New to this topic?
Read these first to get the most out of this guide:
The volume-leveling table above, drawn as running totals. Orders 35 + 12 + 28 + 8 + 17 = 100 parts; leveled production 100 ÷ 5 days = 20 a day. Cumulative orders 35, 47, 75, 83, 100 against cumulative production 20, 40, 60, 80, 100, so the shortfall at each day end is 15, 7, 15, 3, 0: worst 15, on Monday and again on Wednesday. Start the week holding 15 finished parts and nothing ships late: on-hand runs 15 → 0 → 8 → 0 → 12 → 15, back where it began, so the following week starts identically. 15 ÷ 20 = 0.75 of one day's output. Both totals are straight within a day, so the worst gap can only fall on a day boundary. The unleveled column in the table is the same line as the orders; it needs no supermarket, and buys that with a Monday of 35 (75% above the 20 average) and a Thursday of 8 (60% below).
The Batch Scheduling Problem
Most factories schedule in batches: run all of Part A, then changeover, run all of Part B, then changeover, run all of Part C. This creates several problems:
| Problem | Cause | Impact |
|---|---|---|
| Demand amplification | Upstream sees lumpy demand (all A, then all B) | Upstream either overproduces or cannot respond fast enough |
| Inventory peaks | Finished goods of A pile up while B and C are out of stock | High inventory cost + stockouts simultaneously |
| Resource spikes | Monday needs 4 operators; Thursday needs 1 | Overtime followed by idle time; unstable staffing |
| Long lead time | If you need Part C and it runs once per week, wait up to 5 days | Customer lead time = batch interval, not cycle time |
Two Levels of Leveling
Level 1: Volume Leveling
Produce the same total quantity every day, regardless of daily order variation. If monthly demand is 400 parts and there are 20 working days, produce 20 parts every day, not 35 on Monday and 5 on Friday.
| Day | Customer Orders | Unleveled Production | Leveled Production |
|---|---|---|---|
| Monday | 35 | 35 | 20 |
| Tuesday | 12 | 12 | 20 |
| Wednesday | 28 | 28 | 20 |
| Thursday | 8 | 8 | 20 |
| Friday | 17 | 17 | 20 |
| Total | 100 | 100 | 100 |
Same weekly output. But leveled production uses the same resources every day (no Monday overtime, no Thursday idle time) and ships from a small finished-goods supermarket that absorbs daily variation.
How small is that supermarket? Exactly as deep as the orders ever get ahead of the cell: 15 parts, three-quarters of one day’s output. Hold that and nothing ships late; hold much more and you have paid for leveling twice.
Level 2: Mix Leveling
Produce every product type every day (or every pitch period), rather than batching by product. If you make 20 parts/day across 4 types (A=8, B=5, C=4, D=3), produce A-B-C-D-A-B-A-C-A-D-A-B-A-C-A-D-A-B-A-C in a repeating mixed-model sequence.
EPEI: Every Part Every Interval
EPEI measures how frequently you can produce every part number. If your changeover allows running each part type once per day, EPEI = 1 day. If changeovers are too long and you can only run each type once per week, EPEI = 5 days. EPEI is the primary metric for measuring how well you have leveled mix. The goal is EPEI = 1 day or less. SMED is the tool that makes shorter EPEI economically feasible.
The Heijunka Box
A heijunka box is a physical grid that converts a daily production plan into a time-sequenced schedule. Columns are time periods (typically “pitch” increments). Rows are product types. Kanban cards are placed in the grid to create the leveled sequence.
Pitch = Takt time × pack-out quantity. If Takt is 5 minutes and parts ship in bins of 10, pitch = 50 minutes. This is the scheduling increment: the heijunka box has one column per pitch period.
Cell: Produces 4 bracket types. Takt = 6 min. Bin = 5 parts. Pitch = 30 min. 8 pitches per shift.
| Part | P1 | P2 | P3 | P4 | P5 | P6 | P7 | P8 | Daily |
|---|---|---|---|---|---|---|---|---|---|
| Bracket A | A | A | A | A | 20 | ||||
| Bracket B | B | B | B | 15 | |||||
| Bracket C | C | 5 |
The operator takes the next kanban card from the box at each pitch boundary. The sequence ABABABCAB produces all three types every shift in a leveled pattern. Upstream processes see stable, predictable demand instead of lumpy batches.
Leveling in Aerospace
Aerospace has unique challenges for heijunka: low volumes, long cycle times, and high mix. But the principles still apply:
| Aerospace Context | Leveling Approach |
|---|---|
| Final assembly (4 to 8 aircraft/month) | Level the rate: one aircraft enters the line on a fixed Takt cadence (e.g., every 5 working days). Do not start 3 in one week and 1 the next. |
| Component machining (high mix, 200+ PNs) | Group into product families. Level each family: produce every family every day using SMED to minimize changeover. |
| Kit staging | Stage kits on a daily rhythm aligned to assembly Takt. One kit per position per day, not weekly batches. |
| Supplier deliveries | Request daily or twice-weekly deliveries instead of monthly bulk. Smaller, more frequent deliveries = less inventory + faster problem detection. |
The Bottom Line
Heijunka transforms chaotic batch scheduling into a predictable, leveled rhythm. Level volume first (same quantity every day), then level mix (every part type every day). Use the heijunka box to convert daily plans into pitch-by-pitch sequences. The enabling tool is SMED: changeover reduction that makes frequent product switches economically viable. The result: stable upstream demand, lower inventory, shorter lead times, and a production system that runs the same way every day instead of firefighting different crises each week. Next: SMED & Quick Changeover, the tool that enables heijunka by making changeovers fast enough to level mix.
Interactive Demo
The same monthly mix, scheduled in campaigns and then levelled. Watch what each pattern does to inventory, changeovers and the customer waiting for the model you built last.
Heijunka Production Leveling
Toggle between batch and leveled production to see how leveling reduces inventory peaks. Arrange the heijunka box to visualize the mixed-model schedule.
| Day | S1 | S2 | S3 | S4 | S5 | S6 | S7 | S8 | S9 | S10 |
|---|---|---|---|---|---|---|---|---|---|---|
| Day 1 | A | A | A | A | A | A | A | A | A | A |
| Day 2 | A | A | A | A | A | A | A | A | A | A |
| Day 3 | B | B | B | B | B | B | B | B | B | B |
| Day 4 | B | B | C | C | C | C | C | C | C | C |
| Day 5 |
Ready for the full knowledge check? Guided scenarios and data export. Included with every plan.
PlanTake the advanced knowledge checkFrequently asked questions
What is heijunka?
Heijunka is the Japanese term for production leveling: distributing production volume and product mix evenly over time instead of producing in large batches. Instead of making 100 units of Part A on Monday, 80 of Part B on Tuesday, and 60 of Part C on Wednesday, you make 33A+27B+20C every day. This smooths demand on upstream processes, stabilizes inventory, and creates a predictable rhythm.
What is a heijunka box?
A heijunka box is a physical scheduling device: a grid of slots organized by time period (columns) and product type (rows). Kanban cards are placed in slots to create a leveled production sequence. The operator takes the next card from the box to know what to produce. It converts a weekly batch schedule into an hourly mixed-model sequence.
Why does leveling matter if you have pull?
Pull without leveling creates a 'bullwhip effect': small demand variations at the end customer amplify as they travel upstream. Monday the customer wants 40 parts; Tuesday they want 10; Wednesday 60. Without leveling, each upstream process reacts to these swings. Leveling dampens the variation: regardless of daily customer orders, the factory produces the average daily rate, replenishing a finished goods supermarket.