Kanban Board – Visual Work Control

This topic is part of the SG Systems Global manufacturing, quality, and operations glossary.

Updated October 2025 • Flow & Pull Systems • Scheduling, WIP Limits, and Execution Discipline

A Kanban Board is a visual control that makes the flow of work explicit—what’s queued, what’s in progress, where it’s blocked, and what’s done—so the team can control work-in-process (WIP), expose bottlenecks, and deliver predictably. In manufacturing, Kanban boards extend physical card systems that trigger replenishment to include end-to-end visibility across planning, Job Release, execution, and verification. In regulated industries, the board must do more than look tidy; it must reflect authorized reality—only released materials, current instructions, and trained personnel—backed by records in eBMR and governed by Document Control and Audit Trails.

Practically, the board aligns release and capacity: upstream queues buffer variability; WIP limits throttle starts; explicit “Definition of Done” at each column prevents pass-through of hidden rework. Teams see the constraint saturate first; they swarm upstream problems rather than pushing more orders. When paired with JIT and Heijunka, Kanban boards become the day-by-day heartbeat: start only what the system can finish, finish what you start, and prove it with data.

“A Kanban board is not a wallpaper for tasks; it’s a throttle. If everything is ‘in progress’, nothing is under control.”

1) What It Is (Unbiased Overview)

A Kanban board is a column-based visualization of process states—e.g., Ready, Setup, Run, Inspection, Closeout. Work items (cards) flow left-to-right as they meet the entry/exit criteria for each column. Unlike traditional schedules that push work based on due dates alone, Kanban enforces pull: a downstream column signals capacity and “pulls” the next item from upstream when the WIP limit allows. This shifts the control point from theoretical plans to actual system capability, reducing multitasking and aging WIP.

Cards represent batches, jobs, or sub-lots. Each contains identifiers, priority (or class of service), materials status, revision level for methods/labels, and blocker tags. Boards can be physical (racks, magnets, card bins) or digital (MES dashboards). Either way, the rule is the same: the board is the system of record for status, and the data behind it must be defensible during audits and recalls. If the board shows “Run”, the eBMR had better show contemporaneous entries and the materials had better be released.

2) Core Principles & Signals

Visualize the workflow. Map real states, not wishful ones. Include queues explicitly: Ready (kitted), Ready (docs effective), Ready (capacity window). If a column routinely hides rework, split it. If a step is optional, don’t pretend it’s standard.

Limit WIP. Hard caps per column or per resource family prevent overloading the constraint. WIP limits should be sized from observed cycle time and variability, not committee optimism. Violate limits only under a declared expedite policy with visible trade-offs.

Manage flow, not utilization. Maximizing local utilization creates global queues. Flow metrics—lead time, cycle time, throughput, percent blocked—beat utilization percentages when predicting on-time delivery and release velocity.

Make policies explicit. Each column has an entry checklist and an exit checklist (e.g., kit scanned FEFO/FIFO, equipment change-parts verified, effective method version linked, training completed). Ambiguity is the seedbed of delay.

Implement feedback loops. Daily standups at the board; weekly cadence reviews; monthly flow retro to adjust WIP limits, classes of service, and sequencing rules (e.g., allergen family grouping).

3) Card Design, Priority, and Classes of Service

Every card should carry the identity and risks of the work. Minimum data: SKU/recipe, lot/batch, revision IDs for method/label, kit status, expiry/hold-time sensitivity, equipment family, and due date. Add Class of Service to encode how the system treats the card when contention occurs:

• Standard: Default pull per FIFO in the column.
• Fixed-Date: Must complete by a committed date (launch, promotion, customer promise). Sequencing and buffers reflect the stake.
• Expedite: One-in, one-out privilege with strict WIP cost (something else pauses). Limit the number of concurrent expedites.
• Intangible: Technical debt, validation, or maintenance that protects future flow; schedule capacity explicitly.

Use visual tokens for blockers (quality hold, component shortage, equipment down) and risk (high-potency, allergen changeover, sterile boundary). Cards with aging clocks make stale WIP obvious—if it’s been in Setup for two days, that’s a signal, not an accident.

4) WIP Limits, Queues, Sequencing & Release

WIP limits should bind where variability accumulates—usually before the constraint. Tie limits to Job Release: do not release into Run if Inspection or Closeout is already saturated. To avoid thrash, define ready queues distinct from planning queues: planning holds all candidates; ready queues hold only those that have passed kitting, document effectiveness, training checks, and capacity windows. This separation prevents the “visual landfill” of cards that cannot actually start.

Sequencing rules tame changeover cost and risk: group by allergen/color family, by equipment change-parts, or by thermal profiles. In pharma/food, apply FEFO/FIFO in ready queues to avoid expiry and stale kits. If a card’s upstream components near expiry, promote it within the class of service—explicitly, with a recorded rationale.

5) Metrics That Matter

Lead Time: Request to release (or request to done). Confirms customer experience and inventory exposure. Trend by class of service.

Cycle Time: Start to done per step. Reveals where flow actually slows; feeds SPC charts for stability.

Throughput: Completions per period. Watch sustained rate, not peaks. Tie to CPV and first-pass yield.

