8.9 Cycle-Time Reduction: Continuous Flow, SMED, and Heijunka

Why cycle time is mostly waiting

In most processes the unit spends the overwhelming majority of its lead time waiting, not being worked on. Little's Law makes the mechanism explicit:

$\text{Lead time} = \frac{WIP}{\text{Throughput}}$

Reducing lead time therefore means reducing work in process, and the three biggest sources of WIP are batching, imbalance, and demand turbulence. The three tools in this section address them in turn.

Continuous flow

Continuous flow (one-piece flow) moves each unit directly to the next operation as soon as it is finished, rather than accumulating a batch.

The arithmetic is worth doing once. Three operations at 1 minute each, batch of 10:

• Batch processing: all 10 units complete operation 1 (10 min), then operation 2 (10 min), then operation 3 (10 min). First unit finished at minute 21; last at minute 30.
• One-piece flow: first unit finished at minute 3; last at minute 12.

Lead time falls from 30 minutes to 12, and the first good unit appears at minute 3 rather than 21, which also means a defect introduced at operation 1 is discovered within three minutes rather than after ten units have been made.

Prerequisites for flow:

Flow exposes problems that inventory used to hide, which is why lean transformation and Six Sigma variation reduction have to proceed together.

SMED: single-minute exchange of die

Developed by Shigeo Shingo, SMED reduces changeover time. "Single minute" means a single digit of minutes -- under ten -- not sixty seconds.

The central distinction

The three stages

1. Separate internal from external. Simply identifying which activities do not require a stopped machine, and doing those in advance, typically cuts changeover time by 30% to 50% with no capital expenditure. Most organizations do everything after the machine stops purely out of habit.
2. Convert internal to external. Pre-heat dies, pre-assemble the next fixture on a cart, use duplicate tooling, standardize die heights so shimming is unnecessary.
3. Streamline both. Replace bolts with quick clamps and cam locks, use functional clamps and one-turn fasteners, eliminate adjustment through positive stops and numerical settings, run parallel operations with two people on opposite sides of the machine.

The largest single time sink in most changeovers is adjustment and trial runs after restart. Eliminating adjustment through gauge blocks, hard stops, and recorded settings often saves more than speeding up the physical exchange.

Why SMED changes the economics

The economic batch quantity is

$EBQ = \sqrt{\frac{2DS}{H}}$

where $D$ is annual demand, $S$ the setup cost, and $H$ the holding cost per unit per year. Batch size scales with the square root of setup cost, so cutting setup time by a factor of 9 cuts the economic batch size by a factor of 3.

Worked example. $D = 100{,}000$ units, holding cost $H = $2$ per unit per year, setup cost $S = $450$ (a 3-hour changeover at $150 per hour).

$EBQ = \sqrt{\frac{2 \times 100{,}000 \times 450}{2}} = \sqrt{45{,}000{,}000} = 6{,}708 \text{ units}$

After SMED reduces changeover to 20 minutes, $S = $50$:

$EBQ = \sqrt{\frac{2 \times 100{,}000 \times 50}{2}} = \sqrt{5{,}000{,}000} = 2{,}236 \text{ units}$

Batch size falls by two-thirds, and with it WIP, lead time, and the delay between making a defect and finding it. This is why SMED is the enabler for flow and pull rather than an isolated efficiency project.

Heijunka: production levelling

Heijunka levels production so that the same mix and volume are produced in each short interval, converting lumpy customer demand into a repeatable internal pattern.

Two dimensions:

• Volume levelling: produce a consistent quantity per period rather than following demand spikes.
• Mix levelling: produce a repeating sequence of product types rather than long runs of each.

Worked example. Weekly demand is 500 A, 300 B, and 200 C, over 5 days.

Levelled mix requires short changeovers, which requires SMED. The three tools form a chain: SMED makes small lots affordable, heijunka sequences them, and flow moves them.

Heijunka is usually visualized on a heijunka box -- a physical grid of pigeonholes, rows for product types and columns for time intervals, holding kanban cards that release work in the levelled sequence.

The demand amplification argument

The deeper reason for levelling is that unlevelled internal scheduling amplifies demand variation upstream. Each stage adds safety stock and reorder lumpiness, so a modest fluctuation in customer demand becomes a large swing at the supplier -- the bullwhip effect. Levelling the internal schedule dampens the signal instead of amplifying it, which is why heijunka benefits suppliers as much as the plant that adopts it.

Choosing among them