A busy board can hide a delivery system that is slowing down. This episode introduces Kanban through the work itself: what has started, what is waiting, what is blocked, and what has actually finished. Learn how explicit workflow policies and work-in-progress limits support pull…
CURRENT EDITION: PMP 2026. Start with Season 2 for the updated certification preparation course. Previous seasons are archived and may not reflect current exam requirements.
The Certified PMP® Audio Course is your complete, on-the-go study companion for mastering the Project Management Professional (PMP®) certification exam. Designed for busy professionals, this Audio Course transforms every aspect of the Project Management Institute’s (PMI) exam content into structured, easy-to-follow lessons that blend clarity with real-world application. Each episode guides you through the principles of project leadership, process integration, stakeholder management, risk, quality, and agile methodologies—ensuring you understand not only the “what” but also the “why” behind each concept. Whether you’re commuting, exercising, or reviewing between meetings, this series helps you study smarter, stay motivated, and move steadily toward exam success.
The PMP® certification is the global gold standard for project managers, validating your ability to lead teams, deliver results, and manage complex projects under real-world constraints. It measures your proficiency across predictive, agile, and hybrid approaches—emphasizing leadership, strategic thinking, and adaptability. Earning the PMP® demonstrates mastery of the three key domains: People, Process, and Business Environment, and signals to employers that you can manage risk, align projects with organizational goals, and deliver value from start to finish.
Developed by BareMetalCyber.com, the Certified PMP® Audio Course provides structured instruction, practical insights, and exam-focused guidance to help you build the confidence, mindset, and discipline needed to pass the PMP® exam and excel as a certified project management professional.
Kanban makes the flow of potential value visible, controls started but unfinished work, and supports active improvement of the workflow. It can be used in project work as well as operational streams such as maintenance or support. A board alone does not guarantee shorter waits or more useful outcomes. The team must define its workflow, manage the items within it, and use evidence to improve delivery.
Welcome to Certified, the P M P prepcast. This is episode sixty one of season one. Today we are exploring kanban, limits on work in progress, and measures of delivery flow.
Visualizing work and documenting your policies are foundational because they convert opinions into observable facts that teams can act on. A Kanban board’s columns should reflect your actual workflow steps; cards represent individual work items with clear entry and exit criteria. Explicit policies, rules for when to pull an item in, how to treat blocked cards, and classes of service like expedite or fixed-date, prevent silent assumptions from derailing flow. Make blocked items and aging visible with color codes or swimlanes and standardize definitions to reduce “mystery work.” When everyone follows the same policies and sees the same board, coordination happens with less conversation and fewer hidden surprises.
Kanban reduces mystery work by making definitions explicit: what counts as “in progress,” what qualifies as “ready,” and what completion looks like for each column. Standardized definitions prevent people from starting work that isn’t actually actionable and reduce the blame game when something stalls. The board becomes both a work tracker and a policy document: when a card is pulled, the team knows exactly what entry criteria were met and what exit tests must pass. That shared language accelerates onboarding, reduces rework, and raises the quality of handoffs because expectations are agreed and visible before the work begins.
Use consistent workflow boundaries for flow metrics. Work in progress is the number of items that have started but have not finished, including started work that is waiting or blocked. Cycle time is elapsed time from the defined start to finish of an item. Throughput is completed items per unit of time. Work item age measures how long a started, unfinished item has been in the workflow so far. Lead time may use a broader request-to-delivery boundary, including the wait before work starts. Name those boundaries before comparing figures from different teams.
Interpret flow measures together rather than treating any one as a promise. Work in progress reveals unfinished load, while work item age helps identify open items that may need attention now. Completed-item cycle time describes elapsed delivery experience, and throughput counts finished items per period. Any broader request-to-delivery measure should make its boundaries explicit. Changes in work mix, capacity, or completion policy can change how the figures should be interpreted.
Use a board and card format that the actual audience can understand. Make blockers identifiable with text or symbols as well as any color so color is not the only signal. Include the information needed to act, such as the blocking condition, accountable contact, and next review point. A proposed unblock date should be distinguished from a confirmed commitment.
Make aging visible by adding timestamps to key transitions, when a card entered a column or when it was unblocked, so you can detect items that linger longer than expected. A simple aging cue helps teams focus on the right problems: older items often point to hidden dependencies or incomplete information. Use a simple notation rather than complex formulas; teams act faster when the signal is obvious and the next step is clear, who will unblock and by when, rather than when they must interpret a dense report.
Standardize definitions to reduce "mystery work" where team members believe different things by habit. Define what “ready” means for the first column and what “done” looks like for the last column, including any verification or documentation needed. Making these rules explicit reduces the impulse to start partial or poorly specified work and lowers the time spent clarifying as the item flows through the board. Consistent definitions also make metrics meaningful because you measure the same thing repeatedly rather than a shifting set of hidden activities.
