Lean Six Sigma Green Belt – Six Sigma Improve Phase Part 4

8. Design of Experiments

Doe. We are going to discuss about the benefits of factors and levels available. Doe or Design of experiments is a structure and control method used to vary the inputs. Here, inputs are called as factors in order to find the best combination of inputs today, the desired output. What does that mean? Let us understand using this image which is provided here. What do you think this equipment here does? It does the curdling process to generate curd and to buy products.

Here is the curd meow me looking curd, right? Think about any curd manufacturing organizations, companies, big giants such as Britannia or AMU, all right? This generation of curd process is an ongoing process. How do we ensure that the curd is generated? How do you ensure that? Traditionally, what do we do in the lukewarm milk? We are going to put a pinch of curd which will have microbial compensation. And this microbial growth happens when there is lukewarm curd, right? Or lukewarm milk.

Basically, now, curling does not happen when the temperature is cold. Curling does not happen when the temperature of the milk is hot. Curling happens when the temperature is optimum look warm, right? What is this optimum temperature at which curdling happens? Or milk will curdle at a faster peaks, right? That is what your Amol or Britannia would look at, right? Tell me, what is the optimum temperature at which milk will curdle? Microbial growth is maximum. Right. So how do you do that? You do a lot of experiments, right?

Basically, you try to put the temperature at maybe so much coal, so much, maybe you reach lukewarm, maybe you increase it further, further, further. And then, based on all these experiments, you’ll say that milk will curdle a lot at this optimum temperature. That’s the interesting part here. When analyzing a process, experiments are often used to evaluate which process inputs have a significant impact on the process output and what the target levels of these inputs should be to achieve a desired level, desired output, desired result. Experiments can be designed in many different ways to collect this information.

Design of experiments can be used at the point of greatest leverage to reduce costs by speeding up the process, reducing the late and certain changes, and reducing time and human process complexity. We’ll understand these things in more detail in the subsequent slides. All right? So desired experiments are also powerful tools to achieve cost savings by minimizing the process radiation, reducing the rework and the need for inspection. Let me show you that.

Here we have a list of inputs which get into a processor system to generate the output, right? What is the best combination of A, B and C inputs to generate the most optimum output? Here are a few of the learnings from design of experiment. Best optimum settings of input levels how the inputs affect the output, that is, which inputs, if any, have the most influence on the output. You can probably use sensitivity analysis to do that.

What are the parameters used in design of experiments? ABC. These are called as inputs. Y is an output here and these must be expressed in numbers, right? The inputs A, B and C should be controllable. And measurable, we must have a reasonable level of confidence that the input parameters which you have chosen have a significant influence in affecting your output. Hence, you need to have the process knowledge that’s important. Right, here we go.

Let us understand design of experiment. Using this example, I want to come up with a pastry or a piece of cake, right? And I want it to taste well. I want the color to be luring, promising and consistently. I want to manufacture, generate or cook this cake with the same taste, with the same color. These are the examples of characteristics which I want to achieve. And this is the output which I want to achieve. I want a key response that it’s called us. In order to come up with this key, I need microwave oven.

That is the first factor or the first input. I need sugar, I need flour and I need eggs for each factor. These four are called as factors. Sugar, flour, eggs. These are all the factors. And each factor has a level or a setting, right? Owen, temperature. Say I want to maintain the temperature at 40 degrees in 50 degrees, if that is so. This factor, which is called as Owen, has two levels. Level one is 40 and level two is 50. Sugar.

 Say I want to test it with £1 of sugar, £2 of sugar and £3 or ounces of sugar. However you want to term that. Flour, £1. 02 pound, £3. Number of eggs ten x, 20 x, 30 x and say I also want to test it with 40 x. For this x, which is a factor or your input variable x, you have four levels, one, two, three and four. For flower, you have three levels. This is a factor. Basically, this is a factor x, which has three levels, one, two and three. You have sugar, which is a factor which again has three levels, one, two and three. Owen has two levels, 40 and 50. These are the two levels which are available. So what are factors? These are the process parameters or excess inputs levels.

