Key Takeaways
Limit starts, not effort, and the same bottleneck can deliver sooner.
The boats tell the story. In Leopold’s paper-ship exercise, people first pass unfinished work down a row as quickly as possible. Then they repeat the exercise with a limit of one ship at each step. Both rounds finish about 5.5 ships per minute, because the slowest step sets the system’s output. Yet the final red ship takes 3 minutes 18 seconds in the overloaded round and only 45 seconds in the limited round. The second group feels calmer and has time to check quality.
Use slack to improve flow. A bottleneck cannot be made faster by flooding it with work. With less work in progress, colleagues can help the constrained step, reduce handoffs, or improve quality. The customer gets completed work sooner, even if some people are not busy every minute.
The exercise makes a queueing principle visible: when arrivals exceed a constrained system’s capacity, unfinished work accumulates without raising output. That is why busier can mean slower. The useful management question is not whether everyone looks occupied, but where work is waiting and what would relieve that wait. Slack is often treated as waste because it is visible, while delay is hidden in backlogs. Still, a limit is not automatically wise: a poorly chosen cap can starve a genuinely variable process. Teams need to observe the consequences and adjust, rather than treat any particular WIP number as doctrine.
Evolve the process you have instead of importing a universal Kanban template.
Kanban is a change method. It is not a prescribed workflow, a software-development recipe, or a magic board. The method starts with the work people actually do, makes that work visible, and supports successive changes based on what the system reveals. Its principles are to begin with the current situation, improve through observation, and encourage leadership at every level. Its practices include visualizing work, limiting work in progress, managing flow, clarifying rules, creating feedback, and using suitable methods to improve.
A copied solution can misfit. The book’s Toyota anecdote illustrates why: a factory changed how car parts were supplied when it began building more vehicle models. The earlier arrangement was not universally correct; it suited the earlier circumstances. Kanban likewise asks organizations to test whether changes help their own customers and workflow.
This approach resembles experimental science more than a recipe: state what is happening, change a condition, and inspect the result. It also avoids a common reform failure, in which a new process is announced before its problem is understood. The emphasis on local evidence is especially valuable in knowledge work, where tasks differ and rules quickly become outdated. Yet evolution should not become an excuse for endless tinkering. Without a clear customer outcome and measures of progress, a team can make many changes while learning very little.
Map customer value across services, then coordinate at the level causing the delay.
Start with the customer’s request. A Kanban system should show the activities that turn a request into something useful, not simply the company’s reporting lines. For an online book order, the work might move through obtaining the book, packing it, invoicing, and delivery. Those steps may cross departments, but the customer experiences one service. Organizing the board around that path reveals waiting and handoffs that a team-only view can hide.
Choose the right altitude. Leopold’s Kanban Flight Levels describe three views: operational work within a team, coordination among teams, and strategic portfolio choices across initiatives. They are perspectives, not a maturity ladder. A team-level board cannot fix a portfolio that starts too many projects; a strategy board cannot solve a specific team’s daily bottleneck. Start where the problem and authority to change it actually sit.
The service lens shifts attention from organizational boxes to the customer’s end-to-end experience, a move shared with service design and value-stream analysis. It helps explain why an excellent team can still belong to a poor delivery system: customers receive the combined result of interactions, not isolated departmental performance. The flight-level idea is useful because it prevents mismatched solutions. Its risk is abstraction: executives may build a portfolio view that appears comprehensive but lacks operational feedback. The levels matter only when information and decisions travel between them.
Visualize real value-creating work, not a decorative to-do list.
Show the work’s actual journey. A board with only “To Do,” “Doing,” and “Done” can help someone track personal tasks, but it rarely explains how a service creates value. A useful system maps the activities work passes through and represents value-producing work items as tickets. It can also show the option pool, the queue of work ready to start, blocked items, and completed steps awaiting a pull forward.
Limit the whole system. A WIP limit caps how many items may occupy a defined area, including items that are blocked or finished in one activity but waiting for the next. If waiting work is excluded, the backlog simply grows out of sight. One team in the book measured active work at only 1.4 percent of cycle time; the author reports that roughly five percent is common. Less work in the system can reduce waiting and task switching.
