“The lean Six Sigma pocket toolbook” summary — Michael L. George

Introduction

You face constant pressure to improve speed, reduce cost, and eliminate quality defects in your processes. Traditional improvement efforts often stall because they focus on only one dimension of performance at a time.

Michael L. George addresses this challenge by combining Lean speed with Six Sigma precision. The book provides a practical roadmap and toolset for eliminating waste and reducing variation across manufacturing and service environments.

A concise guide to blending Lean and Six Sigma tools to eliminate waste, reduce variation, and accelerate business processes.

You will learn how to navigate improvement projects by answering specific operational questions:

How do you select the right tools for each phase of a problem-solving project?
What methods uncover hidden waste and map value streams effectively?
How do you measure variation and isolate root causes using data?
What techniques reduce lead time and sustain long-term process gains?

1. Using DMAIC and Working With Ideas

DMAIC provides a structured five-phase roadmap for problem-solving, while creative tools help teams gather and prioritize ideas.

The DMAIC Methodology

DMAIC is a structured five-phase problem-solving methodology used to guide teams logically from problem definition to sustainable solutions. The acronym stands for Define, Measure, Analyze, Improve, and Control. It encourages creative thinking within strict boundaries, such as maintaining the basic process or product. It is distinct from DMEDI which is used when designing completely new products or processes.

You can implement DMAIC via a traditional part-time project-team approach or an intensive rapid Kaizen approach. The project-team approach uses full-time Black Belts and part-time team members over 1 to 4 months. The Kaizen approach pushes rapid, intense progress through most phases in 1 week or less with dedicated full-time participants pulled off their regular jobs.

Skipping DMAIC steps to jump to obvious solutions is risky unless the problem is truly simple and implementation risks are minimal. You might feel pushed toward quick fixes, but you should only skip steps if data proves the idea is the best possible solution and that downsides are minor. Otherwise, you must follow full DMAIC.

Define the scope, goals, and targets.
Measure current process performance and collect reliable data.
Analyze data to pinpoint root causes and critical inputs.
Improve by piloting and executing full-scale solutions.
Control by handing off the process and maintaining gains.

Idea Generation and Prioritization

Brainstorming is a foundational tool used to quickly generate a wide range of creative ideas without premature judgment or discussion. The process requires quiet individual thought, capturing every idea on notes or flip charts, and forbidding discussion or criticism until all ideas are fully gathered.

Review the problem definition.
Clarify the goal or question and provide any relevant information.
Give everyone a few minutes of silence to think and write down ideas individually.
Gather ideas via round robin or popcorn format without allowing discussion.
Consolidate similar ideas and discuss the complete set.

Affinity diagrams organize chaotic facts, opinions, and customer statements into natural thematic groups to help diagnose complex situations. By writing ideas on cards and having team members silently group them, you can bypass traditional thinking and identify central themes and overarching header labels.

Gather ideas from brainstorming or customer statements.
Write ideas on cards or self-stick notes.
Post notes randomly on a board or table.
Allow people to silently start grouping the cards.
Create a header label for each group.
Perform an optional second round of clustering.
Complete the diagram and discuss the results.

Multivoting is a structured technique used to rapidly prioritize or narrow down an overly large list of options through sequential voting rounds. You can use a rule of thumb such as giving participants 11 votes per person when there are 33 items on the list, allowing you to identify top contenders like 12 top vote-getters from 33 items in a sample first round.

Number every idea or option being considered.
Write each idea on a visible flip chart or whiteboard.
Decide how many votes each person will have based on the total items.
Cast votes using markers or confidential slips of paper.
Count and record the total votes next to each item.
Decide on a course of action by eliminating low-scoring items.

2. Value Stream Mapping and Process Flow

Process Mapping Fundamentals

A SIPOC diagram provides a high-level visual snapshot of process boundaries and essential operational elements to secure team alignment. It maps out Suppliers, Inputs, Process steps, Outputs, and Customers, serving as a vital communication tool and bridging customer needs with Key Process Output Variables (KPOVs). You can build this overview before diving into granular tasks.