An item does not leave work in progress just because nobody is actively touching it today. A blocked approval, a waiting test, and an unfinished review still count when they are inside the defined start-to-finish workflow. Hiding them outside the count makes the flow look healthier without making delivery faster. Keep both the current item count and the ages of unfinished items visible, then use completed-item cycle time and throughput to assess the broader pattern.
Round examples help explain flow without pretending the arithmetic removes uncertainty. Compare averages over a suitable period, with consistent item definitions and workflow boundaries. A single snapshot of ten open items and two completions last week does not establish the next item's completion time. First ask whether the system and the observations support the relationship being used.
Throughput is easiest to use for simple forecasts: count completed items over recent intervals and report a short-range range rather than a single date. Cycle time distributions show how variable completion is; if cycle time varies widely, forecast ranges must be wider. But avoid heavy statistics, use medians and simple ranges that are easy to explain: “median cycle time is five days; eighty percent complete within ten days.” These summaries describe the observed sample. They are not a guaranteed deadline for the next item, especially when the work mix or workflow changes.
Manage flow proactively by limiting W I P and pulling only when capacity exists rather than pushing work into the system. W I P limits force prioritization and expose bottlenecks: when a column hits its limit, the team must finish existing work or explicitly re-evaluate priorities rather than starting new items. Identify bottlenecks by watching aging and queue length, long queues or many aging cards in a column point to a constrained step that needs attention. Smooth arrival by staggering work intake or implementing explicit intake windows to avoid sudden load spikes.
Reserve capacity for expedite work so critical items do not derail predictable flow; explicitly define what qualifies as expedite and how it interacts with W I P limits so “everything is urgent” does not become the norm. Use visual signals to trigger action, colored cards, blockers, or swimlane flags, to make it clear when intervention is needed. These visual triggers keep the team focused on flow health rather than on heroic, last-minute firefighting that obscures systemic problems.
When bottlenecks occur, apply simple tactics: swarm the bottleneck with help, split large items into smaller pieces, or improve the bottleneck step’s capacity through automation or clarified policies. Don’t assume the fix is more people; often clarifying entry criteria or reducing batch sizes eliminates the queue. Address chronic bottlenecks by adjusting team structure or by changing the policy that created the bottleneck in the first place, sometimes the solution is organizational, not technical.
Classes of service let you treat different kinds of work according to agreed rules so the board reflects reality rather than wishful thinking. Common classes include standard work for normal items, fixed-date for deliveries that must meet a calendar constraint, expedite for urgent production work, and intangible work for technical debt or maintenance tasks. Set explicit policies and limits for each class so stakeholders know what to expect: for example, only two expedite items can be active at once to prevent regular work from grinding to a halt.
Avoid “everything expedite” by enforcing class limits and reviewing outcomes: if too many items are expedited, examine the root cause, are handoffs broken, are priorities unclear, or is the intake policy too loose? Review class outcomes periodically and adjust policies based on observed data: if fixed-date requests routinely miss dates, renegotiate expectations or add capacity for that class. Communicate service expectations clearly to stakeholders so they understand the trade-offs: faster handling of one class reduces capacity for others.
Classes of service are optional policies for differentiating work, not a source of unlimited capacity. An expedited item may use reserved capacity or displace other work, depending on the explicitly agreed arrangement. Make that effect visible to stakeholders rather than assuming capacity has been reserved. Revisit the policy when repeated exceptions undermine ordinary commitments.
Little’s Law relates average work in progress, average throughput, and average elapsed time within a consistently defined system. Under suitable long-run stable conditions, average work in progress equals average throughput multiplied by average cycle time. For example, two completed items per week times an average cycle time of three weeks corresponds to an average of six items in progress. Use the same boundaries, item population, and time units throughout. A single day’s W I P count divided by one unusual week’s completions does not necessarily represent that long-run relationship.
Apply that intuition as a simple diagnostic: if you observe that the number of items in progress is growing while the count of items finished per week remains unchanged, expect single-item completion times to lengthen. In plain terms, more things in flight means each thing waits longer. The practical rule follows: when throughput stalls and W I P rises, trim W I P or increase throughput capability at the constrained step. State the assumptions, including the time window and how items were counted,so stakeholders understand the back-of-envelope nature of the reasoning.
Use Little’s Law to size interventions, not to promise exact dates: you can say, “Given our recent throughput of three items per week and target cycle time of two weeks, an average W I P near six is consistent with that target under comparable stable conditions,” and if W I P drifts higher, take visible action. Emphasize ranges: aim for order-of-magnitude guidance and quick corrective steps rather than exact forecasts that crumble under variability. That makes decisions faster and less defensively statistical.
