Leaderboard

DOE is widely used in manufacturing sectors. However, the usage of the methodology is challenged in the non manufacturing / services sector due to the following 1. Dealing with people over machines - DOE in manufacturing / production sector deals with raw materials, WIP goods and machines whereas in non manufacturing / service industry process steps are performed by people and it is risky to experiment leading to challenging / changing the way people work as it can be counter productive 2. Multiple scenarios, non controlled environment - In a service sector, the guidelines are broad since the objective is primarily to solve the pain area for the customer. This means that the way in which services are provided can change quickly depending on the requirements. Most of the time decisions have to be made on the spot, getting creative but staying within the guideline. There are more noise factors and variables and most of this cannot be measured in statistical terms. 3. Cannot always involve customer - It is not always possible to involve customer as part of the experiment as it involves time and multiple interactions 4. Involves cost and time- Companies dont want to waste money and time on experiments 4.Culture is more human centric than data centric - The culture in the corporate sector, especially leadership is based on past experiences and intuitiveness and less data oriented Below are some practical suggestions to improve acceptance for DOE in corporate sector 1. Promote computer simulation which is more faster, cost effective and visual 2. Establish link between the result of simulation models and customer satisfaction and profit 3. Train all leaders and management staff on Six Sigma techniques, LEAN and basics of data sciences 4. Inform leaders of how competitors are using DOE to get ahead in the market. This might convince leadership to encourage application of DOE

Design of experiments (DOE) in six sigma helps in preventing process variation and thus improving processes. Quality teams works on DOE to identify major factors of process variation and finding variables that can be improved to reduce the variation and delivery high quality services. The three elements of DOE tool are – · Factors – Input to the process which can be either controllable or uncontrollable variable. · Levels - Measurement of how much a factor has been modified. · Response – Output of the experiment The reasons for the limited usage of DOE in corporate sectors can be – · Lack of awareness and knowledge regarding DOE · Measuring the process performance accurately can be a challenge at times as it has multiple noise factors · Seen as time consuming activities The challenges can be overcome by educating and making them aware about the DOE benefits. The DOE can help in reducing the process variations and have a long-term benefits for the process and organization as a whole. Showcasing successful implementation of such tool can help immensely in promoting and buy-in from the leadership or stakeholders.

Design of Experiment is a great scientific tool which is having used extensively in many areas. Medicine, engineering & biochemistry, physics & computer science contributes to 50% of usage of DOE. In Corporate world, we do not see much usage of DOE. There are several factors which make use of DOE more complicated. Some of them are: 1. Target Low hanging opportunities – We must initiate DOE at the earliest so that we can analyze multiple factors and arrive to a best design. However, since the stakeholder / client is always looking for quick result, we sometime investigate the option which are faster to implement. 2. Force fitting DOE – There are multiple automation scope available in the corporate world. We tend to initiate automation. Later we try to force fit DOE and try justifying our investment. 3. Data quality– Data is one of the vital factors. In most of the processes, we have data captured by the users and which may not be accurate. This may result in incorrect interpretation of hypothesis. 4. Sample data – Sometime the sample data we take may not represent the population and therefore, the analysis can go wrong 5. Lengthy steps – due to shorter timeline, most of the experts avoid DOE as there are multiple steps and analysis involved. 6. Personal Bias – There may be cases where we may see some personal bias of the researchers may creep in. 7. Human Error – Because of lengthy statistical steps, there is always a probability of error 8. Human response – This is one of the most common challenge we see in corporate world. The human responses cannot be measured and therefore performing analysis may be difficult. Most of the cases we tend to ignore the factor with human response which is also not a right approach. Therefore, though this is one of the best tools, we do not use in most of the projects in the corporate world.

