Kanban – Pull Replenishment
This topic is part of the SG Systems Global manufacturing, quality, and supply-chain glossary series.
Updated October 2025 • Lean Material Flow • Supermarkets, Signals & WIP Limits
Kanban is a pull-replenishment system that replaces consumed items with precisely sized lots signaled by cards, bins, or electronic triggers. Instead of pushing work or materials based on forecasts, Kanban authorizes production or picking only when downstream consumption creates demand. Properly designed, it stabilizes flow, caps work-in-process, and exposes problems that Heijunka and JIT alone cannot solve. In practice, Kanban takes tangible forms—two-bin loops, carded totes, electronic signals on Kanban Boards, or WMS tasks—each tied to standard pack sizes, defined supermarkets, and disciplined replenishment routes.
Kanban is not a scheduling panacea. If your processes are unstable, lead times swing wildly, or basic accuracy is broken, Kanban will amplify the pain. It shines when the basics are in control: parts are identifiable and scanned via Barcode Validation, shelf life is honored with FEFO/FIFO, and counts are credible (see Inventory Accuracy). With that foundation, Kanban reduces overproduction, frees cash, and keeps constraint resources fed without flooding the floor.
“Kanban replaces ‘make to plan’ with ‘make to signal.’ If there’s no pull signal, don’t start. If there is, replenish exactly what was consumed—no more, no less.”
1) What It Is (Unbiased Overview)
Kanban is a control mechanism that ties production or replenishment authorization to actual consumption. When a downstream user empties a container or scans a card, the signal travels upstream to trigger a standard lot of work or materials. The system’s heartbeat is defined by supermarkets—buffer locations that decouple processes—plus a disciplined route (water spider, milk-run, AGV) and a visual board that shows each card’s status. Because the total number of cards is fixed, so is the WIP. Flow stabilizes, queues shrink, and variability becomes visible rather than hidden under piles of inventory.
Kanban can be physical (cards, bins, shelves) or electronic (eKanban). In regulated or high-mix environments, eKanban within WMS/MES adds traceability, aging controls, and audit trails. Either way, the essence is the same: no signal, no work.
2) Core Concepts & Design Choices
Supermarkets. Dedicated, labeled locations that hold a limited variety with defined max/min. They protect constraints and enable level loading (see Heijunka).
Card/Container. The atomic unit of pull. Each card equals one replenishment lot with a standard pack, label template, and origin/destination encoded (preferably with GS1 keys and, if used, EPCIS events).
Pitch and Takt. Pitch sets the expected time to consume one container; takt sets the customer tempo. Card counts are sized so supermarkets ride between min and max across normal variability.
Routes & Roles. A water spider (or AGV) circulates at a fixed cadence to collect empties and deliver fulls; production teams build to card, not forecast; planners adjust only when demand or lead time realistically shifts.
Signal Integrity. No card duplication, no off-system pulls. Broken discipline breaks Kanban. Exceptions require explicit deviation with reason codes.
3) Sizing the System: The Math That Matters
At its core, Kanban sizing balances consumption rate, replenishment lead time, and variability. A simple starting formula for the number of containers (cards) is: Cards = Ceiling[(Average Demand per Time × Replenishment Lead Time) / Container Quantity] + Safety. Safety is typically 10–30% of demand during lead time, adjusted via SPC on withdrawal and cycle-time charts (see SPC). For perishables, overlay FEFO constraints and cap containers so worst-case dwell stays inside shelf life.
Common patterns: Two-bin for fasteners and low-value items; Carded totes for components; eKanban for high-mix or supplier loops; and CONWIP for controlling total WIP in a routed process when part numbers vary too much for itemized cards.
4) Signals, Movement, and Data
Withdrawal. Operator empties a container and places its card in the pickup lane or scans a “Kanban empty” code. The board shows a new pull request. The water spider collects empties on the route.
Replenishment. Upstream cell builds exactly one container of the specified part to standard; WMS creates a directed task to stage in the supermarket (see Directed Picking and Bin Location).
Traceability. Each container carries lot/exp/date codes, captured at putaway and withdrawal. For regulated contexts, integrate with Batch Genealogy and eBMR so point-of-use scans bind component lots to the work order.
Supplier Loops. For external suppliers, eKanban sends ASN-like signals; receiving executes Goods Receipt and status (Component Release) before cards become available to pull.
5) Governance, Master Data, and Compliance
Kanban lives or dies on master data: standard pack, unit of measure, label templates, home bin locations, and supplier lead times. All of it must be controlled under Document Control with change history. In electronic systems, approvals and edits require Part 11 e-signatures and full audit trails. For food and pharma, FEFO rotation and segregation rules must be embedded so a pull cannot violate shelf-life or status controls.
