Supercharging Kaizen & Six Sigma With Computer Vision (Smart Shop Floor Part 3)

Table of contents:
Kaizen for Continuous Improvement
Kaizen definition, qualities and the seven-step cycle.
Kaizen Blitz and Six Sigma For Bigger Problems
The difference between Kaizen, Kaizen Blitz and SIX SIGMA.
The Limitations of Kaizen, Kaizen Blitz and Six Sigma
A common challenge: the accuracy and robustness of data sets used for improvements.
Computer Vision: the Equalizer in Reaching a Single Source of Truth
“Smart cameras are unique in being able to watch all physical aspects of a process or operation.”

Kaizen for Continuous Improvement

Kaizen is the Japanese term for change for the better, or for improvement. It was developed in Japan after World War II based on the work of Walter Shewhart and W. Edwards Deming, which Toyota applied to its Lean philosophy of continuous improvement.[1]

As envisioned by Toyota, Kaizen’s philosophy can be combined other process improvement frameworks to drive continuous improvement. The Kaizen way is to take smaller, and commonsense steps forward in a never-ending process of saving money, improving quality, reducing accidents, or increasing customer satisfaction. 

The never-ending philosophy of Lean and Kaizen are fundamentally different from traditional Western project-based approaches which take an ad hoc process of fixing problems. It is also the major difference between Lean and Six Sigma.   

Kaizen is not designed to drive major process improvement projects or other radical changes. This is a major difference between Japanese and Western thinking. In the West, major and abrupt change is often the preferred method to drive big operational improvements. The problem comes in making them stick, avoiding the Hawthorne-type of backsliding discussed earlier. Kaizen and Lean thinking is also built into the Agile philosophy and framework to take on smaller pieces of work that can be achieved in days versus weeks or months.  

Major change efforts also suffer from rapidly changing priorities. In running or supporting over 30 process improvement projects, it was not unusual to be pulled out of one crisis-inspired project to take on the next crisis or grand idea to transform the organization. It came to be known internally as the chasing the shiny ball syndrome. Unfortunately, grand ideas and crisis-inspired projects do little to create a continuous improvement culture.  The gradual Kaizen is often associated with a system of events lasting from two to five days. While Kaizen events are called “Kaikaku” in Japanese for radical change, the events typically take on smaller problems that can be resolved in less than a week, which is also similar to the sprints used in Agile projects. . As a comparison, even lower-level Lean Six Sigma projects are rarely completed in less than a few months and major projects can take several months to complete.

To be successful, a Kaizen event should have these qualities:

  • The event is preceded by a period of detailed preparation.
  • The event is well prepared, with a very specific scope and achievable goal.
  • All resources are fully dedicated and involved with the event; no outside work is permitted.
  • Like a Six Sigma project, the solution is not known but the customer is known.
  • The problem is an immediate and pressing one.
  • The event is followed by an implementation period of a few weeks.
  • The Kaizen ends with the implementation.

There is a good argument for combining Six Sigma with Kaizen. Six Sigma will teach Kaizen leaders and participants the disciplines and tools to define problems and solutions, find the best metrics to measure improvements, and the basics of root cause analysis. Participating in a Kaizen event is a great way for new Six Sigma practitioners to start implementing process improvement before they take on major Lean Six Sigma projects.

Following a seven-step Kaizen cycle can help make process improvement a continuous reality.

  • 1. Get people involved. Engage with people closest to the processes that are impacted by the Kaizen. By treating them with respect, as subject matter experts, they will be open to sharing their candid views and making relevant suggestions. This also helps them buy in to the changes under discussion.
  • 2. Discover problems. Talk to everyone involved with the processes that are part of the problem under discussion. Combining confidential individual interviews and brainstorming group meetings is a good combination to get candid views and achieve a consensus.
  • 3. Create solutions. The same people who help uncover the problems can help create the solutions to those problems. Strive for a consensus as to the optimal solution.
  • 4. Test solutions. Implement the chosen solutions with Kaizen participants and as many stakeholders as possible.
  • 5. Analyze results. On a periodic basis, check the progress and work to keep the participants motivated and engaged. Determine whether the solution is worth adopting.
  • 6. Adopt the solution. If the solution is acceptable, adopt it organization-wide.
  • 7. Plan for the future. Look at the opportunities for related improvements the Kaizen exposed.

Kaizen Blitz and Six Sigma For Bigger Problems

While Kaizen strives for gradual but continuous improvement, there are times when more immediate, intensive,  and impactful change is needed. That is when some organizations turn to the Kaizen Blitz, also known as the Kaizen Event.  The general idea is to throw overwhelming and dedicated resources at a major problem. This necessitates cross-functional teams with members from all stakeholder disciplines, i.e., manufacturing, engineering, supply chain, finance, quality, product management, etc. The key here is that the resources assigned are truly dedicated, i.e., removed from all other duties for the duration of the Kaizen Blliz, typically one week.[2]

A Kaizen Blitz generally follows Six Sigma’s DMAIC methodology (Define, Measure, Analyze, Improve, and Control) but in a more simplified form of Preparation, Event, and Followup. The simpler framework is to fit the shorter time-frame.