Identify process boundaries and key activities keeping to roughly 6 activities at most.
Identify key outputs and customers of those outputs.
Identify inputs and suppliers.
Identify critical-to-quality requirements for inputs, steps, and outputs.

Value Classification and Takt Time

Mapping process steps and classifying value help uncover hidden operational waste and establish customer-driven pacing.

Processes can be categorized into customer value-added, business non-value-added, and non-value-added waste to isolate hidden costs. Customer value-added tasks enhance form or feature that clients pay for; business non-value-added tasks are required for compliance or financial risk management; non-value-added tasks represent pure waste such as excessive handling, rework, and waiting.

Classify each process step as value-added, business non-value-added, or non-value-added.
Add up time spent in each category using a Time Value Map or Value-add Chart.
Optimize value-add tasks, minimize business non-value-add tasks, and eliminate non-value-add activities.

Takt time calculates the mandatory operational pace required to meet customer demand by dividing available work time by customer orders. It sets the rhythm for the production or service line so that no step operates slower or faster than client demand, highlighting bottlenecks known as time traps. You take 480 min total available work time derived from 8 available hours and divide by 60 customer orders to be filled used in the example calculation to reach an 8 min resulting takt time.

Follow value-add analysis instructions to classify process work.
Collect time spent on each work type per process step.
Visually display results in a bar chart.
Calculate takt time by dividing available work time by customer demand.
Act on results by eliminating NVA activities, improving VA activities, and reducing BNVA activities.

3. Voice of the Customer and Data Collection

Segmenting customer needs and establishing rigorous data collection plans ensure accurate measurement of process performance.

Voice of the Customer Analysis

Customer segmentation allows organizations to target subgroups that generate the highest value, as all customers do not create equal value. High-volume versus low-volume customers, or West Coast versus Midwest versus Southeast customers serve as common examples.

To use customer segmentation, you follow these steps:

1. Identify the output (product or service) being studied
2. Brainstorm to identify the customers of that output
3. Identify the segmentation characteristics that you think may influence how a customer or group responds to your company and its products or services
4. Develop profiles of the segments you will seek out for your projects
5. Include representatives from each segment in whatever customer contact you initiate (interviews, surveys, focus groups, etc.)
6. Document the results

Kano analysis helps categorize customer requirements into dissatisfiers, satisfiers, and delighters to properly prioritize product and service features. By asking customers how they would feel if a need was addressed or not addressed, features are classified into basic requirements, performance requirements, and excitement requirements.

Data Collection and Measurement Systems

Data types are split into continuous variables and discrete or attribute data, with continuous data generally preferred due to more powerful statistical analysis tools. Lead time, cost, price, and call duration are continuous, whereas on-time delivery pass/fail and department names are discrete.

Creating a structured data collection plan ensures that gathered data remains both useful and statistically valid. You follow a rule of thumb limiting the data collection process duration to 15 minutes or less a day to ensure completion.

Gage R and R evaluates the repeatability and reproducibility of a measurement system to ensure data variations stem from actual differences rather than measurement error. The automotive industry action group short form uses 2 operators measuring 5 items 2 times for 20 total measurements. The long form standard uses 3 operators measuring 10 items 3 times for 90 measurements total. Between 30% to 50% of measurement systems are not capable of accurately or precisely measuring the desired metric. Less than 10 percent Study Var is good, 10 to 30 percent may be acceptable, and more than 30 percent is unacceptable.

The Kappa calculation is used to analyze measurement system agreement when dealing with non-quantitative attribute or discrete data. You consider an attribute measurement system inadequate if the result falls below 0.7, while a score of 0.9 or above is considered excellent.