These are the possible values for the factors. For example, just assume that temperature has two possible options. Two possible values 20 degrees and 30 degrees. Assume you have only two levels that you have chosen, right? Then this shall be called two levels for the factor temporary. And how many outcomes are expected out of that number of levels raised to the power of factors? If I have two levels for each of these four factors, then the number of outcomes will be 16. What does that mean? I have only two levels for x. Assume I have only two levels for floral, two levels for sugar.

Just assume for now, right, then the number of experiments that you might perform is two raised to the power four, which is 16. So how do you do that? With 40 degrees temperature, £1 of sugar, £1 of flour and ten x, I perform one experiment. With 40 degrees, £2 of sugar, £1 of flour, ten x, I perform another experiment. With 50 degrees. With two kflower, one kg sugar, 20 eggs, I perform another experiment. In that way, I look into all the possible combinations. I perform approximately 16 experiments. And all of all these 16 experiments, which experiment? Which combination of inputs is giving me the optimum output, the best output, right. By using least the cost that is achieved or identified using design of experiments. That is very important. All right.

9. Piloting

Now is the time that we pilot this best solution which we have identified. Right? So how do you do that? What is a pilot? Basically, a pilot is a test of the proposed solution. And this type of test has the following properties. In what are those properties? It’s performed on a small scale. Scale. You do not want to fail on a big way, right? Do you know that you have the optimum solution? It’s always better for you to first pilot that. If not, the failure is going to be extremely costly.

Pilot is used to evaluate both the solution and the implementation of the solution. Just because I have a solution does not mean it’s going to work wonders for me, right? There might be problems with the implementation of the solution as well. The purpose is to make the full scale implementation more effective. Hence, you first try it on a small scale and if it is working well, if it does not have any problems, that is when you can go with the full scale implementation, right?

 It gives you the data about expected results and it exposes the issues in the implementation plan. If at all there are any problems while you implement your solution, this is when you come to know if you do a pilot. Instead of going the big bang way, the pilot should test both if the process meets our design specifications and the customer expectations in a total check in process.

The design specification, for example, is a target two minute check in time, right? This corresponds with the customer’s need for quick checking. This is just an example. There are a lot of other examples. Pilot, right? You probably want to implement at a single location. If your organization is located in multiple locations, you just want to implement at a single location, right?

You want the product or service markups or models, small prototype, a working model, right? You probably provide a limited time offers to the customers and see whether the solution that you have brought in in sales and marketing is working well or not. You try to do this early evaluation by end users. Windows Ten, right? Which is going to come now, is free for few days. Free upgrades for a few days, right?

So early evaluation by end user, that is what people do. Beta versions of software are an example of this implementation. For a select customer group, instead of maybe providing a sale offer to everyone, I’m going to release it only to a section of customers. And based on that, probably I’ll select more and if it is a success, probably I’ll implement that for all the customers. Walk throughs Ryan’s or breast rehearsals. Implementation for one work area. If there are multiple departments, probably I’ll select to implement the solution for only a single department.

Probably I’ll release the product only in the test market. These are all the various examples of piloting, basically. All right. Now. Why should I pilot? What’s the reason? Why do I do that? Because I want to improve the solution. I want to understand what are the risks involved in implementing the solution. I want to validate whether the solution is giving me the expected results or not. I want to ensure that there is smooth implementation. When I go with full scale implementation, I want the people to buy in the change.

If I’m going to push it the big way, probably there’ll be a lot of resistance to the change. If I want less resistance, then I’ll do a pilot. And this will also ensure that the buying happens. People are stakeholders impacted by this change, are on the line of this particular solution which you intend to perform, basically, right? Identify the previously unknown performance problems. Probably there was no problem and you have come up with the solution. Once I implement, there might be a lot of additional problems which might come up, right? So should be wary of all that. This is a typical pilot roadmap. First you create a pilot plan. You ensure that there is a strong leadership support for you. You communicate the plans to key stakeholders. You train the pilot group and you implement the pilot. Then after implementation of the pilot, you collect and analyze the feedback. Then you try to diagnose the gap and revise the solutions and also you implement the solution.