Background Performance or outcome variables of process, product depend up on many parameters and their variability. These independent X variables would have impact on output Y variables. For example; Fuel efficiency of an automobile depend on – Quality of Fuel (V1), Speed of Vehicle ( V2), Maintenance of Vehicle ( V3) etc Output quality and yield of batch in chemical / pharmaceutical manufacturing impacted by - Temperature, pressure in reaction, quality, purity of input raw materials etc. During design or process development stage – many experiments are conducted in R&D laboratories or pilot labs to understand impact each variable (Factors) on the outputs at different input range (levels). As long as these experiments are less cost intensive, less time consuming with less resource utilization, multiple tests with combination of factors and levels can be done easily. However when cost , time, resources involved in such experiments are significant, then it is extremely important that few experiments are conducted to see results as per experiment protocol. The criticality of such experiments would go up exponentially wrt to cost / time/ resources/ outcomes, when conducted at scale-up level from 1L in laboratory to 1 KL at plant scale. A statistical method that is used in selecting such impact-full experiments is known as “Design Of Experiments”. Design of experiments – play very important role in pharmaceutical manufacturing process especially in active pharma ingredients (API ), oral solid dosage forms ( OSD) . In API synthesis manufacturing process, Temperature, pressure of key reactions and purity of input play important role in outcome of API with respect to yield, purity. For 2 different ranges of ( levels) of these factors , there would be 8 combination of experiments that can be conducted as per below formula. No. of Experiments = (Levels)Factors Conducting such number of experiments at Plant scale is very difficult as cost, time, resources used are very high. In actual manufacturing process, these type of variables (factors) affecting output variables / quality run into 10s. In such scenarios, a random way of conducting experiments are just not acceptable. Even design of experiments has be done with proper back ground work. Following would help in improving Design of Experiments and their outcome. Understanding interactions with in Xn and between Xn and Yn. Conduct as many gram scale laboratory experiments as possible to understand all interactions of factors ·Gain process knowledge as much as possible with these multiple experiments. In some cases especially for safety evaluations “what-if” experiments are done to understand extreme conditions, deviations. Runways Arrive at Critical Process parameters / factors affecting the output ( may 80-20 rule, pareto analysis, Risk Priority number etc) Define nuisance random variables which can be blocked that can mitigate interference in experiment outcomes ( fraction factorials). Take guidance from subject matter experts within or outside industry to gain best in process knowledge Conclusion All above (but not limited to) , one can arrive at fractional factorial design of experiments and same can be conducted with higher probability of success rate with minimal impact on resources. That is the reason one says “Good planning is Half the job done” and same specially true in case of Design of experiments.

Even though DoE is one of the best methods for process optimization, the effort behind the execution and repetition of the experiments and the costs associated are not favoring its widespread use in the industry. The pre requisite being identifying the parameters associated with the process input and then generating models to identify if it is an optimized one. The complexity and effort will increase depending on the number of parameters and that is the drawback when a process with multi parameter input above 10 or 20 , it will be very complex. This is not only used in the process optimization but also for the engineering product development, simulation and in calibration of systems widely. Some of the alternatives used in the industry for DoE is Active DoE, which is using symbolic regression to select and propose specific parameter combination rather than executing a full factorial model to be executed. usually a 1/3rd or slightly more might be the outcome of applying Active DoE. This also has drawback that the number of parameters that can be considered can go up to 20 and if something goes beyond 20 it gets complicated. This is used widely in many calibration tools in automotive and avionics industry. Alternatively, another approach used is the parameter combination specific to the edge cases can be selected and simulated to evaluate the model and its behavior. In many applications, Mote Carlo simulation is used rather than DoE for the parameter selection and simulation in the industry.

reason:- People not following the six sigma based structured approach for problem analysis. Solution:- Sig sigma consulting agencies should broadcast free webinars to atleast make people aware about the scientific approach to play within variables to come up to the conclusion .

There are number of reason why DOE is not commonly used in the coroporate world or service settings. The most fundamental challenges in applying Design of Experiment in a service set-up are the following: Lack of Awareness, Knowledge and Misunderstanding which tends to discourage experimentation in most service organizations Performance of a corporate or service industry is very difficult to measure accurately Corporate or Service Organization's Process Performance has a great dependency on human behaviour involved in the process Impalpable parts with the delivery of service, outcome may not be as stable as manufacturing. Is often simultaneously contructed and consumed and unapparent dimensions are important indicators of service quality context. This means experimental control of the inputs and measurement of the results needs careful validation and consideration In any service process, it is very common that a clear description and distinction of process is needed for quality improvement and control. Understanding of front/back-end processes and customer requirements are essential for these improvements. Practical Solutions to Overcome These Reasons 1. Start with the basics by defining all processes and operational definitions 2. Standardize and Set Clear Quality Guidelines and Process Documents available to everyone 3. Build Robust Awareness Drive across the organization 4. Influence other Leaders and Empower People to do small DoE in their processes to practice the tool and see potential improvements