4. Descriptive Statistics and Variation Analysis

Statistics are divided into descriptive methods that summarize a group of observations and inferential methods that draw conclusions about a population from a sample. Descriptive statistics represent characteristics of a large group of observations using sample statistics denoted by Latin letters, whereas population parameters are denoted by Greek letters. Inferential statistics use a small set of numbers to make broader inferences.

Statistical measures and control charts distinguish between common and special cause variation to evaluate process stability.

Variance possesses a major advantage over standard deviation because it is additive across process steps. While standard deviations cannot be added together to find total process spread, variances can be summed. The total process variance is determined by adding variances for all process steps and then taking the square root to find the standard deviation.

Descriptive Statistics

The distribution of sample means approaches a normal distribution regardless of the shape of the parent population if the sample size is large enough. Under the Central Limit Theorem, calculating averages on subsets of data allows valid statistical inferences and normality even when raw data is skewed. Sample sizes of 30 are sufficient to make sample means approximately normal in the worst-case scenario.

When you collect data, you should apply specific guidelines:

If the population is normal, Xbar will always be normal for any sample size
If the population is at least symmetric, sample sizes of 5 to 20 should be OK
Use sample sizes of 30 to make Xbar approximately normal no matter how far the population is from being normal
Use a standard subgroup size to calculate Xbars
Ensure sets of data used to calculate Xbar are rational subgroups

Control Charts and Process Capability

Processes experience two types of variation: common cause arising from random stable factors, and special cause arising from temporary or assignable factors. Common cause variation is inherent to the system and requires fundamental process redesign to reduce. Special cause variation produces identifiable patterns and instability, requiring analysts to track down specific root causes.

Continuous data control charts evaluate both process accuracy through averages and process precision through ranges or standard deviations. Practitioners typically use ImR charts for individual data points with low output rates, Xbar and R charts for subgroup sizes from 3 to 9, and Xbar and S charts for subgroup sizes of 10 or more.

When you set up these charts, you follow a clear method:

Determine sampling plan
Collect samples at specified intervals of time or production
Calculate the mean and range or standard deviation for each subgroup
Plot the data on separate charts for averages and variation
Calculate control limits for the range chart after 20 or more sets of measurements
Calculate control limits for the Xbar chart only if the range chart is in control

Attribute control charts track count or proportion data instead of continuous physical measurements. Binomial data uses p-charts or np-charts for pass/fail classifications, while Poisson data uses c-charts or u-charts for tracking defect counts where non-occurrences cannot be counted.

For attribute data, you follow a similar sequence:

Determine an appropriate sampling plan
Collect sample data at specified time intervals
Calculate the relevant metric
Calculate the appropriate centerline
Plot the data
Calculate control limits after 20 or more measurements

Specific pattern tests applied to control charts detect shifts, trends, or stratification based on standard deviation zones. Zones divide the control chart into Zone C at plus or minus 1 standard deviation, Zone B between 1 and 2, and Zone A between 2 and 3. Rules like points beyond Zone A or runs of consecutive points on one side identify special causes.

To identify special causes, you look for specific behaviors:

Check for 1 point beyond Zone A to detect shifts or single aberrations
Look for 9 points in a row on one side of the average in Zone C or beyond
Identify 6 points in a row steadily increasing or decreasing for trends
Check for 14 points in a row alternating up and down for systematic effects

Process capability indices compare actual process variation against customer specification limits to evaluate capability. Cp and Pp measure total variation relative to tolerance, while Cpk and Ppk account for centering and separate upper and lower specification limits. Short-term metrics use subgroup variation while long-term metrics use overall data variation.

Before evaluating capability, you must complete preparation steps:

Calibrate measuring system
Perform Measuring System Analysis
Ensure the process is in statistical control
Confirm customer requirements to establish specification limits
Check short-system capability first and implement fixes if unacceptable

5. Identifying Causes and Confirming Effects

Structured root cause analysis tools combined with advanced hypothesis testing isolate critical process inputs and drivers.