For more complex, mission critical problems, Six Sigma is a tried and proven framework for data-driven problem solving in which the solution is not known, but the customer (voice of the customer) is well known.  Six Sigma became popular in the 1990’s first at Motorolla and then at GE with Jack Welch, GE’s CEO its most vocal advocate.

While DMAIC is the most well-known framework for improving an existing process, it has its limitations in designing a new process or fixing a process that is badly broken. In these situations, Design For Six Sigma or DMADV is a good option. DMADV stands for Deign, Measure, Analyze, Design, and Verify.  DMADV uses many of the same tools of DMAIC, but adds such design tools as Pugh Matrix, and Quality Function Deployment, commonly known as House of Quality.

The table below is a summary of the major features of the three problem solving frameworks. These are only general estimates and wide variations are not uncommon. For example, I a have led successful six sigma projects that took a year to complete.

Duration is ContinuousDuration of 3-5 DaysDuration of 3-6 Months
Part-time ParticipationFull-time ParticipationPart-time Participation
Problem Defined Before KaizenProblem Defined Before BlitzProblem Define at the Start
Bias for ActionBias for ActionComplete Detailed Analysis
Team Size FixedTeam Size FixedTeam Size Can Change
Quickly Implement SolutionsQuickly Implement SolutionsImplement Solutions Over Time
Uses Basic DataUses Basic DataIntense Data Analysis

The Limitations of Kaizen, Kaizen Blitz and Six Sigma

All three frameworks for process improvement have proven their effectiveness over decades and not just in manufacturing, but in a wide range of industries. For anyone who has participated in one or more of these frameworks, a major obstacle typically surfaces. It is the accuracy and robustness of the data sets used to drive the improvements. It is very often the case that there are multiple sources of the truth with participants and stakeholders unable to agree on one truth. Unfortunately the result is that senior management often times picks what source of the truth will be used. Ironically my experience over 35 years in manufacturing, supply chain, and process improvement is that the folks closest to the process have the most accurate vision of the truth. These are the floor supervisors and experienced individual contributors whose views are often the softest voice in the room.      

Computer Vision: the Equalizer in Reaching a Single Source of Truth

Smart Manufacturing includes IoT, robotics, additive manufacturing, edge computing, mobile computing, big data analytics, and computer vision. While each of these technologies can contribute to more accurate data sets, computer vision plays a unique and invaluable role in achieving the elusive goal of one source of truth. Today’s computer vision, also known as machine vision, is an application of artificial intelligence that uses deep learning algorithms and neural networks to create smart cameras. Smart cameras are unique in being able to watch all physical aspects of a process or operation. They can not only watch parts going down an assembly line, but they can also watch people interacting with parts, assemblies or materials on the line, and they can watch people interacting with equipment and vehicles.

Computer vision can thus create a complete digital twin of any observable physical process. Now there is one irrefutable version of the truth and not just as snapshot of an operation as was traditionally captured by industrial engineers and continuous improvement teams. The data capture is continuous over extended periods of times and across different operators, different physical locations, and different shifts of the day.

I witnessed the power of computer vision on one of my earliest deployments over five years ago. It was at a luxury mattress manufacturer whose products were priced higher than its competitors resulting in a loss of market share. Their cost of goods sold were based on labor standards using their manufacturing execution system’s (MES) labor reporting. Our computer vision observed that actual labor times on a 24-hour day basis. It consistently showed actual labor rates lower than the standards. Further investigation revealed lax reporting practices by employees. Lowering  the cost of goods sold allowed prices to be reduced to a more competitive level, increasing sales.  Regardless of how disciplined the labor reporting, computer vision provides an irrefutable single source of truth.  

One final benefit of computer vision.  Traditional Lean and Six Sigma suffer from backsliding, the tendency of users to revert back to their old ways of doing things once a project or initiative ends. This resulted in many stakeholders judging their process improvement projects as failures.  The tendency to revert back is known as the Hawthorne effect, named after studies in the 1950s showing folks did better work when being observed, but reverted back to their traditional ways once the observers depart. With computer vision, the observers never depart. They become imbedded in the process. If backsliding does occur, the deviations from the standard are flagged in real time creating alerts.

Regardless of the process improvement approach or combination of approaches you are using in your operations, computer vision will go a long way in creating a single source of the truth and assuring that improvements you implement will be perpetuated.

Read More
4M1E & Computer Vision-1 (Smart Shop Floor Part 4)
Time and Motion Study (Smart Shop Floor Part 2)
What is Industry 4.0? (Smart Shop Floor Part 1)

Anthony Tarantino, PhD
Six Sigma Master Black Belt, CPM (ISM), CPIM (APICS)
Adjunct Faculty Member, Santa Clara University
Author of Wiley’s Smart Manufacturing, The Lean Six Sigma (May 2022)
Senior Advisor to PowerArena


[1] Tarantino, A., Smart Manufacturing, The Lean Six Sigma Way. Hoboken, NJ: Wiley & Sons, Inc.
[2] iSixSigma, Kaizen Blitz Definition (,all%20aspects%20of%20your%20organization