Consulting firms are hired very often to help optimize processes and improve the quality of the products delivered. One of the frameworks used is the so-called Six Sigma which tries to find the root cause of problems using simple statistical analysis

In this post, I will show you the main concepts. For more details, I recommend checking the presentation on Six Sigma or my online course Essential Six Sigma for Management Consultants

What is Six Sigma?

Six Sigma is a data-driven method used to improve the process. Using this method you reduce the so-called defects, and you increase the consistency of results. This in turn will lead to improvements in quality. Six Sigma helps identify what is slowing down the process, how to eliminate possible delays, how to improve the process and remove additional problems that occur. You lower the number of defects by either: improving the average results or increasing the consistency of results.

Since defects are so important let’s see what we actually mean by this:

In Six Sigma you can use 2 main approaches: DMAIC (used for improving existing processes) and DMADV (used to develop new processes or products at highest quality levels).

We will concentrate on the DMAIC method. Let’s see what it is and what stages it has:

Define Stage

Now let’s briefly see what you do in each and every stage of the Six Sigma project if you use the DMAIC approach. We will start with the Define Stages. In this stage you concentrate on defining well the process you are trying to improve and the problem connected with the process you are trying to solve. In the Define Stage you will do the following things:

1. Define the problem. To optimize a process, you first have to define what the problem you are trying to solve. Let’s briefly look at how we do that.
   
2. Understand what customers need. The process you are trying to improve has to satisfy the customer, so in order to optimize it you have to understand the customer well. To achieve that you will use things like Voice of Customer, Define Needs & Requirements, and Critical-to-Quality Approach (CTQ). All of those methods try to translate information provided by the customer into concrete things that can be measured and later optimized.
3. Describe the process linked to the problem. The next stage is to describe the process linked to the problem. Ideally, you want to map it. One of the most often used methods is the so-called SIPOC. Let’s look at a short example of a SIPOC:
   
4. Pick the KPI you will concentrate on. After we have the necessary information, you can finally decide what you want to improve. In previous stages, you were looking at many parts and KPIs of the process. Now you have to decide on what you will focus on. Do you want to make the process faster? Or maybe you want to improve the perceived quality of the process? The choice should be made in line with the problem you want to solve. So, if the customer wants a faster process, most likely this is what you should concentrate on.
5. Guess what the root cause could be. Once we know what we will try to improve, it’s time to play the role of detective and try to guess what may be causing the problem (for example slowing down the process). You will use for that, 3 main methods: Fishbone Diagram, 5 Why, and Issue Tree. Let’s look at the Fishbone Diagram:

and an example of this method applied to a restaurant chain

Measure Stage

In the second stage, we have to start gathering data to confirm or deny our guesses from the first stage (Define Stage). Just as a reminder, in the previous stage, we defined the problem, the process linked to it, and the KPIs that we want to improve. After that, we tried to guess what generated the problem. To confirm whether we are right we need lots of data.  In the Measure Stage we will do the following things:

1. Define things you want to measure. Firstly, you have to name the things you want to measure. Usually, you will try to collect data on the KPI (output variable) you want to improve and on factors that may influence the KPI (factors guessed by you – input variables)
2. Define Sources of Data. Once you know what to collect, you have to define where you will get the data from. Some data will be available in financial systems, MRP systems, and CRM systems. Other data you will have to collect additionally for example by conducting customer research
3. Create Data Collection Plan. In some cases, you will have to define a formal data collection plan. In this plan, you must describe in more detail what data you need. In the data collection plan, apart from defining what you want to measure, and where you will get the data from, it’s a good idea to also define the level of detail of the data, and the type of data.
4. Define Sample Size. You will not always have the chance to gather all the data for the whole population. That’s why you quite often limit yourself to a sample of observations. Obviously, the bigger the sample the better.
5. Collect Data. Finally, you collect the data according to the data collection plan and the sample size. The more digitalized the process, the faster it goes. However, in many situations, the collection of data may take weeks or even months.

Analyze Stage

The most interesting stage is the third stage where you turn data into conclusions. Thanks to this stage you are able to find out what the thing you want to improve (output variable) depends on. Once you know the cause-and-effect relationship you can find ways to solve the problem and improve the situation.

Let’s have a look at the list of the most popular analyses that you may be interested in running.

