Just-in-time and kanban

Why it matters

Indian automobile and two-wheeler plants, following Toyota and Maruti Suzuki, run with a few hours of parts inventory instead of weeks. Just-in-time (JIT) production achieves this by making only what the next process has actually used, and kanban is the simple card-or-bin signal that makes it work on the shop floor. Understanding both explains why lot sizes, set-up times and supplier reliability matter so much.

Key ideas

JIT philosophy. Produce the right item, in the right quantity, at the right time, with zero defects – and treat everything else as waste. Inventory is seen not as an asset but as something that hides problems (long set-ups, breakdowns, poor quality, unreliable suppliers), like water hiding rocks in a river. Lowering inventory exposes the rocks so they can be removed.

Push vs pull.

• Push (MRP-style): production is released by a schedule based on forecasts; each stage pushes output to the next whether or not it is needed.
• Pull (JIT): nothing is produced until the downstream process signals that it has used something. The final-assembly schedule pulls parts through the plant; WIP is capped by the number of kanbans.

Building blocks of JIT.

• Levelled production (heijunka) – a stable daily schedule with mixed-model sequencing, so demand on upstream processes is smooth.
• Small lots and set-up reduction – SMED (single-minute exchange of die) cuts set-up times so that small lots become economical (smaller S gives smaller EOQ).
• Quality at source (jidoka) – machines and workers stop the line when a defect appears (andon, poka-yoke); there is no stock to hide bad parts.
• Preventive and total productive maintenance – breakdowns are not buffered by inventory.
• Cellular/flow layouts and multi-skilled workers.
• Supplier partnerships – few, nearby, certified suppliers delivering small quantities frequently, often directly to the line.
• Continuous improvement (kaizen) and respect for people.

Kanban. Japanese for "card" or "signboard". A kanban authorises a fixed quantity (one standard container) to be produced or moved.

• Withdrawal (conveyance, move) kanban – authorises moving a container from the supplying process's outbound stock to the using process.
• Production kanban – authorises the supplying process to produce one container to replace what was withdrawn.
• Signal (triangle) kanban – used where the supplying process works in batches (e.g. a press); signals when stock falls to a reorder level.
• Supplier kanban – used with outside suppliers. In a single-card system only withdrawal or only production kanbans are used; in a dual-card (Toyota) system both.

Kanban rules (Toyota).

1. The downstream process withdraws only what it needs, when it needs it.
2. The upstream process produces only the quantity withdrawn, in the order withdrawn.
3. No item is made or moved without a kanban.
4. A kanban is always attached to the container.
5. Defective parts are never sent to the next process.
6. The number of kanbans is reduced gradually to expose problems and improve the process.

Number of kanbans. Enough containers are needed to cover demand during the time a container takes to go round the loop (waiting + processing + moving), plus a safety allowance. Maximum inventory in the loop = N × C.

Limits. JIT needs stable, level demand, reliable suppliers and short set-ups. It is vulnerable to supply disruptions (as seen with natural disasters and pandemics), so many firms now keep strategic buffers for critical parts while running JIT elsewhere.

Formulas

N = D·(T_w + T_p)·(1 + α) / C

• N = number of kanbans (containers), rounded up; D = demand rate (units/day or units/h); T_w = waiting time of a container (days or h); T_p = processing time per container including move (same unit); α = safety (policy) factor (fraction, typically ≤ 0.10 in mature plants); C = container capacity (units per container).

Maximum inventory in the loop = N · C

• Units.

Dual-card system: N_w = D·T_w·(1 + α)/C, N_p = D·T_p·(1 + α)/C

• T_w here = withdrawal (conveyance) cycle time, T_p = production cycle time.

Mixed-model cycle: ratio of daily quantities reduced to smallest integers; cycle time = available time / total units

• For levelled (heijunka) sequencing.

Worked examples

Example 1 (standard) – number of kanbans. A machining cell supplies an assembly line using 2400 shafts per day. Containers hold 60 shafts. A container waits on average 0.15 day and takes 0.10 day to process and move (total 0.25 day). Safety factor α = 10 %.

1. N = D(T_w + T_p)(1 + α)/C = 2400 × 0.25 × 1.10/60.
2. N = 660/60 = 11 kanbans.
3. Maximum WIP in the loop = 11 × 60 = 660 shafts, about 0.28 day of demand.
4. If set-up reduction cuts the loop time to 0.15 day: N = 2400 × 0.15 × 1.10/60 = 6.6 → 7 kanbans, so WIP falls to 420 shafts.

Example 2 (GATE level) – dual-card system and levelled sequence. (a) A feeder process supplies 600 units/day to assembly; container size 30 units; α = 0.20. Conveyance cycle T_w = 0.1 day; production cycle T_p = 0.3 day.

1. Withdrawal kanbans N_w = 600 × 0.1 × 1.2/30 = 2.4 → 3.
2. Production kanbans N_p = 600 × 0.3 × 1.2/30 = 7.2 → 8.
3. Total containers = 11; maximum inventory = 11 × 30 = 330 units. (b) The same line assembles three models daily: A 400, B 200, C 100 units in a 480-minute shift.
4. Ratio 400 : 200 : 100 = 4 : 2 : 1, so the minimum repeating cycle has 7 units, e.g. A–B–A–C–A–B–A, repeated 100 times a day.
5. Cycle time = 480/700 = 0.686 min per unit (about 41 s).
6. Upstream processes now see a steady mix every 7 units instead of large batches of one model.

Common mistakes

• Forgetting to divide by container size and quoting the number of units as the number of kanbans.
• Rounding the number of kanbans down; it must be rounded up.
• Using inconsistent time units (demand per hour with waiting time in days).
• Calling MRP a pull system: MRP pushes orders from a schedule; kanban pulls from actual use.
• Thinking JIT means "zero inventory": it means minimum necessary inventory, reduced continuously.
• Ignoring set-up time: without SMED, small lots make JIT uneconomic.

For GATE PI

Expect direct numericals on the number of kanbans (with or without a safety factor), maximum inventory in a kanban loop, and conceptual questions on push vs pull, kanban types and rules, heijunka, SMED, jidoka and the features of JIT purchasing. Practise the kanban formula with careful unit conversion and rounding up.

Quick check

1. D = 500 units/day, loop time 0.4 day, α = 0, container 25 units. How many kanbans?
2. In Q1, what is the maximum inventory in the loop?
3. Which kanban authorises a supplying process to make more parts?
4. What does SMED reduce, and why does that help JIT?

Answers: 1. 500 × 0.4/25 = 8 kanbans. 2. 8 × 25 = 200 units. 3. Production kanban. 4. Set-up (changeover) time; it makes small lots economical.