K
KnowMBAAdvisory
OperationsAdvanced7 min read

Heijunka

Heijunka (Japanese for 'leveling') is the Toyota practice of producing a small mix of every product every day rather than batching production by model. Instead of building 1,000 sedans Monday and 1,000 SUVs Tuesday, a heijunka schedule builds Sedan-SUV-Truck-Sedan-SUV-Truck in a continuous mix at the same hourly takt. Why it matters: large batches amplify demand swings into massive operational shocks (the bullwhip effect). Small mixed batches dampen swings, allow tighter inventory, surface defects faster, and let the system absorb demand changes without panic. Heijunka is the prerequisite that makes just-in-time inventory possible; without leveled demand, JIT collapses into chronic stockouts and firefights.

Also known asProduction LevelingWorkload SmoothingMixed-Model SchedulingDemand SmoothingLevel Loading

The Trap

Trying to apply heijunka to upstream demand without first stabilizing the downstream signal. If your sales team books huge end-of-quarter orders to hit numbers, your factory CAN'T level — it's responding to a spiky signal. Real heijunka requires fixing the demand-generating behavior first (changing sales comp, building order pipelines, setting daily commits with customers) and THEN smoothing production. Companies that try to 'level' production while sales remains spiky just push the chaos to inventory or backlog instead. The other trap: leveling so aggressively that mix changes every minute, creating massive setup-time waste. Heijunka requires that you've also reduced setup times (SMED) to make small batches economical.

What to Do

Step 1: Map your current production schedule by week — what % is run as large batches of single SKU? Step 2: Pick the top 3 SKUs (typically 70-80% of demand). Build a leveled schedule that produces all 3 every day in proportion to demand (e.g., if Sedan is 50%, SUV 30%, Truck 20% of demand, build them in a 5-3-2 daily ratio). Step 3: Reduce setup/changeover times so smaller batches are economical (target: under 10 minutes per changeover via SMED — Single Minute Exchange of Die). Step 4: Lock the schedule for a fixed window (e.g., 2 weeks frozen, then re-level). Track inventory turns, on-time delivery, and overtime as the leveling deepens.

Formula

Heijunka Mix Ratio = (Daily demand for SKU A : SKU B : SKU C) — produced in this ratio every shift

In Practice

Toyota's Tahara plant in Japan builds 8 different models on the same line, mixed in approximately demand proportion every day. Setup time per model changeover: under 90 seconds (vs. typical industry 30-45 minutes). The benefit: when a model's demand suddenly drops 20%, the line just shifts the mix the next day — no inventory buildup, no overtime, no scramble. Compare to U.S. competitors of the 1980s, who would batch 5,000 of model A then switch to 5,000 of model B; if demand shifted mid-batch, they'd be stuck with thousands of unsold cars and a six-week lead time on the now-popular model. Heijunka was a key reason Toyota's working capital requirements were 60-70% lower than GM's per vehicle produced.

Pro Tips

  • 01

    Heijunka is downstream of SMED. If your changeover takes 4 hours, you literally can't run a leveled mix — the math forces large batches. Reducing changeover time from hours to minutes is the unlock that makes heijunka possible. Toyota spent 20 years getting die-press changeovers from 4 hours to under 3 minutes.

  • 02

    The 80/20 rule applies: level your top SKUs (which are 70-80% of demand) and run low-volume SKUs in dedicated weeks. Trying to level every SKU in a 200-product catalog is overkill. Heijunka is for runners (high-volume), not for one-off custom orders.

  • 03

    Use a heijunka box (a physical scheduling board with kanban cards in time-slotted columns). Every two hours, the next batch of cards triggers production. When a card is missing or extra, the deviation is visible immediately. Excel can't replace this for shop-floor leveling — visibility is the whole point.

Myth vs Reality

Myth

Heijunka means producing the same amount every day regardless of demand

Reality

Heijunka means producing a leveled MIX every day at a rate that matches average demand. If demand swings, the mix and total adjust on weekly cycles — but within the cycle, hourly production is leveled. It's smoothing, not freezing.

Myth

Heijunka only works in mass production

Reality

Heijunka principles apply to any repeatable workflow. Software teams use 'flow leveling' — release a steady stream of small features daily instead of monthly mega-releases. Hospitals level surgery scheduling instead of front-loading Mondays. Consulting firms level project intake to avoid quarterly capacity crises. The principle — never amplify natural variation into operational shocks — is universal.

Try it

Run the numbers.

Pressure-test the concept against your own knowledge — answer the challenge or try the live scenario.

🧪

Knowledge Check

Your factory currently batches: Monday-Tuesday make 1,000 of Product A, Wednesday-Thursday make 600 of Product B, Friday make 400 of Product C. Your sales lead time is 10 days and on-time delivery is 78%. Most likely root cause?

Industry benchmarks

Is your number good?

Calibrate against real-world tiers. Use these ranges as targets — not absolutes.

