Aarna Appliances: Kanban Cards, WIP and a Quality Spend the CFO Will Not Fund
Aarna Appliances assembles mixer grinders on one line: eight models, average demand 400 units a day over 250 working days. A motor sub-assembly feeder line supplies the main line in containers of 20 units. Replenishment lead time is 1.5 days, made up of processing, waiting for the winding changeover and transport; planning adds a 10% safety margin. The plant runs batches of 500 because the winding changeover takes 240 minutes, and it averages six changeovers a week. Work in process between the two lines is 3,200 units. Last year's quality costs (Rs lakh): prevention 6, appraisal 4, internal failure 18 (rework 11, scrap 7), external failure 12 (warranty 8, dealer returns 4). The plant manager proposes Rs 10 lakh more on prevention - poka-yoke on winding and a supplier audit - expecting internal failure to fall 60% and external failure 50%. The CFO says quality spending never pays back in the same year.
1. Size the kanban for the motor feeder and use the answer to diagnose what is actually wrong on the line.
Read the worked answer
Kanban size: N = D × L × (1 + safety) ÷ C = 400 × 1.5 × 1.10 ÷ 20 = 660 ÷ 20 = 33 cards, that is, 660 units of authorised inventory.
The diagnosis lies in the gap between authorised and actual inventory. WIP is 3,200 units, about eight days of demand and roughly five times what 33 cards authorise. So either the loop is not being obeyed and cards are being added informally, or the process cannot respond within the stated lead time and people buffer with stock. Both point away from the card arithmetic.
The lead time of 1.5 days is not process time; it is dominated by waiting for the 240-minute changeover and for a batch of 500 to be completed. With six changeovers a week on eight models, the feeder is a batch shop pretending to be a flow line. A kanban sized on that lead time simply institutionalises the batch.
Reduce the changeover first. If SMED halves winding changeover time and the batch falls, the lead time might drop to 0.5 days, giving N = 400 × 0.5 × 1.10 ÷ 20 = 11 cards - a two-thirds cut in stored parts from the same demand. Also, one loop across eight models is wrong: each model or model family needs its own card count, or a levelled sequence at the heijunka point, otherwise cards migrate to the fast movers and the slow movers starve. Finally, the 10% safety margin is a fixed allowance for variability that should be reduced as variability is eliminated, not treated as permanent.
2. Evaluate the Rs 10 lakh prevention proposal on the case numbers and answer the CFO's objection.
Read the worked answer
Current split: prevention 6 + appraisal 4 = Rs 10 lakh of prevention and appraisal; internal 18 + external 12 = Rs 30 lakh of failure. Total quality cost Rs 40 lakh, of which 75% is the price of not doing it right first time.
Proposal: spend Rs 10 lakh more. Internal failure falls 60% of 18 = Rs 10.8 lakh; external failure falls 50% of 12 = Rs 6.0 lakh; total avoided failure cost Rs 16.8 lakh. Net benefit Rs 16.8 − 10.0 = Rs 6.8 lakh a year, and total quality cost falls from Rs 40 lakh to Rs 33.2 lakh.
The payback is inside twelve months, so the CFO's blanket statement is wrong. But two qualifiers stop the number being taken at face value. First, composition of the saving: the Rs 11 lakh of rework includes absorbed labour that will still be paid, so the cash saving is smaller than Rs 16.8 lakh; the genuine cash is scrap material (Rs 7 lakh, of which 60% = Rs 4.2 lakh), warranty (Rs 8 lakh × 50% = Rs 4 lakh) and released winding capacity, which has value only if it is used to make more grinders or drop overtime. Second, the benchmark is understated: recorded warranty of Rs 8 lakh excludes dealers who do not claim, and the lost repeat purchase from a returning customer is not in the ledger at all, so external failure is worth more than the books show.
Fund the Rs 10 lakh as a staged release, with the first tranche tied to the supplier audit and a measured fall in winding rework within one quarter, and track failure cost and first-pass yield monthly rather than relying on the annual accounts.
3. The winding supplier is 400 km away and delivers once a week. What does that do to the lean plan, and what would you change first?
Read the worked answer
A weekly delivery from 400 km away puts roughly five days of inbound stock outside the plant, so inventory is not 660 units in the loop - it is the loop plus a week of steel and magnets in transit and at the dock. The value stream being optimised stops at the gate while the waste sits upstream. Correcting the internal kanban while leaving the inbound pipeline weekly gives a local optimum and a fragile one: any transport delay stops the feeder.
Sequence of action. First, stabilise the schedule: a levelled weekly sequence across the eight models is the precondition for any of this, because a supplier cannot be asked to deliver frequently to a schedule that changes daily. Second, attack the inbound pipeline with the supplier - a milk run that collects from two or three suppliers on one route, or twice-weekly deliveries with supplier-owned kanban, converting the delivery frequency decision into a transport-cost versus inventory-cost calculation. Third, move the cheap, high-volume, low-value items (packaging, fasteners) close to point of use, and keep the high-value motor parts on a controlled loop. Fourth, do not try to fix supplier frequency while the 240-minute changeover remains, because long changeovers force large batches, and large batches need large deliveries; the internal SMED work has to come first.
The judgement the examiner is looking for: lean is a system, and the binding constraint here is the changeover plus schedule instability, not the kanban formula.