DMAIC, pronounced as one word Dah-May-Ick, is a continuous improvement methodology used within Six Sigma. It is a very empirical based process (or should be) set to standardize work output to a set level of quality. DMAIC relies on consistent data to work and improve processes.

Define

This is where you seek to identify areas for improvement. Perhaps you have realized that your business is growing and you are spending a lot of money on overtime or bringing in temporary workers who may be making more mistakes than usual.

Within your project, it may be testing out improved ways to do things or changes to a software application that may have a murky predicted outcome.

1. What’s your business problem? Lay it out in detail asking specific questions about specific areas.
2. What do you currently know about the problem?
3. Who/what is impacted by the problem?

Measure

Look at the improvement objectively and try to quantify it. Collect measurable data about the current state of any related processes. If you want to improve the lives of your workers, and your costs, by reducing how much time they spend at the office then you measure the current hours worked in overtime.

Within your project, it might be measuring the impact on a web application and the servers running it. It may also mean a meeting mid-iteration to ask what your team or even other stakeholders think about the current state and propose some adaptive changes using quantifiable evidence to help support your position that the changes could be of benefit. (*Figuring out what to measure is often the most difficult part, in it may frequently require replanning as new information is learned)

1. Collect data to establish a baseline of the current state of your process
2. Avoid using generic and non-quantifiable terms – such as “The overtime is high” (High compared to what?)

Analyze

Create process maps to help identify inputs and outputs to the defined problem area. You may have to get in and work with the people who perform the job functions, in fact, you probably should work with them and ask for their input. Collect as much information about how the tasks are done as you possibly can.

1. Investigate and learn the details surrounding the problem area
2. Question the people who perform the work
3. List out possible causes to the defined problem

Improve

Work to find a solution to the most likely cause of the defined problem. The process tends to better if you only make one change and then move on to determining if that helped solve the problem. There are ways to implement multiple changes and investigate the outcome; they, however, come with a steeper learning curve when trying to identify whether you have made a positive change and which changed item was responsible for the impact.

You can use the Deming Cycle (Plan, Do, Check, Act) here, but generally, you want to follow the below (Which is pretty much the Deming Cycle):

1. Plan your improvement – how are you going to achieve the preferred result?
2. Put the plan into action – ideally one change at a time
3. Verify that the change is working as expected
4. If the change is not working as expected, adjust the change or implement a different solution (go back to planning)

Control

In the Control step, you want to make the changes last. If your plan was to reduce overtime by implementing more efficient processes you would have put in place controls to make sure the people carrying out the tasks knew of the changes and could carry out the new directions. You would also continuously monitor the overtime to make sure it didn’t rise above your defined acceptable level.

Within your project, you want to ensure your changes are well implemented and are not causing any negative or unintended impact on the whole process or product. You want to make sure that when released, users are actually using it and not reacting negatively.

If you don’t make the implemented solution easy for others to follow or give them the tools to follow it, it likely won’t be followed.

1. Put controls in place to ensure improvements can be and will be followed
2. Monitor for the reoccurrence of the original problem