A well-designed board functions as a shared instrument panel: it turns private, scattered knowledge into something a group can inspect and act on. That is more than visualization for its own sake. It changes the conversation from individual busyness to system behavior. The low active-work share is a sharp reminder that effort is not the same as elapsed time. However, a metric such as flow efficiency can be misused as a productivity target. Its strongest role is diagnostic: it prompts investigation into queues, dependencies, and waiting, not pressure to keep every minute occupied.
Treat recurring blockers as clues about the process, not personal failures.
Keep the evidence. A blockade is an unexpected condition that stops committed work from moving forward. Mark the blocked item, record why it stopped and for how long, and retain the markers instead of discarding them when work resumes. Grouping repeated causes is blocker clustering. It helps distinguish a one-off interruption from a system problem that repeatedly delays delivery.
Find causes before remedies. In one web project, missing customer photos and text accumulated 109 days of waiting. Asking why revealed that customers were not told early enough what material was needed. The process was changed to request it during analysis and include it in the activity’s completion rules. When choosing which issue to address, consider whether it constricts flow, how difficult it is to fix, and the economic effect of the delay.
Blocker clustering applies a basic principle of safety engineering: near misses and recurring interruptions are valuable data, not merely inconveniences. The 109-day example is powerful because the remedy is a change in the interface with the customer, not a demand that employees work faster. This also helps move retrospectives away from blame. Records can still distort reality if people label ordinary waiting as exceptional or stop reporting blockers. Clear definitions and a climate where inconvenient facts are safe to surface are prerequisites for useful analysis.
Include customer feedback and specialist waits inside the delivery process.
Completion belongs to the customer. Work that leaves an internal board may still fail to meet the customer’s needs. A validation step makes expected review and feedback part of the workflow instead of treating them as surprise rework. In a production-robotics company, reports of errors were most likely during the first few days after installation and dwindled after five days. The team used that pattern to keep work in validation for six days before calling it complete.
Make expertise less fragile. Knowledge work often depends on specialists who are not immediately available. A limited parking area can keep externally blocked work visible without clogging the active workflow. Pairing a specialist with a generalist on suitable tasks transfers knowledge, although it may take longer at first. That extra time can prevent future work from waiting for one person’s availability.
The validation step reflects a broader product principle: quality is defined in use, not merely at the point of internal handoff. Making feedback normal also avoids the organizational habit of classifying customer learning as an exception. Knowledge transfer addresses a different form of fragility, the single expert whose absence stalls a service. Pairing is not a universal substitute for deep specialization; some work genuinely requires rare expertise. The practical aim is not to eliminate specialists, but to identify which work can safely spread knowledge and which requires deliberate protection of expert capacity.
Forecast from completed work, and communicate odds instead of a single date.
Measure what actually flows. The book recommends tracking work in progress, cycle time, and throughput. Cycle time is elapsed calendar time between agreed start and finish points, not just hours spent working. Throughput is the number of work items completed in a time period. Once a system has enough historical data, a scatterplot can show the range of observed cycle times. A quantile expresses a confidence threshold: for example, an 85th-percentile forecast says that 85 percent of comparable past items finished within that duration.
Simulate a range. For many items, a Monte Carlo simulation repeatedly samples historical results to show possible completion outcomes. In one example, forecasts for a 304-story backlog shifted as new weekly throughput data arrived. The forecast was updated rather than defended as a fixed promise. Such methods are most useful when work in and out of the system is reasonably stable.
Probabilistic forecasts are more honest than a precise date unsupported by evidence. They also create a better customer conversation: the decision-maker can choose whether an 85 percent chance is sufficient or whether the cost of missing a date requires a higher confidence level. This is an instance of calibrated uncertainty, a practice used in fields where decisions must be made without certainty. The caveat is crucial: historical data cannot rescue a process whose workload or completion rules are changing unpredictably. Forecasts need both transparent assumptions and frequent recalibration.
Sequence work by the value lost while waiting, not by loudest priority.
Price delay in customer terms. Cost of Delay estimates the economic loss or missed opportunity caused by postponing a deliverable. The book suggests asking what value the work creates, how that value changes over time, and how long completion may take. A feature that prevents recurring losses may be urgent immediately; a seasonal feature may have value only during a particular window. Sequencing must consider both delay cost and completion time, rather than assigning vague labels such as “high” or “urgent.”
