, experimenters working to improve processes have found two level experiments to be very effective. The drawback is that unlike three or four levels experiments, two level experiments cannot detect non-linearity. This can be overcome by doing several two level experiments in place of one three or four level experiment. The data from the series of two level experiments can be used to estimate non-linearity.

The effects of high variability on experimental results can be reduced by performing an experiment more than once which is known as replication. Replications are beneficial because average values have less variability than individual measurements. Replications are done by taking repeated measurements during each experimental trial or by taking one measurement during each experimental trial and then repeating the process several times.

Very often the way in which a factor affects a response variable is independent of the levels of the other factors included in the experiment. But factors do influence each other sometimes. When the effect of one factor is influenced by the level of one or more factors we say that there is an interaction among the factors. It is possible to estimate interaction effects without performing additional trials.

Inspection, Test, and Measurement

Inspection and test include measurement of an output and comparison to specified requirements to determine conformity. Inspection is performed under static conditions on items such as components and can vary from simple visual examination to a series of complex measurements. Conformance to a standard is emphasized here.  Test on the other hand is performed under either static or dynamic conditions and is carried out on complex items such as subassemblies or systems. Test results determine conformance and are used as inputs for other analyses in the form of evaluating new design and diagnosing problems.

The purpose of inspection is to obtain product acceptance. Product acceptance is based on three factors: (1) Conformance- judging whether a product conforms to specifications (2) fitness for use- deciding whether nonconforming products are fit for use and (3) communication- deciding what to communicate to insiders and outsiders about non-conforming products.

Inspection planning is the activity of designating stations at which the inspection should take place and providing those stations with the means for knowing what to do and the facilities for doing it. For simple routine quality characteristics the planning is done by the inspector. For complex products the planning is usually done by specialists such as quality engineers. The tool used for choosing the location of inspection stations is the flowchart.

Inspection accuracy depends on (1) the completeness of inspection planning (2) the bias and precision of the instruments used and (3) the level of human error. Human errors in inspection arise from multiple causes such as technique errors, inadvertent errors, conscious errors and communication errors. The amount of inspection to decide the acceptability of a lot can vary from no inspection to a sample to 100% inspection. The decision is governed by the amount of prior knowledge available on quality, the homogeneity of the lot and the allowable degree of risk. Previous quality history of the product item is also helpful in deciding on the amount of inspection required.

Statistical Process Control

Statistical process control is the application of statistical methods to the measurements and analysis of variation in a process. Control charts are used to track the accuracy (via the mean) and precision or variability (via the range or standard deviation) of a process. Control charts were developed by W.A.Shewhart of Bell Telephone Laboratories in 1924 for process control.

The factors that contribute to product variation can be classified as either inherent (random or chance) causes or assignable causes. Chance causes are considered a natural, consistent part of the process difficult to isolate or even too small to worry about. Some examples are variations in material chemistry or property, measurement errors, machine vibrations and variations in human performance. Assignable causes are events that produce detectable changes in the behavior of the process, measured by the accuracy and precision of that process. These changes are usually large in magnitude and controllable. Examples of assignable causes are tool change in the form of tool wear, temperature fluctuations and pressure variations. When only chance causes are present the process is said to be under control.

Control charts can be used to find out whether variations in product quality are the result of random chance causes or due to a major process change indicating an assignable cause. Many types of control charts are available but the most commonly used one is the X (bar) R chart. X (bar) chart monitors the process mean and the R chart monitors the process variability. Control limits for both charts are set at three standard deviations above and below the process average. The horizontal axis is represented by time or sample number. The following is a representation of an X (bar) R chart:

Figure 5: X-R Control Chart Sample

If a point falls outside the control limits it is possibly due to an assignable cause. A run of seven points above or below the process mean is also an indicator of assignable cause. A common misconception about control charts is that they indicate what went wrong in a process. Control charts should be used as detection devices to indicate when something went wrong but not what went wrong.

Quality Improvement

Six Sigma

The six sigma approach is a collection of managerial and statistical concepts and techniques developed by Motorola in the 1980’s which aims at reducing process variations and preventing deficiencies in the product. Variation in the process is labeled sigma which refers to the standard deviation of measurements around the process mean. A process that has achieved six sigma capabilities has a variation that is very small with respect to the range of specification limits. Six standard deviations can be placed between the process mean and either specification limit. The normal curve for a six sigma process is shown below:

Figure 5: Normal Curve

A six sigma process has no more than 3.4 defects per million opportunities taking into account a 1.5 sigma shift in the process mean with time. The key focus of six sigma approach is the relationship defined between input and output variables which is expressed as:

\( Y=f(X_1,X_2, \ldots , X_n) \)

Product results \( Y \) are a function (f) of many process variables \( X_1……, \ldots , X_n \). \( Y \) the output is the dependent variable and X are inputs, causes that are independent variables. Six sigma approach is about finding the crucial \( X \)’s that have a maximum bearing on \( Y \). Five phases are used in six sigma:

  1. Define: This step identifies potential projects, selects and defines a project and sets up a project team.
  2. Measure: this step documents the process and measures the current process capability.
  3. Analyze: This step collects and analyzes data to determine critical process parameters.
  4. Improve: This step conducts formal experiments to focus on the most important process variables and determines the process settings for optimal results.
  5. Control: This step measures the new process capability and documents the improved process. Controls are instituted to maintain the gains.

The following table shows the defects per million opportunities corresponding to the various sigma levels:

Sigma level

Defects per million opportunities(DPMO)

Percent defective

Percentage yield





























Table 2: Defects versus Sigma Level