Identifying Potential Causes

You can use Pareto charts to focus improvement efforts on the vital few sources of trouble. The horizontal axis represents categories like defects or causes, while the vertical axis represents counts, percentages, or impacts such as cost or delays. A cumulative percentage line can also be added. If no Pareto effect is present, you should convert counts to cost or time impacts, or reconsider stratification factors.

Collect data on different types or categories of problems.
Tabulate the scores, determining total and category-specific numbers or impacts, combining minor ones into an 'other' category if needed.
Sort the problems by frequency or level of impact.
Draw a vertical axis divided into increments equal to the total number observed.
Draw bars for each category from largest to smallest, putting 'other' last.
Add the cumulative percentage line by converting raw counts to percentages and plotting points above each bar.
Interpret the results for clear Pareto effect or no Pareto effect.

The 5 Whys is a quick brainstorming technique that pushes your team past superficial answers down to actionable root causes. Starting from a known problem or cause from a Pareto chart or fishbone diagram, you repeatedly ask why an outcome occurs until reaching a potential root cause you can act on. There is no strict requirement to stop at exactly five iterations.

Select any cause from a cause-and-effect diagram or a tall Pareto bar and ensure common understanding.
Ask why this outcome occurs.
Select one reason and ask why that occurs.
Continue until reaching a potential root cause that the team can act on.

Cause-and-effect diagrams, or fishbone diagrams, provide structured formats that encourage broad thinking to uncover potential root causes. The specific problem is placed at the head of the skeleton, and potential causes are brainstormed and categorized using the 6 Ms. You review the completed diagram to eliminate inapplicable causes and mark critical ones for follow-up investigation before generating action plans.

Name the problem or effect of interest as specifically as possible and write it at the head of the fishbone skeleton.
Decide major categories for causes using the 6 Ms: manpower, machines, materials, methods, measurements, and Mother Nature.
Brainstorm more detailed causes by working through categories or doing open brainstorming, writing suggestions on sticky notes.
Review the diagram for completeness, eliminating inapplicable causes and brainstorming sparse categories.
Discuss the final diagram to identify critical causes for follow-up investigation and mark them.
Develop plans for confirming that potential causes are actual causes before generating action plans.

A C&E Matrix connects process inputs to prioritized customer requirements to identify key variables that drive process improvement. Customer requirements and their priorities are listed across the top, while process steps and inputs are listed down the side. Inputs are rated for correlation strength against each output, then cross-multiplied with priority scores and summed to highlight critical variables. You typically list 5 or fewer outputs across the top, use a 1 to 10 priority score scale, apply a 1, 3, and 9 correlation rating scale, and ensure 50% to 60% of cells are blank.

Identify key customer requirements from process maps or Voice of the Customer studies and list them across the top.
Assign priority scores from 1 to 10 to each output based on customer importance.
Identify process steps and key inputs and list them down the side of the matrix.
Rate each input against each output using blank for no correlation, 1 for remote, 3 for moderate, and 9 for strong correlation.
Cross-multiply correlation scores with priority scores and add across for each input.
Create a Pareto chart focusing on variables with the highest total scores.

Confirming Causal Effects

Hypothesis testing uses statistical confidence intervals to determine whether observed data differences reflect true population differences or random variation. It evaluates a null hypothesis against an alternative hypothesis. Type I errors involve rejecting a true null hypothesis, while Type II errors involve failing to reject a false null hypothesis. The p-value indicates the exact probability of making a Type I error. You might use a common alpha level of 0.05 leading to a 95% confidence level, or an alternative alpha level of 0.10 leading to a 90% confidence level.

Define the null hypothesis representing the baseline assumption that values or means are equal.
Define the alternative hypothesis representing reality if the null is rejected.
Establish an alpha risk level to determine the required confidence level.
Collect sample data and calculate test statistics or review software outputs.
Compare the p-value to the chosen alpha level to decide whether to reject the null hypothesis.