1. Chart analysis. The simplest method is simply to look at the chart and based on what you see decide whether two variables (an input variable X and output variable Y) are connected. Using a chart, you can guess the link between the two variables and the direction. For example, if we look at the duration of a service time of a customer in a restaurant (an input variable), most likely it will have a negative impact on customer satisfaction (the output variable). This should be visible on a chart if you plot customer satisfaction against the service time.  Have a look at our guide on data visualization to learn more about chart analysis
2. Comparison of averages. Another simple way to find out the relationship between 2 variables is to calculate the averages for different levels of a specific input variable and compare them. If the difference is significant, then most likely the input variable has an impact on the output variable. For example, if you compare the average salary (output variable) for different levels of education (an input variable), you will discover that the average salary is much higher for people with a master’s or bachelor’s degree than the salary for people with only elementary education.
3. Correlation analysis. Correlation is a measure that helps check how 2 variables are related, and whether they behave in the same way. Correlation also tests the strength of the movement. The higher the absolute value of the correlation measure, the stronger similarity in behavior. It’s worth mentioning that we can check whether things behave in the same way, but we cannot check whether they influence each other and what is the direction of the influence. Let’s look at a movie that will show us how it works:
   
4. Regression model. In regression models, we want to check the relationship between one output variable and many input variables. Moreover, regression models will help us also estimate the impact that a specific input variable has on the output variable. In a sense, a regression model is an extension of the correlation analysis. Regression models also can be used for forecasting.
5. Testing. Testing can be also used as a way of analyzing what actually influences the output variable we try to improve. To use this method in practice, you should divide the sample into two groups. In the first group, you change the input variable you think is influencing the output variable you want to improve. In the second group, you don’t change the input variable. Thanks to such a test you can quite fast establish whether an input variable influences the output variable. Btw this is how medicines are tested. This method enables you to find out whether there is an actual cause-and-effect relationship. However, it takes a lot of time to get good results.

At the end of this stage, you should know what input variables influence the output variable you want to improve. Once you have identified the link you can decide what you should do.

Improve stage.

One of the most difficult stages is the improvement stage. You know what the problem is so now you have to come up with potential solutions, test them and implement them if they generate the expected results.

Let’s look at the main thing you have to do:

1. Generate Potential Improvements. Firstly, you have to come up with potential improvements. Usually, you will try to change the input variables that are impacting the output variables. There are many ways to generate ideas for potential improvements. I recommend especially brainstorming internally, analyzing reviews and feedback from customers and employees, and analyzing solutions used by competitors. For inspiration, you can check our post on coming up with business ideas
2. Test Improvements. From the long list of improvements, you should pick the most promising ones. You identify their potential for example by creating a simple ranking of potential ideas. For those ideas with high potential, you should test by changing the level of input variables and checking the impact on the output variable. I recommend using the funnel approach for managing the test. Below is a short movie on that:
   
3. Analyze results of Tests. Once the tests are done, you’ll get plenty of data that you can analyze and draw conclusions. Based on that, you will decide whether a specific idea makes sense or not. In other words, you should conduct a cost-benefit analysis based on the data generated by tests. Let’s look at an example of a cost-benefit analysis:
   
4. Decide what we implement. After you have conducted and analyzed the tests, you will have sufficient data to decide which potential improvements should be introduced in the whole organization and which ideas are too costly given the impact they generate
5. Implement the improvements. Finally, you implement the most promising ideas that have been tested. In most cases, implementation will be difficult because you will have to change processes and teach people how to implement the change. In many cases, you will also have to make significant financial investments.

Control

Once you have improved the output variable (you have solved the problem), it’s a good idea to make sure that the progress you have made is not wasted. Therefore, you need to implement some control mechanisms.

Let’s look at the things you have to do during this stage:

1. Creating mechanisms preventing errors. If you want to keep the improvements that you have achieved in the previous stages, you have to build mechanisms that will prevent people from reverting to old bad habits and committing errors. Pretty helpful can be things like checklists and Poka Yoke. Let’s check some examples of Poka Yoke:
   
2. Pick the Controlling Method. Apart from that, you have to add some sort of control mechanism. You should especially decide whether you use constant or regular controls.
3. Set the frequency of controlling. In the case of regular controls, it is also a good idea to decide how often you will be checking the input and output variables. This can be done daily, weekly, monthly or annually.
4. Check performance. Once you establish the frequency of controls, you should make sure that the people that are supposed to do them, perform the checks as planned.
5. Modify & Act if necessary. Finally, it is a good idea to define what will happen if you find a discrepancy between how the process should look and how it looks in reality. Without this feedback loop and defined processes, all the checking and controlling don’t actually make any sense. The point is not just to prevent and control, but also to act if there is a problem.

That’s in short. As always, I recommend checking for more examples of our online course Essential Six Sigma for Management Consultants. You will find there +3 hours of content and more than 70 lectures, that will teach you all the things you need to know if you are conducting a consulting project in this field.

Have a look also at our post on Top Management Consulting Techniques, Tools, and Frameworks.