So these are the eight steps which you traditionally follow for a pilot. Basically, when should a pilot you pilot when you need to confirm the expected results and practicality of the solution. Is it practically possible? If yes, then what are the results? If you want to reduce the risk of failure, if you’re going to implement a big way full scale implementation, what if it fails? If you are implementing on a small scale, if it’s going to fail, the cost of failure is going to be less in comparison to full scale implementation failure, right? The scope of the change is large. If you implement the big way and in order to revert, it’s going to be extremely difficult. Implementing the change will be costly.

Changes would have far reaching unforeseen consequences. Right, so that’s about that. And then we have pre pilot and post pilot. Before piloting a solution, what should I do? Postpiloting the solution, what do I do? Right, let’s look into pre pilot. Ensure that all elements of the design are complete. Ensure that all the design elements are well integrated and interfaces between different parts of the design are tight enough. Ensure that prepilot review is needed to identify the possible failure points and areas of vulnerability to be tested in the pilot. Pre pilot is needed to review the predictability. To review predicted design capability.

Prepilot review is needed to review the pilot and the implementation plans. Extremely important. Post that you implement the pilot, you’re going to pilot your solution and then you also look into the post pilot. The pilot plan should be designed to include steps to analyze the reasons root causes of performance gaps may need to be analyzed.

Analyze the root Causes even a small change in a business process can affect many other processes during a pilot, the team may need to double check for this ripple effect. You need to check for the impact on other processes. Make sure the new design has not somehow caused problems for your internal supplier and customer process. Ensure people working in such support areas as planning, customer service operations and quality control need to know that you’re doing things differently so they can adjust their work when necessary to communicate with all the parties which might be impacted with your change.

FMEA can be used to anticipate the potential problems, right, due to the impact on other processes and people. And we can often take countermeasures to reduce or eliminate the risks. That’s the important part. Finally, we complete the pilot review. Once the pilot is complete, you will have to review that. Once all the pilot data has been collected and the results verified, the team can determine the next steps towards solution implementation only. And only after an objective and comprehensive assessment of the pilot can responsible next step decisions be made. Some questions few of the questions a team should ask upon the completion of a pilot to help guide them towards identification of the proper next steps are follows right? Did pilot yield you the anticipated results? Was a plan for conducting the pilot effective? What improvement can we make to the solution?

Can the solution be implemented as is? Right? Should it be? Can the solution remain in place at the pilot location? What lessons learned and best practices can we apply during solution implementation? Right? And did the solution achieve the required design goals or not? So these are few steps which you need to check after the pilot is implemented. Now let us look into the key outputs of the improve fees. We have come to the end of the Improf East. The key output is the selected solution, right? However, we have also generated a list of potential solutions and then we have selected the solution. We have optimized the solution settings.

We have looked into FMEA to identify what are the risks associated with the selected solution. We have checked the impact of the selected solution and the benefits and the new pilot is required improve Phase in a nutshell, we have generated alternate solutions to control the inputs using these techniques primarily brainstorming, round robin fashion benchmarking, creative thinking or probing process mapping. These are a few tools and techniques which we have used to generate alternative solutions.

Then we have evaluated the alternative solutions to identify the optimum solution. We have used Pew Matrix which may be used to compare alternative solutions. We have used FME failure mode effects analysis to analyze the risks of the solution. We have built consensus on alternative solutions. Right. Using Delphi, we try to build consensus.

Using multivoting and NGT nominal group technique, we have tried to build a consensus. Finally, we have selected the solution which we want to pilot. And if the pilot is successful, we do a full scale implementation. Pause the full scale implementation. Performance of your output is reevaluated to check whether significant improvement has taken place or not. All right, my dear friends, we have come to an end of improved phrase. I hope you have enjoyed the session. We are looking forward for you to go through the control phase. And there are a lot of interesting examples which are awaiting you in the control fees. See you there. Bye for now. Thank you so much for listening to this session.