Make risk explicit. Work can face technical risk, such as whether a solution can be built, and market risk, such as whether customers will use it. Estimate both the potential impact and its likelihood, document assumptions, and reduce the largest risks early. Small prototypes or customer checks can reveal whether an attractive idea is technically feasible and wanted before extensive investment.
Cost of Delay translates competing claims into a common decision language, making it harder for the most senior voice or most persistent requester to win by default. The method’s most valuable output may be the discussion: disagreements about a feature’s value reveal different assumptions that can be investigated. But monetary estimates can create false precision, especially for innovation with no established market data. Ranges and explicit uncertainty are better than invented exactness. A portfolio also needs room for uncertain experiments, or comparisons based only on confident short-term returns can systematically crowd out renewal.
Use Kanban’s visibility to change the work, then verify the customer benefit.
The board is not the intervention. At STUTE Logistics, an IT department of 30 people handled about 25,000 tickets a year. Its old practice assigned new tickets immediately, even while previous requests remained unresolved. Workers switched among many open items, and customers faced handoffs. After introducing limits and changing how work entered the system, the open-ticket count fell, throughput rose, and the team could examine its processes more clearly.
Improve what the evidence exposes. STUTE added a single customer contact for issue resolution, automated parts of employee onboarding and departure, and built a tool that showed customers forecasted completion times. The company also used regular meetings and shared measures to coordinate teams. These were context-specific changes, not a template to copy. Kanban provided visibility and feedback; employees and leaders used them to decide what to change.
STUTE’s story grounds the book’s central distinction between adopting a method and improving a service. The gains came from changing ticket intake, reducing handoffs, automating repeatable work, and giving customers clearer information, not from a board alone. The case also suggests that measurement becomes more useful when employees help define it, rather than experiencing it as surveillance. One caution is that an impressive organizational account cannot isolate which intervention caused each result. The transferable lesson is the learning process: expose the workflow, agree on evidence, make a change, and check whether customers actually benefit.
Analysis
Practical Kanban argues that slow delivery is usually a property of the work system, not a shortage of individual effort. Its paper-ship exercise establishes the central counterintuitive claim: restricting work can preserve throughput while sharply reducing the time a customer waits. The rest of the book follows that logic outward, from visible workflow and team coordination to forecasting, portfolio choices, and economic risk.
The book’s strongest contribution is its refusal to equate Kanban with a board or a team-level productivity technique. A board matters because it makes the customer-facing service inspectable. That shift changes the unit of improvement from local activity to the interactions and queues between activities. The Flight Levels vocabulary gives managers a way to locate a problem at operational, coordination, or strategic scale without pretending that every organization should follow the same rollout plan.
Its practical tools reinforce one another. WIP limits support stability; stability makes historical measurements more useful; probabilistic forecasts make uncertainty discussable; and Cost of Delay gives prioritization an economic basis. Blocker clustering and retrospectives create feedback loops through which the system can adapt. The STUTE example shows these ideas as organizational practice rather than abstract doctrine.
The argument is most persuasive when it challenges utilization as the dominant goal. A full workload can conceal queues, rework, and coordination costs, while creating pressure to start yet more work. Still, limiting WIP does not by itself guarantee better outcomes. The right limits depend on variability, demand, skill availability, and the customer’s definition of value. Likewise, economic estimates are aids to judgment, not objective truth, especially for innovative work. The method therefore depends on leadership willing to reveal bad news, revise policies, and protect experiments from being judged by the same certainty standards as routine maintenance. Kanban is less a mechanism for controlling work than a disciplined way to see consequences and make better decisions.
Review Summary
Practical Kanban receives high praise for its comprehensive coverage of Kanban principles and practices. Readers appreciate the mix of theory and practical examples, making it accessible for both beginners and experienced practitioners. The book is lauded for its insights on large-scale Kanban implementation, forecasting, and risk assessment. Many reviewers find it essential reading for modern managers, highlighting its value in improving organizational efficiency. While some suggest it's best for those already familiar with Kanban basics, others praise its clarity and depth, making it a valuable resource for continuous improvement in work systems.
Glossary
Kanban Flight Levels
Organizational improvement perspectivesA communication model for identifying where Kanban can be applied and where a problem should be addressed. Flight Level 1 concerns operational work, Flight Level 2 coordinates work across teams, and Flight Level 3 manages strategic portfolios. These are different perspectives, not stages of organizational maturity, and several can operate at once.