One-sample and two-sample t-tests determine whether process averages or sample means differ significantly from target values or from each other. The t-test calculates a t-statistic and compares it against a t-distribution, which is wider and flatter than a normal distribution for small sample sizes. Rejection of the null hypothesis is signaled when confidence intervals exclude zero or when p-values fall below the alpha threshold.

An automobile manufacturer targeting camshaft lengths of 599.5 mm found that Supplier 2 had a mean of 600.23 mm with a p-value of 0.00, indicating the supplier was not on target.

Select the appropriate t-test type based on comparing a sample to a target value or comparing two samples.
Formulate the null and alternative hypotheses for the mean or difference between means.
Gather normally distributed sample data and execute the test using statistical software.
Examine the resulting p-value and confidence intervals to evaluate statistical significance.

6. Reducing Lead Time and Complexity

Operational levers like pull systems, setup reduction, and complexity mapping eliminate lead time and non-value-add costs.

Distinguishing between time traps and capacity constraints is critical because time traps insert delays while capacity constraints limit total output. Time traps create long lead times and excess inventory due to poor policies, long setups, or downtime, but they might not restrict overall output. Capacity constraints cannot produce at the rate required by customer demand and must be fixed first before addressing time traps.

Property appraisers can evaluate 120 properties per day, but customer demand is 130 applications per day, making appraisers a capacity constraint.

Gather needed data including aggregate customer demand, net operating time available, and net resource capacity by process step
Calculate takt rate as the number of units to process divided by net operating time available, and takt time as its inverse
Analyze figures to identify the step with net resource capacity closest to the takt rate as the time trap, and check if it meets customer demand to determine if it is a capacity constraint

Lead Time Reduction Tools

A Generic Pull System places a cap on work-in-process to make lead times predictable and drive down inventory. By calculating a target process cycle efficiency based on world-class benchmarks, a WIP cap is established.

Replenishment pull systems automatically replace consumed items to eliminate shortages and stockouts for repetitive, low-variability items. This system calculates work demand rates, replenishment lead times, and optimal safety stock levels assuming normally distributed demand, often executed simply through a two-bin system.

Determine work demand rate based on average daily or weekly usage and historical data
Determine replenishment lead time and order interval
Determine optimal safety stock level using demand variability, desired service level, lead time, and the lead time reduction factor
Set bin quantity in a two-bin system based on max lead time or order interval multiplied by demand plus half the safety stock

Complexity Mapping and Analysis

Complexity value stream mapping and analysis diagnose how product and service variety drives non-value-add time and costs. Because low-volume offerings have a disproportionate impact on efficiency, tools like product family grids, complexity maps, and the complexity equation help isolate whether poor process cycle efficiency stems from process issues or product complexity.

An electric generator company compared bills-of-material for two frames and discovered they shared all but four parts, revealing hidden commonality that could be exploited to simplify design.

List subprocesses across the top of a matrix and products or services down the side to build a product/service family grid
Sort products into 3 to 10 families based on similarity of process flow
Select representative products from each family and gather cost, processing time, setup time, queue time, and demand data
Calculate PCE destruction for each offering to determine the impact of specific process steps or product lines

7. Selecting and Testing Solutions

You must exhaust all possible solution ideas because the best ones often emerge toward the end of intense creativity efforts. After generating ideas, you must narrow the list using affinity diagrams or multivoting, synthesize them by combining best characteristics, and use evaluation criteria.

Decision matrices, risk assessments, and pilot testing ensure that implemented solutions meet project goals and sustain gains.

Before analysis, showstoppers and options that conflict with strategy or go beyond charter scope must be removed. Remaining ideas are evaluated for organization fit and project goal impact, categorized as standalone or coupled solutions, and scored against criteria including process impact, time impact, cost-benefit, and other impacts.

