Kaizn Studio
Continuous Improvement Toolkit
Nine guided tools, six reference guides and 21 printable cheat sheets for practitioners at every belt level. No sign-up. Your work saves in this browser as you go, and you can save a project file to keep or share.
New to Lean Six Sigma? Read the overviewWelcome back. Your work in this browser has been restored.
Pick the phase your project is in. Each phase lists its tools and the guides behind them.
Agree the problem, goal and scope.
Establish a reliable baseline.
Find and verify the root causes.
Test and implement countermeasures.
Sustain the gain and hand over to the process owner.
Tools that run the length of the project. Keep the A3 current at every tollgate and capture lessons as you go, not only at close.
The formal authorisation document for your project. Defines scope, goals, team and timeline to get everyone aligned before improvement work begins.
The charter prevents the most common project failure mode: starting without alignment. It answers three questions. What exactly are we fixing? (problem statement). How will we know we've succeeded? (SMART goal and metrics). What is and isn't included? (scope). If the Sponsor won't sign it, the project isn't ready to start.
| Metric (Y) | Baseline | Target | Unit |
|---|
| Phase | Milestone | Target Date | Owner |
|---|
Map each process step with its linked suppliers, inputs, outputs and customers. Work from the Process column outward for best results.
1. Start with Process steps and define 4 to 7 high-level steps using verb + noun format (e.g. "Receive request"). 2. For each step, add the Inputs required and the Outputs it produces. 3. Then identify Suppliers who provide those inputs and Customers who receive the outputs. Each row represents one step and its directly linked elements.
| Suppliers Who provides inputs? |
Inputs What goes in? |
Process Step Verb + noun |
Outputs What is produced? |
Customers Who receives outputs? |
|---|
Iteratively ask "Why?" to move from symptom to root cause. Each problem gets its own collapsible chain and you can extend beyond 5 when needed. The number is a guide, not a rule.
Visual cause and effect analysis. Enter your problem, add causes by category and the diagram updates live. Category labels are editable.
Start with a clear problem statement before the team assembles. During the session, work through one category at a time with a 2-minute silent sticky note round followed by a share-out. Defer judgement during the brainstorm, then use dot voting to identify the 2 or 3 most likely root causes. Investigate each selected cause with 5 Whys. See the Cheat Sheets page for the full Fishbone Facilitation Guide.
Tell the complete improvement story on a single structured page. The A3 format was pioneered by Toyota and forces real clarity. If it doesn't fit on one page, the thinking isn't ready yet.
| Action | Owner | Due Date | Status | Notes |
|---|
Failure Mode and Effects Analysis. Identify what could fail and prioritise risks using RPN = Severity × Occurrence × Detection.
For each process step, identify potential failure modes. For each, describe its effect, then rate Severity (S), Occurrence (O) and Detection (D) on a 1 to 10 scale. The RPN (S × O × D) tells you where to focus. Address high-RPN items first, especially where Severity is 8 or above.
| Process Step | Failure Mode | Effect | Risk | RPN | Cause | Controls | Recommended Action | Owner | New RPN | |||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| S | O | D | ||||||||||
21 print-ready reference cards covering every tool and DMAIC phase. Click any card to preview and print as PDF.
A complete guide to the methodology, its origins, the 5 Lean principles, belt structure, and how DMAIC drives lasting improvement.
Lean originated in the Toyota Production System (TPS) in post-war Japan, pioneered by Taiichi Ohno and Shigeo Shingo. Its core insight: the only work worth doing is work the customer would pay for. Everything else is waste.
Six Sigma was developed at Motorola in the 1980s and popularised by General Electric under Jack Welch in the 1990s. It uses statistical methods to reduce process variation, targeting fewer than 3.4 defects per million opportunities.
Lean Six Sigma combines both: Lean's speed and waste elimination with Six Sigma's rigour and defect reduction. Together they are more powerful than either alone.
A guide to all common process mapping styles, from high-level SIPOC to detailed swimlane maps and value stream maps. Choose the right tool for the right question.
You cannot improve what you cannot see. Process mapping creates a shared understanding of how work actually flows versus how people think it flows. It surfaces waste, handoffs, delays and rework loops, and gives you the baseline from which improvement is measured.
High-level 5-column view. 4–7 process steps. Used in Define phase to scope the project and align stakeholders on boundaries. Start here before any detailed mapping.
Starting a project, aligning a team on scope, communicating to senior stakeholders.
Sequential steps as boxes and diamonds (decisions). Shows logical flow including decision branches, loops, and endpoints. Good for documenting procedures and standard work.
Documenting a procedure, identifying decision points, creating standard work instructions.
Flowchart organised by lane (person, team, or system). Handoffs become immediately visible, which is a key source of waiting and errors. The most powerful map for identifying where delays occur.
Multiple departments involved; when you need to surface handoff delays and ownership gaps.