WIP Limit
Cap on active workA rule limiting the number of work items allowed in a defined part of a Kanban system. Limits apply to blocked items and work waiting in completed stages, not only tasks someone is actively handling. By restricting starts, a WIP limit helps reduce overload, task switching, and queues while encouraging completion.
Cost of Delay
Economic cost of waitingAn estimate of the costs or lost economic value that result when a deliverable is delayed or arrives too late for its opportunity. It considers the value at stake and how that value changes over time. Teams use it, alongside completion time and risk, to compare work and create a sensible sequence.
Spice Girls
Work-selection decision groupLeopold’s informal name for the people who decide which options enter a Kanban system and in what order. The group may include business stakeholders, customers, technical experts, or managers. Its members act as gatekeepers for the input queue and use shared evidence, including Cost of Delay, to coordinate demand with available capacity.
Monte Carlo Simulation
Randomized outcome simulationA forecasting technique that runs many trials of a defined model, using random selections from a range of observed or assumed values. The results show a distribution of possible outcomes and their likelihood, rather than a single predicted date. In the book, it is used to forecast delivery for multiple work items from historical throughput.
Quantile
Probability threshold valueA threshold in a distribution below which a stated proportion of observations falls. For example, an 85th-percentile cycle time is a duration within which 85 percent of comparable completed items finished in the measured data. Quantiles let teams communicate forecasts at chosen confidence levels instead of relying on an average.
Little’s Law
Relationship among flow measuresA queueing relationship connecting the average number of items in a system, the rate at which items arrive or depart, and the average time they spend there. Leopold uses it to explain how more work in progress generally corresponds to longer cycle time when flow is stable. It describes system behavior and past averages, not a delivery forecast by itself.
Blocker Clustering
Grouping recurring work stoppagesA method for recording the causes and duration of work blockades, then grouping similar incidents to reveal repeated systemic problems. Teams can examine the largest or most flow-constraining clusters, investigate root causes, and compare possible remedies by effort and economic benefit. The method turns interruptions into evidence for improving the work system.
Option Pool
Uncommitted work choicesA collection of ideas or requests that could be undertaken but have not yet been committed to the work system. Stakeholders review options and select work for a coordinated input queue. Keeping options outside active implementation supports late commitment, allowing priorities to change before substantial effort is invested.
Definition of Done
Completion criteria for an activityA shared description of the result required before a work item can leave an activity on the Kanban board. It may be a checklist of necessary outputs or quality conditions. Making the criteria explicit reduces misunderstandings, supports consistent handoffs, and gives teams a basis for revising process rules when exceptions become routine.
Flow efficiency
Active time within cycle timeThe proportion of a work item’s total cycle time spent in active work. The remaining time is inactive, often because the item waits in a queue, is blocked, or depends on another person or service. The measure helps reveal delay hidden by an emphasis on effort, but should be used diagnostically rather than as an individual performance target.
FAQ
What is "Practical Kanban: From Team Focus to Creating Value" by Klaus Leopold about?
- Practical Kanban guidance: The book provides actionable solutions for common challenges in Kanban implementation, drawing on Klaus Leopold’s extensive real-world experience.
- Shift from team to value: It emphasizes moving beyond optimizing individual teams to improving the entire value creation process across organizations.
- Comprehensive Kanban coverage: Topics include Kanban fundamentals, system improvement, scaling, forecasting, prioritization, risk assessment, and real-world case studies.
Why should I read "Practical Kanban" by Klaus Leopold?
- Real-world expertise: Klaus Leopold shares insights from designing and implementing over 100 Kanban systems, offering proven solutions to typical problems.
- Clarifies misconceptions: The book dispels myths, such as Kanban being only for team optimization, and explains the risks of local versus global optimization.
- Continuous improvement focus: Readers learn how to evolve, scale, and make their Kanban systems more predictable and valuable for customers.
What are the key takeaways from "Practical Kanban" by Klaus Leopold?
- Kanban as evolutionary change: Kanban is presented as a flexible, continuous improvement method, not a rigid framework.
- Value stream optimization: The book stresses optimizing the entire value chain, not just individual teams, to deliver better customer outcomes.
- Data-driven decision-making: Emphasis is placed on using real metrics (WIP, cycle time, throughput) and economic principles (Cost of Delay) for prioritization and forecasting.