Variability is the fuel of queues: when work sizes, priorities, or arrival patterns vary a lot, steps queue up and bottlenecks emerge. In practice, protect the identified bottleneck by reducing variability feeding into it, split large items, avoid batching, and make readiness criteria explicit so only work that can proceed enters the constrained step. This protects flow health by smoothing what arrives at the bottleneck and giving the constrained resource consistent, actionable work rather than unpredictable spikes that cause long queues.
Bottlenecks are easiest to spot with simple visuals: watch which column’s queue keeps growing, which cards age the longest, and where throughput temporarily stalls. Act quickly with practical tactics: swarm the constraint with extra focus, remove unnecessary work feeding it, or change the policy that created the queue, sometimes the fix is clarifying acceptance criteria rather than more hands. Treat these actions as experiments: try a change, measure whether queue length shrinks, and either adopt or revert based on simple evidence.
Address chronic bottlenecks with a mix of policy and capability changes. If a column is constantly overloaded because of required approvals, consider whether authorized simplification, automation, delegation, or better preparation could reduce waiting without removing a required control so fewer items hit the bottleneck unprepared. If capability is the issue, invest in training, pairing, or dedicated capacity for that class of work. The right solution matches whether the root cause is process, policy, or skill, not a reflex to add bodies.
A throughput forecast needs history relevant to the expected work and clearly defined completion rules. Comparable item sizes can help, while a known mix of different sizes may require segmentation or another model. A range such as three to five items describes a forecast interval only under stated assumptions. Calling it medium confidence does not establish a probability. Explain the method and limitations instead of presenting an unsupported confidence claim.
Use simple visual tools, throughput bars, median cycle time, and a range forecast, to communicate delivery expectations. Don’t overpromise; present a short rationale describing the window and the assumption that item sizes remain similar. Narrow a range only when the evidence and forecast method support doing so. Always qualify forecasts with the key caveat: if arrival rates or item sizes change, the forecast must be revisited. This keeps stakeholders aligned with the model’s limitations and avoids blaming the team for normal variability.
Refine forecasts incrementally: when new data arrives, update the short-range window and communicate the change with a one-line explanation. Avoid complex probability cones or heavy statistical language; instead show a simple chart of recent throughput with a proposed range and let stakeholders ask clarifying questions. This plain-English approach preserves credibility while giving leaders the information they need to plan with contingency.
A fictional museum digitization team provides a useful flow example because its work includes preparation, scanning, description, and permission review. The team defines a work item as one collection packet ready for publication, not each individual photograph. Work starts when a packet is selected for active preparation and finishes when the packet has passed the agreed publication checks. Requests waiting for selection remain visible in a separate intake queue. This definition matters because counting photographs in one week and packets in another would make throughput comparisons meaningless.
On Monday morning, six packets have started but not finished. During the week, four more packets start and three finish. Assuming no cancellations or other boundary changes, seven remain in progress at the end of the week. Six plus four minus three equals seven. The team has been busy, and it completed useful work, but unfinished work increased by one. The calculation does not identify the cause of delay. It simply prevents a cheerful completion report from concealing that the active system is accumulating work faster than it is finishing it.
Two of those seven packets are waiting for permission responses from rights holders. Removing them from the active count because nobody is scanning them today would change the apparent result without changing the service delivered. Their waiting time remains part of their elapsed cycle time under the team's chosen boundaries. The board can identify them as blocked and distinguish the reason, but the overall work-in-progress count still includes them. A permission request that has not even started the defined workflow belongs in intake instead. The distinction follows the policy, not whichever location makes the dashboard look better.
The oldest unfinished packet has been active for twelve calendar days. That is its current age, not its completed cycle time, because it has not finished. Suppose a recent comparable sample of twenty finished packets contains seventeen that completed within ten calendar days. Seventeen divided by twenty is eighty-five percent. The team can use that observation as an initial basis for a service level expectation while acknowledging the small sample and checking whether the current work is comparable. It is not proof that this twelve-day packet will finish tomorrow, nor that every future packet has an independently established eighty-five percent probability of meeting that period.
The age tells the team to investigate while the item is still open. They discover that the rights holder received an incomplete request and has been waiting for a description of the intended publication. The actionable issue is missing information, not a lazy scanner or a mysterious statistical outlier. An owner supplies the information and confirms receipt. The team also adds a preparation check to prevent the same omission. Recording that correction preserves the real wait and the learning; resetting the start date would erase both.
Now the team considers lowering its active-work limit from seven to five. A rule change does not instantly make two open packets disappear. The transition might temporarily stop new starts while people finish or unblock existing packets. If an item is legitimately canceled, record that disposition separately from a successful completion. Do not count the cancellation as delivered publication value merely to increase throughput. The group needs a workable transition policy and stakeholder agreement about which requests wait, especially if some have genuine exhibition deadlines.