WIP: Count per column vs. limit; time-in-state distributions.

Blocker Time: Percent of time items are blocked and top blocker categories (materials, equipment, quality review, labeling).

Flow Efficiency: (Value-add time)/(Lead time). Brutal but honest—most systems hover in the single digits until WIP is cut.

6) Common Failure Modes (and How to Avoid Them)

Everything is “in progress”. No WIP limits, or limits ignored. Countermeasure: set limits where pain is felt; stop starting, start finishing; make policy to break limits explicit and rare.

Board drifts from reality. Teams update weekly, not live. Countermeasure: integrate with MES so state changes auto-move cards; require move-with-scan at gates (kit scan, setup complete, IPC passed) to update the board.

Pretty columns, sloppy policies. No entry/exit criteria; hidden rework. Countermeasure: define checklists per column; add sub-columns (e.g., Run: first-article/steady-state); force rework to re-enter earlier states.

Expedite addiction. Everything urgent, nothing reliable. Countermeasure: hard cap on concurrent expedites; measure expedite cost in delayed standard work; escalate root causes via CAPA.

Visual landfill. Planning and ready items mixed; stale cards linger. Countermeasure: purge non-ready items to planning view; add aging WIP alarms; conduct regular “done or delete” triage.

Regulatory mismatch. Card says “Run” but materials are on hold or instructions obsolete. Countermeasure: bind card movement to Hold/Release status and Document Control checks; block moves without evidence.

7) Digital Kanban in MES, ERP & WMS

Digital boards eliminate transcription and give instant context: click a card to see eBMR step status, device readings, IPC results, label version, and training completions. Connect to ERP for demand and ATP/CTP signals; to WMS for Directed Picking and kit readiness; to QMS for deviation blocks and Change Control effective dates. A serious board won’t let you drag a card into Run if the equipment is down for maintenance, the component is quarantined, or the label master revision changed overnight.

Automation patterns: “auto-promote” from Ready to Setup when the cell frees up; “auto-block” when a linked lot flips to Hold; “auto-split” when a batch divides into sub-lots for parallel equipment; “auto-merge” for bulk-to-pack conversions. All automation must leave a trace in the Audit Trail.

8) Regulated Context: Evidence and Defensibility

In pharma, med-device, and food, the board is part of the story you tell an inspector: what was planned, what was released, what ran, and what passed. Each column’s entry/exit criteria should reference controlled masters (methods, sampling plans, label art), and the card should link to the executed evidence—operator IDs, device IDs, results, exceptions—captured contemporaneously under ALCOA+. If you cannot reconstruct which version of the instruction was in force when the card moved, your board is theater, not control.

Labeling and serialization require extra discipline: bind cards to GS1/GTIN masters; enforce previews with lot/expiry parameters; reconcile labels at line clearance; block Closeout until reconciliation balances. For perishable or potency-driven products, incorporate FEFO logic into ready queues and aging alerts.

9) How This Fits with V5

V5 by SG Systems Global provides an integrated Kanban view across MES, WMS, and QMS so the board is both visual and authoritative. Columns are tied to executable states in the eBMR; card movement is gated by checks: kit verified via Barcode Validation, materials status from Hold/Release, current methods under Document Control, training completion, and equipment availability. WIP limits can be enforced at column or resource family level; attempts to exceed limits require a reason code and appear in dashboards for management review.

Upstream, the Job Queue feeds “Ready” with finite-capacity awareness, and Job Release changes state only when prerequisites pass. Downstream, Closeout compiles yields, IPC summaries, label reconciliation, and deviations for QA review-by-exception; Finished Goods Release is blocked until evidence is complete. Analytics show lead time, cycle time by step, blocker time by category, WIP vs. limit violations, and trends in flow efficiency, feeding CPV and APR/PQR style summaries.

10) FAQ

Q1. How is a Kanban board different from a Gantt chart?
Gantt shows planned timelines. Kanban shows actual flow and constraints. Kanban limits WIP, exposes blockers, and controls starts; Gantt often encourages overcommitment and ignores variability.

Q2. Where should we set WIP limits?
Start at the constraint and the step immediately upstream. Use historical cycle time and variability to size limits. Reduce gradually while watching lead time and flow efficiency; if lead time improves and blocker time doesn’t spike, keep going.

Q3. What belongs in a column’s entry/exit criteria?
The minimum checks that prove the work is ready for that state and done with that state: kit verified FEFO/FIFO, component IDs scanned, effective method and label versions confirmed, equipment availability, IPC plan in place, required signatures captured.

Q4. Can we run Kanban with paper?
You can, but in regulated environments paper quickly drifts from the master and weakens traceability. A digital board tied to eBMR, WMS, and QMS is far more defensible and scalable.

Q5. How do expedites fit?
Define a small, visible expedite lane with a hard WIP cap. Every expedite requires a reason code and shows its cost (which standard cards were delayed). Repeated expedites for the same root cause flow into CAPA.

Related Reading
• Flow & Scheduling: JIT | Heijunka | Job Queue | Job Release
• Execution & Records: Job Traveler | eBMR | Barcode Validation | Finished Goods Release
• Quality & Governance: Document Control | Change Control | Audit Trail (GxP) | Data Integrity | CPV