Shows material and information flow from supplier to customer. Includes cycle time, wait time, inventory levels. Distinguishes value-adding from non-value-adding time. See the VSM page for detail.
Measure phase. Use it for calculating total lead time, identifying bottlenecks and inventory buffers.
Tracks physical movement of a person, product, or information through a space. Drawn on a layout map. The tangled lines ("spaghetti") reveal unnecessary motion and transport waste.
Physical layout is suspected as a waste source, particularly in manufacturing or service environments.
Maps responsibilities against process steps in a grid (rows = roles, columns = steps). Shows RACI-style involvement at each stage. Identifies over/under-involvement of teams.
Clarifying roles across a complex process; preparing standard work; RACI alignment.
VSM is the most powerful tool for seeing and eliminating waste across an entire process, from customer demand back to the first input.
A VSM shows the complete flow of materials and information required to bring a product or service to a customer. Unlike standard process maps, VSM captures time data including cycle time, wait time and total lead time, making waste visible in a way that flowcharts simply cannot.
Time to complete one unit at a single step. Measured directly at the process.
Total time from start to finish, including all waiting. This is what the customer actually experiences.
Available time ÷ Customer demand rate. The heartbeat of the process and how fast you must produce to meet demand.
Value-Adding Time ÷ Total Lead Time × 100%. World class: >25%. Most start at 1–5%.
The Measure phase lives or dies on data quality. Key tools for collecting reliable, representative data that tells the truth about your process.
Before collecting data, answer three questions: What exactly are we measuring? (operational definition), How will we collect it? (sampling strategy), and How will we verify the measurement system is reliable? (Gage R&R). Poor data produces false root causes and wrong solutions.
A pre-designed form for tallying defects, issues, or events by category and time. Reduces transcription error. The simplest and most reliable data collection tool.
Columns = categories (defect types). Rows = time periods or locations. One tick = one occurrence. Include who, when, where fields.
Records what actually happens during direct observation. Captures cycle times, wait times, motion, and deviations from standard work. Never rely on self-reporting alone.
Step name, start time, end time, cycle time, issues observed, operator, date/shift, notes column.
Collect data broken down by the factors most likely to cause variation: shift, machine, operator, supplier, day of week, location. Averaging across strata hides the real source of variation.
"Average defect rate 4%" hides "Machine 3 night shift has 18% defect rate."
You rarely need 100% of data. Choose an approach that gives reliable conclusions without excessive effort.
Structured methods for capturing what customers actually need:
Determines how much variation in your data comes from the measurement system itself.
Same person, same part, same equipment. Does it give the same reading?
Different people measuring the same thing. Do they get the same reading?
Rule of thumb: %GRR <10% = acceptable. 10–30% = marginal. >30% = fix the measurement system before collecting data.
Turning collected data into insights that point to root causes and guide improvement decisions. Covers Pareto, control charts, capability, and supporting tools.
Typically 80% of defects come from 20% of causes. Pareto analysis helps you focus effort where it will have the greatest impact. Always stratify data before building a Pareto chart. Pooling data from different shifts, machines or teams hides the real pattern.
A control chart is a run chart with statistically derived upper and lower control limits (UCL/LCL). Its purpose is to distinguish common cause variation (random noise inherent to the process) from special cause variation (something unusual that warrants investigation). Only special causes should be investigated. Tampering with common cause variation makes things worse.
| Data Type | Subgroup = 1 | Subgroup 2–9 | Subgroup ≥10 |
|---|---|---|---|
| Continuous (variable) | XmR / ImR | XbarR | XbarS |
| Defective (pass/fail) | — | p-chart (variable n) · np-chart (fixed n) | |
| Defects (count) | — | c-chart (fixed area) · u-chart (variable area) | |
Constants 2.66 and 3.27 derived from d2 and D4 for n=2 subgroups.
Any single rule violation warrants investigation. These rules detect patterns statistically unlikely under common cause variation.
1 point outside UCL or LCL. Probability 0.27%. Most obvious special cause signal.
9 consecutive points on the same side of the centreline. Suggests a sustained shift in the process mean.
6 consecutive points continuously increasing or decreasing. Suggests gradual drift such as tool wear, degradation or seasonal effect.
14 points alternating up-down. Suggests two alternating processes (two machines, two operators).
2 of 3 consecutive points beyond 2σ on the same side. Early warning of a shift.
4 of 5 consecutive points beyond 1σ from centreline on the same side.
15 points within 1σ of centreline. Suggests stratification and that data may be from mixed processes.
8 consecutive points on both sides, none within 1σ. Suggests a bimodal distribution from two mixed processes.
Measures how well the process spread fits within specification limits, assuming it is perfectly centred. It ignores any off-centring.
Measures actual performance and accounts for off-centring. When Cpk is lower than Cp the process is not centred on target.
