A common technique used to analyze and divide data into groups is called stratification. Stratification helps to collect data about a problem or event that needs to be treated separately. Stratification looks at the process data, splits it into distinct layers and does an analysis to see what a different process would entail. Stratification allows you find the metrics to measure your current processes and decide how to change them.

Often a stratifer is used to separate the collected data into subgroups. The subgroups are used to investigate whether a certain factor is causing problems and can be changed. The stratifiers allow you to design the manufacturing experiment.

Designing the manufacturing experiment will involve using your collected data subgroups and deciding what factors inter-relate. Typically the design of experiment (DOE) is an organized, structured method to determine the relationship between factors (x) affecting a process and the output of that process (y).

Design of experiment will involve designing a set of 10-20 experiments that are relevant factors to be varied systematically. The result of the DOE's will identify the optimal conditions, what influences the results, what does not influence the results, and what are the details of interactions between the different DOE's.

DOE was first developed by Sir Ronald A. Fisher in his book, The Design of Experiments. In his book he described how to test the hypothesis that a certain lady could distinguish by flavor alone whether the milk or tea was first placed in the cup.

This allowed Fisher to illustrate the important means of his design which include:

• Comparison
• Randomization
• Replication
• Blocking
• Orthogonality

Comparison refers to reproducing measured results as closely as possible. Comparisons between treatments are reproducible and easier to measure. Randomization utilizes mathematical theories that explore the consequences of making the allocation of units to decide on tables to measure random numbers. Random allocation will be calculate the data and allow it to be managed easier. Replication will be used to evaluate if a subject is associated with a certain criteria that can be estimated. Blocking is the arrangement of experimental units into blocks that are similar to one another. Typically this looks at the variation between units and allows for greater estimation and precision of the source of variation under a study. Orthogonality concerns the contrasts that can be legitimately carried out. The contrasts can be represented by sets of orthogonal contrasts and distribute the data sets.

DOE methods require well-structured data matrices.

When you are designing your manufacturing experiment, you need to consider the following:

1. Define the objective to the investigation

2. Define the variables that will be controlled during the experiment

3. Define the variables that will be measured to describe the outcome of the experimental runs

4. Choose a standard design that has a reasonable cost

A standard design can be generated automatically as soon as you have your objective, the number of variables, and the nature of the responses and experimental runs you can afford. DOE's are commonly used in research and development to solve optimization problems. The way to minimize your optimization costs are to conduct as few experiments as possible.

A properly designed and executed experiment will generate better, more precise data while using fewer experimental runs that alternative approaches. The results can then be interpreted using statistical techniques. The reason to have statistical results over observational results is because the information from an observational study can be difficult to interpret.

Using the process validation concept you can design your manufacturing experiment. The process validation will be used to accumulate data that will provide results to a predetermined requirement. It is necessary to make sure your products will perform safely and effectively. This is why process validation can come in helpful.

Start by deciding on a sequence of events in the product development phase that lead to tasks that need to be accomplished at different phases and the tools that need to be used. You will determine what the consequences will be for different tests and conditions. The key elements of the process should include some of the following: equipment, materials, people, etc.

There are traditional and factorial designs that need to be taken into consideration when you design your manufacturing experiment. A traditional method is to evaluate one variable at a time. This will allow you to find the effect of that chosen variable under different circumstances.

The factorial design combines all the different levels for all the different factors. With a the factorial design, the testing will reveal what the effect of one variable will be with changing factors. The factorial design is also efficient. Only one test will be needed with a factorial design versus the numerous tests that you have with a traditional design. A factorial design also makes it possible to estimate the interaction between 2 different factors. Your manufacturing experiment should be carefully planned and statistically designed. The DOE concept is easy to understand and implement into your manufacturing organization. The factorial designs are also easy to construct and implement. The end results will be easily interpreted and will lead to justified conclusions.

You can purchase computer software that will help you implement your DOE strategy.

When you are selecting DOE software it is important to look for software that offers the following:

• An easy to use interface
• A well-written manual that includes start up support
• Designs for screening and optimizing process or product formulations
• A flexible spreadsheet where data can be entered
• A spreadsheet that can deal with missing data and changing factors
• Software that allows for experimental runs to be tested

DOE software should determine which terms you want included in your model and will compute all the coefficient values that are associated with errors and values. A good DOE software package will offer the users the capability to augment an existing design. DOE software should also include a thorough training process. Most of the potential DOE software buyers are intimidated by the perceived difficulty of using DOE software.

One DOE software company offers a 3 day computer workshop in addition to extra courses to properly train the DOE user. By using DOE software, you will save your company thousands of dollars on a yearly basis. Having proper telephone support will also help you use the software properly and adequately.

Ask your colleagues about what type of DOE software they are currently using. Do your research on different DOE software programs and find one that is tailor made to your particular data sets. The software should allow you to easily input your data and run the data and receive easy to read results. The goal of DOE software is to tell you how to change the inputs in order to change the outputs.

With the use of DOE software, you have designed your manufacturing experiment. The next step is measuring and charting your data. DOE software can help you measure and chart your data. The Six Sigma approach stresses the importance of results that come directly from your data that has been measured and properly charted.

The place to begin to measure and chart your data is with proper training. Your staff must be properly trained in order to maintain consistency throughout your entire manufacturing process. The training must be planned out and a training session must be practiced on management before sent down to the rest of the company. Taking into account several types of questions will help to evaluate the training plan.

Some of the questions are as follows:

• What are the training objectives?
• Who will receive training?
• How long is the training?
• What methods will be used?
• Who will be qualified to conduct the training?
• What is the criterion for instructors?

A number of variables will be created and each variable need to have its own measuring and charting procedure. Decide which variables are the most pressing and need to be charted first. These variables should be directly tied to immediate results. Each division in your manufacturing organization should have a reporting and data collection process. The different data sets can be measured and charted by each branch of your company or they can be reported to a trained individual who is in charge of the data.

The Six Sigma methods require your manufacturing organization to work as a team. You cannot obtain the proper results without clear communication and hard work. If you have each division in charge of reporting different data sets, you need to have a clearly defined procedure for this process. If there is the slightest bit of confusion, you can loose data or it can become flawed and will not be useful for your company.

In the end, your goal is to reduce the costs, eliminate the waste, and improve the productivity and customer satisfaction in your manufacturing organization. The Six Sigma method combined with the DOE approach will help you to achieve these goals.