DMAIC| Analyse Stage
What is it… The Analyse Stage of DMAIC is about understanding the Root Cause(s) of our problem.Now we get to analyse the data we collected in the previous Measure Stage. The intention is to step beyond having data and to drill-down and analyse what the data is telling us. This stage uses several analytical tools to understand the root cause(s) of our problem. The tools which are deployed will depend on the problem being solved, the type of data you have and what this data is telling you.
How long should the Analyse Stage take?
For the Analyse stage a guide for completion is somewhere between 2-4 weeks depending on the complexities within your project and organisational set-up. Targeted analysis will help to keep this stage focused and reduce unnecessary waste (see what we did there!) in analysis. Some organisations have a lack of data (or inconsistent application) which may unnecessarily add to the time-frame. Conversely others may be data rich but intelligence poor where focus will help to reduce the burden.
What are the key steps…
Step 1 | Process Analysis – Here we delve into more detail on how the process really works in practice. Our Gemba walk insight from the Measure Stage will help to inform this. You can expect to do detailed process maps, value stream mapping and review the 8 wastes. By the end of this step you should have a solid understanding of how the process truly works today.
5 Why’s is a simple technique to understand – just think of a young child consistently asking why to obtain an understanding. In practice, teams (and even seasoned Lean Six-Sigma Practitioners) can struggle with this in deployment. The following are helpful pointers –
- There could be more or less than 5 whys
- Back-up the 5 whys with supporting data – where applicable and resonable to support assumption
- Split a team in two or more to tackle the exercise, you may be surprised at the different routes each team goes down when not constrained by the flow of another session
- Encourage all team members to engage, participate and challenge
- Don’t be afraid to reset a previous question when clearly off track
- Don’t give up or settle for assumed knowns
Step 2 | Root Cause Identification – The trick after Step 1, is understanding which ‘tools’ you are going to utilise to support the analysis. The following can be used as a guide to support this decision making:
- Identifying Possible Root Causes – 5 Why’s; Brainstorming; Cause and Effect Diagrams (otherwise known as Fishbone for its resemblance to a fish or Ishikawa diagram its Japanese term)
- What Could Go Wrong – Utilise the FMEA to assess what could go wrong and where it could go wrong. Negative brainstorming could also be deployed here.
- What Does Go Wrong – Measles Charts and Affinity Diagrams (the later being grouped up issues/feedback)
Failure Mode and Effect Analysis (FMEA) is effectively a risk analysis tool utilised within several stages of the DMAIC improvement lifecycle (Analyse, Improve and Control). Its application or reason for use in each stage may differ but the overriding principle is to prevent an event from ever happening (think safety i.e. keeping trains aligned on a track) or to identify where possible defects could occur.
Step 3| Analysis – This is where we understand what the data is telling us. There are a myriad of statistical techniques and tools that are utilised within this step. A little like the root cause identification step above, it can depend on our data and problem for which of these tools we will deploy. As a guide the below is offered:
- Graphical Analysis: Histograms; Dot Plots; Probability Plots
- Statistical Analysis: Normality Test
- Time Changes:
- Graphical Analysis: Time Series Plot
- Statistical Analysis: SPC (Statistical Process Control – Charts); Run Charts
- Graphical Analysis: Box Plots; 100% Stacked Charts
- Statistical Analysis: Confidence Intervals; Hypothesis Testing
- Graphical Analysis: Scatter Plot; Matrix Plot
- Statistical Analysis: Correlation; Regression Analysis; Multiple Regression; Binary Logic Regression
- Graphical Analysis: Pareto Charts
- Process Adjustment
- Statistical Analysis: Design of Experiments
The above is intended as a high-level overview and categorisation. Each method and tool requires its own explanation and practice to be efficient within.
Step 4| Verify Root Causes – Now we verify the root cause(s) and we understand the cause and effect.
- We utilise Hypothesis Testing to identify the significant factors on our problems
- For identifying relationships we utilise Regression and Correlation analysis
- For experimenting to test main effects and interactions we utilise Design of Experiments (DoE) and analyse those results. From an analytical perspective the calculations are similar to regression the main difference here is testing something out in a controlled manner beyond what we currently do today. This is most common in the technical environments as opposed to the service sector.
This sets up for the next stage to insure we choose and implement initiatives which will solve our problem.
Step 5| Gain Approval to move onto next Stage
At the end of the Analyse Stage the following would have been completed or identified:
|Detailed Process Map||FMEA||RCA Analysis||Graphical Anaslysis||Benchmarking|
|Hypothesis Testing||Statistical Analysis||Design of Experiments||RCA verification|
Key Deliverables of the Analyse Stage
- Data analysis on the root causes
- Hypothesis Testing
- DoE – where applicable
The DMAIC Analyse phase is quite in-depth and you can benefit from a Green Belt or other statistical course to support your learning for implementation.
DMAIC Analyse Stage FAQ’S
What is meant by the 'Process Door'?
What is meant by the 'Data Door'?
The DMAIC Analyse stage is sometimes referenced in two parts. Part 1 refers to the Process (Process Door) and Part 2 to Data (Data Door).
As there are a myriad of tools and techniques to apply the Data Door acts as a signpost and selector for the appropriate tool to match the task. Therefore ensuring that the correct statistical tools are deployed.
What is a Measles Charts?
This technique compliments other metrics like First Time Yield (FTY) by pin pointing the problem areas. This in turn allows the project team to focus in and resolve the root causes.
What are the 3 types of Regression Models?
The most common regression model type is Linear – think Celsius and Fahrenheit, when one rises so does the other.
Quadratic – represents relationships that rise and then fall
Cubic – represents relationships which rise, fall then rise again
The full | Six-Sigma A-Z Glossary
Join Our email list
Join the InvisibleConsultant email list today for support direct to your inbox. Don’t delay Level up Your Capability today.
Though we hope you will find our material helpful, reliance on this material and any related content is at your sole risk. InvisibleConsultant.co.uk assumes no responsibility for any errors or damages arising.