Shows the distribution of continuous data including shape (normal, skewed, bimodal), spread and whether data falls within specification. Always look at shape before using mean/SD. A bimodal histogram suggests two mixed processes.
Plots two variables to explore correlation (X vs Y). Does not prove causation, but a strong correlation is worth investigating. Look for direction (positive/negative), strength, and outliers. Used to validate suspected X → Y relationships.
Shows median, interquartile range, and outliers. Excellent for comparing distributions across groups (shifts, machines, operators). A box plot comparison is almost always more informative than comparing means alone.
Originally 7 wastes from the Toyota Production System, expanded to 8 with Non-utilised Talent (Skills). Every waste consumes resources without adding value the customer would pay for. Learning to see waste is the first step to eliminating it.
In Lean thinking, work falls into three categories. Value-Adding (VA) covers activities the customer would pay for if they understood them. Necessary Non-Value-Adding (NNVA) covers activities required by the current system but not valued by the customer, such as regulatory compliance or system limitations. Pure Waste (NVA) covers activities that can and should be eliminated immediately.
Unnecessary movement of materials, products, information, or people between locations.
Moving files between shared drives · Sending documents between departments for sequential review · Physical movement of goods to interim storage before final destination
Does this item need to travel this far? Can the next step be co-located? Can we remove the handoff entirely?
Excess stock, work-in-progress, or information queuing beyond what is immediately needed.
Emails sitting unread in shared inboxes · Partially completed work batched before the next step · Spare parts stockpiled beyond safety stock levels
Why is this waiting? What is causing the build-up? Can we reduce batch sizes or implement pull?
Unnecessary movement by people. Walking, reaching and searching for tools, information or equipment.
Searching for files across multiple systems · Walking to a printer across the office · Switching between applications to complete a single task
Is everything needed within easy reach? Can layout or system design reduce searching and switching?
Idle time when people, equipment, or information is delayed between process steps.
Waiting for approvals or sign-offs · System downtime · Waiting for a meeting to start · Dependencies on another team's deliverable
What is the person or process waiting for? Is this a capacity issue, a policy issue, or a communication failure?
Producing more than needed, sooner than needed, or faster than the customer requires. Often considered the worst waste as it creates or hides all others.
Reports produced but never read · Preparing more stock than can be sold · Sending notifications to people who don't need them
Does anyone actually use this? Are we producing based on demand or just "in case"?
Doing more work, adding more features, or applying more quality than the customer actually values or requires.
Reformatting reports into multiple templates · Multiple approval layers for low-risk decisions · Polishing a document that will only be read once
Is this step explicitly required by the customer? Would removing it be noticed?
Errors, mistakes, or non-conformances requiring rework, correction, scrapping, or customer complaint handling.
Incorrect invoices requiring reissue · Data entry errors causing downstream rework · Software bugs requiring patches · Mis-shipped orders
What causes this error? Is there a poka-yoke (mistake-proofing) solution? Why wasn't it caught earlier?
Failing to use the knowledge, creativity, skills, and experience of people in your organisation. The "8th waste" added when Lean expanded beyond manufacturing.
Not involving front-line staff in improvement decisions · Qualified staff doing tasks below their capability · Ignoring improvement ideas from operators
Are we listening to the people closest to the work? Are skills being used to their full potential?
Walk the process with fresh eyes. For each step, ask which of the 8 wastes is present. Use a tally sheet to count occurrences by type. A Pareto of waste types will tell you where to focus.
As you build your current-state Value Stream Map, mark each non-value-adding step with its waste type. This connects the 8 Wastes framework directly to your improvement priorities.
Use TIMWOODS in the Analyse phase to classify root causes. Most root causes trace back to one or more waste types and naming them helps teams prioritise countermeasures in the Improve phase.
Decide what to measure, exactly how, how much and by whom before anyone collects a single number. A clear plan is the difference between a baseline the team trusts and one it argues about.
Start with the project Y from your Charter, then add the Xs you suspect drive it. For each, write an operational definition precise enough that two people measuring the same item get the same answer. Decide how much data you need and what to record alongside it (shift, site, product type) so you can stratify later. Then check the measurement system before you trust the numbers.
| Measure | Y or X | Data type | Operational definition | Source and location | Sample size and frequency | Who collects | Stratify by |
|---|
Plot your measure over time against its median. Simple rules tell you whether you are looking at a real change or ordinary variation, with no statistics software needed.
Collect at least 10 points before reading signals, and 20 or more for a solid baseline. Points that sit exactly on the median are ignored by the rules. Three rules are checked for you: a shift (6 or more points in a row on one side of the median), a trend (5 or more points in a row all going up or all going down) and too few or too many runs. Check a fourth by eye: an astronomical point, a value so different that everyone would agree it is unusual.
| # | Label or date | Value |
|---|