Demand Variance Reduction After Heijunka Implementation

Manufacturing and assembly plants 12-24 months post-implementation

Toyota-Class

60-80% variance reduction

Strong Implementation

40-60% reduction

Partial Implementation

20-40% reduction

Failed (sales not aligned)

< 20%

Source: Toyota Production System studies / Womack & Jones research

Real-world cases

Companies that lived this.

Verified narratives with the numbers that prove (or break) the concept.

🚗

Toyota (Origin of Heijunka)

1960s-Present

success

Heijunka was developed by Taiichi Ohno as a counter to the American mass-production model that batched cars by model. Toyota's insight: if Camry, Corolla, and RAV4 are built in proportion every day, a sudden shift in demand mix doesn't require shutting down a line or building 5,000 unsold cars. Toyota's Tahara plant builds 8 models on a single line with sub-90-second changeovers. The result: working capital per vehicle is 60-70% lower than competitors, lead times are 50% shorter, and the system absorbs demand shocks (like 2008-2009 financial crisis) with proportionally less disruption than U.S. peers.

Models per Line (Tahara)

8

Changeover Time

< 90 seconds

Working Capital per Vehicle vs. GM

60-70% lower

Demand Shock Recovery (2009)

6 months vs. competitors' 18+ months

Heijunka is what makes lean inventory possible. Without leveled production, JIT becomes JISO (Just-In-Stockout). Toyota's competitors copied JIT and crashed; they had skipped heijunka.

Source ↗
🥫

Hypothetical: Mid-Size CPG Co-Manufacturer

Recent

success

A contract food manufacturer ran 14 SKUs on 4 production lines, batching each SKU in 2-week runs. End-of-quarter brand orders created chronic crises: overtime costs ran 22% of labor, on-time delivery was 76%, finished-goods inventory was 42 days. After 9 months of work — first reducing changeovers from 6 hours to 45 minutes (SMED), then leveling top 5 SKUs to daily mix — overtime dropped to 6%, on-time delivery rose to 94%, and finished inventory shrank to 16 days. The business case was driven by the inventory release: $4.2M in working capital freed, plus $800K/year in overtime savings.

Overtime % of Labor

22% → 6%

On-Time Delivery

76% → 94%

Finished Goods Inventory

42 days → 16 days

Working Capital Freed

$4.2M

Heijunka in CPG and food manufacturing usually pays for itself in working capital release alone, before the labor and service-level benefits show up. SMED first, then level, then watch the cash flow change.

Decision scenario

The End-of-Quarter Production Crisis

You're VP Operations at a 400-person CPG manufacturer. End-of-quarter is brutal: 18% overtime, frequent stockouts, customer escalations weekly. The CRO insists his team needs the quarterly cadence to hit numbers. The CFO wants production to 'just absorb' the variability. Your COO predecessor lasted 14 months.

Overtime (avg)

18% of labor

Stockouts/qtr

32 SKU-weeks

Inventory

38 days finished goods

On-Time Delivery

79%

01

Decision 1

The CEO has given you 18 months. The CRO won't easily change quota cadence; the CFO is breathing down your neck about working capital. You have to choose your sequencing carefully.

Implement heijunka in production immediately — level the schedule and force the rest of the org to adaptReveal
You smooth production but sales still spikes at quarter-end. Now you're stuck with high inventory in week 1-7 and stockouts in week 12-13. Customer complaints rise. CRO blames you. CFO is angry about working capital. You become the problem. Six months in, you're rolling back heijunka under pressure.
Inventory: 38 days → 51 daysStockouts: 32 → 41 SKU-wksYour political capital: Burned
Spend month 1-3 building the data case: visualize cost-per-unit by week showing 2-3x cost in week 12-13. Take it to CEO and CRO together. Negotiate monthly quotas + early-order discounts. THEN reduce changeovers (SMED) in months 4-9. THEN level production in months 10-15.Reveal
Data case shifts the CRO — he agrees to monthly quotas because his AEs were burning out anyway. SMED work in months 4-9 cuts changeovers from 4 hrs to 35 min. By month 12, top 5 SKUs are leveled. Overtime drops to 5%, stockouts to 8/qtr, inventory to 19 days, on-time to 95%. CFO writes you a glowing review. You stay 5+ years.
Overtime: 18% → 5%Inventory: 38 days → 19 daysOn-Time Delivery: 79% → 95%Working Capital Freed: $3-4M

Related concepts

Keep connecting.

The concepts that orbit this one — each one sharpens the others.

Beyond the concept

Turn Heijunka into a live operating decision.

Use this concept as the framing layer, then move into a diagnostic if it maps directly to a current bottleneck.

Typical response time: 24h · No retainer required

Turn Heijunka into a live operating decision.

Use Heijunka as the framing layer, then move into diagnostics or advisory if this maps directly to a current business bottleneck.