Solution Selection Methods

Generate potential solutions and exhaust all team ideas.
Narrow the list and synthesize by combining best characteristics.
Select the best solution through an iterative process.
Remove show stoppers from your list of alternative solutions.
Consider organization fit for each remaining idea.
Determine project goal impact for each remaining idea.
Narrow the list using multivoting, CDAM, Pairwise Ranking, or Force Field Analysis.
Enter criteria and top solution alternatives into a solution selection matrix.
Score alternatives on each criteria.
Use FMEA or any risk-evaluation technique as appropriate.
Identify the solution ideas you want to compare and label them with numbers or letters.
Draw a matrix with as many columns and rows as there are items and mark off diagonal boxes.
Review or develop criteria for comparison.
Compare each item to every other item one at a time until you fill out the upper or lower half of the matrix.
Record the results by putting the letter or number of the preferred option in the box.
Tally the number of times each option won out or tally total votes to interpret results.
Develop potential solutions and capture them in drawings, word descriptions, or other characterizations.
Identify criteria relevant to customer requirements, design parameters, and project goals.
Weight the criteria.
Select one alternative as the baseline and enter B in all cells under it.
Prepare an evaluation matrix listing alternative solutions across the top and criteria in the first column.
Score each alternative solution against the baseline using plus, minus, or same ratings.
Sum the pluses, negatives, and sames, and compute the weighted total for each concept.
Focus first on alternatives with the most pluses and fewest minuses, looking for ways to attack weaknesses and incorporate strengths.
Identify the strongest possible solutions and repeat analysis to confirm first-run results.
Perform Controlled Convergence analysis to enhance positives and eliminate negatives.

Risk Assessment and Testing

List your highest priority risk.
List the controls that would help mitigate or eliminate the risk.
Determine if those controls exist.
Discuss the appropriateness and sufficiency of each control.
Identify actions needed to address any control deficiencies.
Review the product, service, or process starting with steps contributing the most value.
Brainstorm and sort possible failure modes.
List one or more potential effects for each failure mode.
Assign ratings for severity and occurrence on a scale of 1 to 10.
List current monitoring and controls, then assign a detection rating on a scale of 1 to 10.
Calculate a risk priority number (RPN) by multiplying severity, occurrence, and detection ratings.
Use RPNs to select high-priority failure modes and prioritize actions.
Plan to reduce or eliminate risk by identifying causes, recommended actions, and responsible persons.
Carry out the plans and document actions taken.
Recompute RPN.
Plan what needs to be piloted, where, who will be involved, when, and how long.
Review design to ensure elements are complete, integrated, and vulnerable points are identified.
Implement design changes identified in the review and move to pilot testing.
Carefully observe all activities, manage expectations, check for ripple effects, and establish reliable baseline performance data.
Evaluate the test and verify results using applicable statistical tools, and celebrate success.

10 Key Ideas

DMAIC provides a structured five-phase roadmap guiding teams logically from problem definition to sustainable solutions while avoiding the risks of skipping steps.
Voice of the Customer analysis segments customer groups and categorizes requirements using Kano analysis to properly prioritize product and service features.
Structured data collection plans and Gage R&R studies ensure that gathered data remains statistically valid and that measurement systems are capable of precise evaluation.
Distinguishing between common and special cause variation using descriptive statistics and control charts allows organizations to evaluate stability and process capability.
Root cause analysis tools like Pareto charts, the 5 Whys, and fishbone diagrams help teams isolate critical process inputs and drivers.
Hypothesis testing utilizes statistical confidence intervals and p-values to determine whether observed data differences reflect true population differences rather than random variation.
Replenishment and pull systems place caps on work-in-process to make lead times predictable and eliminate shortages for repetitive items.
Complexity mapping and analysis diagnose how product and service variety drives non-value-add time and costs to reveal hidden commonality.
Failure Mode and Effects Analysis calculates risk priority numbers by multiplying severity, occurrence, and detection ratings to prioritize risk reduction actions.
Pilot testing requires careful observation of activities, management of expectations, and verification of results using applicable statistical tools before full-scale implementation.

Book summary “The lean Six Sigma pocket toolbook”: summary

Summary text