An exhibition opens in three weeks, and a curator labels a newly arrived packet urgent. The team first clarifies what the exhibition needs and whether a different already-cleared image could satisfy it. If an expedited packet is justified, the policy makes its effect on ordinary work visible. Expedite does not create an extra scanner, an extra reviewer, or permission to ignore rights checks. It changes how limited capacity is allocated. An exception record should explain the decision and the displaced work, so recurring urgency can be examined rather than silently becoming the normal intake route.
For a longer-run illustration, suppose comparable records show an average of six active packets and an average throughput of three packets per week under reasonably stable conditions. Those averages correspond to an average cycle time of two weeks. The units cancel: packets divided by packets per week gives weeks. If the team targets an average of four active packets while sustaining the same throughput, the relationship corresponds to about one and one-third weeks. This is an arithmetic consistency check under stated conditions, not a causal guarantee that imposing a limit of four will produce that result.
The practical constraint may prevent the target from working. For example, if the only permission reviewer visits once a week, reducing the number of active packets does not eliminate that calendar wait. The team could improve request completeness, arrange more frequent authorized review, or redesign which work is prepared before selection. Each option has a different consequence. Moving the start boundary later can reduce the measured cycle time while leaving the customer's request-to-publication wait unchanged. That may be a legitimate new measurement, but it must be disclosed and should not be advertised as faster service without end-to-end evidence.
The museum also checks quality and usefulness. A new scanning shortcut might raise packet throughput while producing images that curators reject for unreadable labels. A completion rule that includes the necessary quality checks prevents such output from being counted as finished prematurely. Conversely, insisting on unnecessary formatting for every packet could delay an exhibition without increasing its value. The team examines the requirement and decision owner before changing the rule. Improving flow means delivering suitable results more effectively, not merely moving cards into a final column sooner.
At the next review, the group compares ages, completed cycle times, throughput, and work in progress over comparable periods. It also looks at publication defects and the exhibition's actual needs. A short-lived improvement after clearing one unusually old packet is not enough to establish a stable new service expectation. The team records what changed, keeps observing, and updates the forecast as evidence develops. This example shows why the four flow measures belong together: completed work tells you what happened, while open-work age and current load help you act before the next delay becomes history.
Scenario: cycle time has risen over the past month while W I P has increased, and stakeholders urgently demand firm delivery dates for several high-priority items. The board shows several aging cards piling up in the integration column and an expanding queue at the test stage. Stakeholders expect exact dates rather than ranges, and pressure is building to start more work to hit commitments. Pause and think: what does the board show about the system’s health, and which lever will reduce uncertainty fastest?
Verify the workflow data and discuss the aging integration and test items with the people doing the work. Check whether the queue reflects missing information, a dependency, inadequate capability, or an intake policy that starts too much work. An agreed limit on new starts and targeted help may stabilize the flow, but setting a smaller number does not instantly remove work already in progress. Explain the current forecast and its uncertainty now, then define an improvement experiment with an owner and review point. Do not promise that cycle time will improve within one or two weeks without evidence for that timing.
Follow that immediate step with short experiments: split the largest items into smaller slices, assign explicit owners to unblock aged cards, and reserve a small, documented expedite slot for genuinely urgent work. After a week of tighter W I P and targeted unblock actions, revisit cycle time metrics and report back with updated ranges. This sequence shows both decisive action and evidence-based recovery rather than waving deadlines or adding more work that would worsen the situation.
Pitfalls start with letting W I P grow unchecked, which fuels multitasking and unpredictable delays that erode trust; hidden queues, work waiting in people’s inboxes or undocumented handoffs, create blind spots that make flow metrics meaningless. Another trap is treating every request as an expedite, which collapses class differentiation and kills predictability. Finally, vanity charts that count activity but not flow mislead stakeholders into believing throughput is healthy when cycle time tells a different story.
A compact playbook is to define the workflow, visualize the work, control work in progress, and act on evidence. Track work in progress, throughput, work item age, and cycle time using the agreed boundaries. Explain what changed, why, and how its effect will be assessed. Improvements may be small experiments or larger justified changes, provided the team preserves necessary controls and evaluates the result.
Keep decisions plain and people-first: when you reduce W I P you ask people to focus, not to do less important work; when you change policies, explain how the change protects predictability and reduces frantic firefighting. Use the board as the conversation starter rather than a report to hide behind, and make the next action obvious for anyone who reads it. That practical discipline turns flow metrics into useful levers you can apply every week to improve delivery, reduce stress, and keep commitments credible.
For more certification study resources, books, and podcasts, visit Bare Metal Cyber dot com. Thank you for listening to Certified, the P M P prepcast.