What are the core principles and practices of Kanban according to Klaus Leopold?
- Three core principles: Start with what you do now, pursue evolutionary improvement, and encourage leadership at all levels.
- Six essential practices: Visualize work, limit work in progress (WIP), manage flow, make policies explicit, implement fast feedback loops, and drive improvements using models and data.
- Flexible, context-driven approach: Kanban adapts to each organization’s unique context, supporting gradual, sustainable change.
How does Klaus Leopold’s "Practical Kanban" explain the importance of limiting Work In Progress (WIP)?
- Controls system flow: WIP limits prevent overload, reduce multitasking, and help maintain a smooth workflow.
- Shortens cycle times: By reducing WIP, cycle times decrease, enabling faster delivery of completed work.
- Improves predictability and quality: WIP limits create a stable system, making delivery times more reliable and allowing teams to focus on quality.
What is the Kanban Flight Levels model in "Practical Kanban" and how is it used?
- Three levels of focus: Flight Level 1 covers team operations, Level 2 addresses coordination across teams, and Level 3 manages strategic portfolio work.
- Not a maturity model: Flight Levels are communication tools to identify where Kanban can be applied and where issues exist, not a stepwise progression.
- Optimizes value streams: The model helps organizations coordinate work across teams and initiatives to maximize customer value.
How does "Practical Kanban" by Klaus Leopold recommend scaling Kanban in large organizations?
- Scalable by design: Kanban principles and practices apply at any scale, focusing on managing dependencies and coordination rather than just multiplying teams.
- Coordination over local optimization: The book warns against optimizing teams in isolation, which can lead to delays and inefficiencies.
- Use higher-level boards: Flight Level 2 and 3 boards visualize and manage work across multiple teams and strategic initiatives, improving overall flow.
How does "Practical Kanban" by Klaus Leopold address forecasting and estimation in knowledge work?
- Forecasting over estimation: The book advocates using real data and models (like Monte Carlo simulations) for probabilistic forecasting, rather than relying on inaccurate estimates.
- Key metrics: Work in Progress (WIP), cycle time, and throughput are used for forecasting, providing a functional interface to customers.
- Quantile-based predictions: Forecasts use quantiles (e.g., 85th percentile) to set realistic deadlines and Service Level Agreements (SLAs).
What is Cost of Delay (CoD) in "Practical Kanban" and how should it be used for prioritization?
- Economic prioritization: Cost of Delay quantifies the monetary impact of delayed work, combining value and urgency.
- Sequencing work: CoD is divided by estimated duration to calculate a CoD Weight, guiding the order of work to minimize total delay costs.
- Avoids prioritization chaos: Using CoD provides a rational, transparent basis for decision-making, replacing subjective or arbitrary prioritization.
How does Klaus Leopold’s "Practical Kanban" recommend handling risk assessment in Kanban systems?
- Risk as impact and probability: Risk is defined as the potential impact of uncertainty on profit or loss, considering both magnitude and likelihood.
- Types of risk: The book distinguishes between technical risk (feasibility) and market risk (customer acceptance), both of which must be managed.
- Quantification methods: Risks can be estimated through brainstorming, expected value calculations, or multi-dimensional models that include CoD and product life cycle.
What practical examples or case studies does "Practical Kanban" by Klaus Leopold provide?
- STUTE Logistics case study: The IT department at STUTE Logistics used Kanban to improve workflow, transparency, and customer service across five teams.
- Cultural and process transformation: By adopting Kanban principles like WIP limits and visualization, the organization overcame skepticism and improved efficiency.
- Tangible benefits: The case study demonstrates reduced ticket load, improved throughput and cycle time, and increased employee engagement and customer satisfaction.
What are the best quotes from "Practical Kanban" by Klaus Leopold and what do they mean?
- "Be the change you want to see!" — Management must actively participate in Kanban improvements, not just delegate.
- "A bad system will beat a good person every time." — System and workflow improvements are more effective than pushing individuals harder.
- "Stop starting, start finishing." — Limiting WIP helps teams focus on completing tasks, reducing multitasking and delays.
- "Forecasting is a tool for proactive project management." — Data-driven models are essential for predicting outcomes and managing projects effectively.
- "Cost of Delay helps steer away from arbitrariness and towards economically rational considerations." — Decision-making should be based on measurable economic impact, not subjective prioritization.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.