Operational governance is simple: daily tier meetings review the board, shortages, and aging by card; weekly reviews assess sizing against actuals; monthly, leadership checks KPI trends (turns, service level, WIP, expedites) and approves resizing when sustained shifts occur.
6) Common Failure Modes (and How to Fix Them)
Phantom inventory. Counts don’t match reality, so cards lie. Fix: lock down IDs, mandatory scans, cycle counting (see Cycle Counting), and visible discrepancies on the board until resolved.
Over-sizing. Too many cards hide problems and swell cash. Fix: remove cards systematically; watch service level and lead time; hold at the smallest set that meets target.
Under-sizing. Stockouts and line starvation. Fix: analyze withdrawal SPC for spikes, increase safety temporarily, then address root causes (setup time, supplier variability).
Bypass behavior. People “just grab one” from bulk or create ad-hoc pulls. Fix: eliminate shadow stock, enforce putaway, and require deviation reason codes.
Label/lot errors. Wrong revs or mixed lots in one container. Fix: print from master data only, require scans at fill and withdrawal, reconcile labels at supermarket like mini line-clearance.
7) Metrics That Prove Pull Is Working
Service level at the point of use (fill rate without expedite). Stockouts per 1,000 withdrawals. Turns and days on hand by item family. Average & 90th percentile lead time from signal to availability. WIP (cards in system). Expedites as a percent of moves. FEFO adherence and ageing distributions for perishables. Trend with control limits; react to signals, not vibes.
8) How This Fits with V5
V5 by SG Systems Global makes Kanban executable. In V5 WMS, each card becomes a digital object with part, pack size, home bin, and status. Empties trigger directed tasks to replenish from the supermarket or supplier; Directed Picking guides pathing; FEFO logic blocks expired/soon-to-expire lots. In V5 MES, component scans at the station bind lot IDs to the order and update Genealogy. The Kanban Board shows signals, in-process replenishments, blockers, and ageing; Job Queue protects constraints by limiting WIP even when demand spikes.
Upstream changes—new pack size, new bin—are controlled under Document Control with audit trails. Analytics in V5 trend KPIs like stockouts, lead time, and turns; CPV/SPC dashboards separate noise from real shifts so card counts are resized deliberately, not emotionally.
9) Practical Walkthrough (Example)
A nutraceutical plant struggles with line stops from cap-and-seal components. The team designs a Kanban loop:
- Scope & Data. 12 high-use SKUs, average 180 withdrawals/day. Replenishment lead time from receiving to supermarket: 90 minutes; from molding cell to supermarket: 4 hours.
- Design. Supermarket near fillers with two lanes per SKU; standard tote = 600 caps; water spider route every 30 minutes. Initial cards per SKU: demand × lead time / pack + 20% safety → 180 × 1.5hr / 600 ≈ 0.45 → 1 card minimum; for molding replenishment, 180 × 4hr / 600 ≈ 1.2 → 2 cards; combined loop set to 3 cards per SKU to start.
- Execution. Labels carry GTIN, lot, expiry; V5 WMS enforces FEFO and location moves; scans at fill and withdrawal update the board.
- Stabilization. First week: stockouts drop 80% but ageing rises on two SKUs. SPC shows over-estimated safety; remove one card each, add supplier delivery window alert.
- Lock-in. Update PFEP item master, bin diagrams, and water spider SOP under Document Control. Metrics feed the site’s KPI deck and APR.
10) FAQ
Q1. Is Kanban the same as scheduling?
No. Scheduling allocates time and resources (see Job Scheduling); Kanban authorizes replenishment based on consumption. Use both: schedule the constraint, pull everything else.
Q2. Two-bin, min/max, or cards—which should we use?
Two-bin is simplest for low-value, high-velocity items. Min/max with eKanban suits mixed demand and supplier loops. Physical cards work when visual control is paramount. Start simple; upgrade only when the problem demands it.
Q3. How do we size safety stock in Kanban?
Use demand-during-lead-time variability. Start with 10–30% of average DDLT, then tune using service-level targets and SPC; remove safety when process variability drops.
Q4. Can Kanban work with perishables?
Yes—embed FEFO, limit card count so dwell stays within shelf life, and design smaller packs with more frequent routes.
Q5. What breaks Kanban fastest?
Bypassing scans, hoarding, inaccurate masters, and undisciplined exceptions. Fix the system, not the people: make the right action the easiest.
Related Reading
• Flow & Leveling: Heijunka | JIT | Job Queue
• Materials & Identity: Directed Picking | Bin Location Management | GS1/GTIN | EPCIS | Barcode Validation
• Quality & Records: Batch Genealogy | FEFO | Audit Trail | Document Control
• Ops & Visibility: Kanban Board | Inventory Accuracy | KPI | Live Production Visibility