Lean teams use heijunka to stop the peaks and troughs that create overtime, idle time and excess inventory. A steady schedule also makes pull systems and standard work stable.
Where it comes from
Heijunka is a foundation of the Toyota Production System, sitting under just-in-time and jidoka. A heijunka box is the common tool for releasing kanban in a leveled sequence.
At a glance
From weekly blocks to a leveled mix
Swipe sideways to see the whole diagram.
How to apply Heijunka
- 1
Understand demand
Analyze demand by product over several periods. Separate steady runners from low-volume or erratic items.
- 2
Calculate takt time
Divide available time by average demand. Takt sets the pace the leveled schedule must hit.
- 3
Shorten changeovers
Use SMED so the line can switch products often. Leveling mix is impractical while changeovers stay long.
- 4
Build a leveled pattern
Create a repeating sequence that mixes products in proportion to demand. Decide the pitch, or time interval, for releasing work.
- 5
Use a heijunka box
Load kanban into time slots by pitch so work is released at a steady rate. Material handlers pull one slot per interval.
- 6
Buffer and review
Hold a small finished-goods buffer to absorb short-term swings. Review the pattern when demand shifts.
Worked example (illustrative)
A line sells 200 of model A, 200 of B and 100 of C each week, and builds them in blocks: A on Monday and Tuesday, B on Wednesday and Thursday, C on Friday. Stock of A peaks near a full week of demand while C runs short by midweek. The team cuts changeover time and moves to a daily A B A B C pattern of 40 A, 40 B and 20 C. Finished stock of each model now stays near one day of demand, about a fifth of the old peak, and the downstream schedule becomes predictable.
Common mistakes
- Trying to level the mix before reducing changeover times.
- Chasing every short-term order change, which reintroduces the peaks heijunka should remove.
- Leveling final assembly while upstream processes still run in large batches.
When Heijunka is not the right tool
Heijunka needs some repeat demand to level. Highly custom work with no repeating products is hard to level by mix.
A common approach is to split products into runners, repeaters and strangers, then level the runners and repeaters first.
Heijunka box and board
A heijunka box is a board with a grid of slots. Columns are time intervals, set by the pitch, and rows are products. Kanban cards are placed in the slots to show what to make in each interval, in a mixed and level sequence.
A material handler collects the cards for each interval and takes them to the line. Because the box holds the plan, everyone can see whether production is on time just by looking at which slots are empty.
Pitch sets the width of each column: takt time multiplied by the pack quantity. In the illustrative example above, 100 units a day in 450 minutes gives a takt of 4.5 minutes. With 20 units per container, the pitch is 90 minutes, so the box has five slots a day for the A B A B C pattern.
Art Smalley's guide Creating Level Pull describes how leveling, pull and kanban work together.
Leveling volume and mix
Volume leveling makes the same total amount each day, even when orders arrive unevenly. Mix leveling spreads the different products across the day instead of in long batches.
Both need short changeovers, which is why SMED usually comes first.
Volume leveling with a buffer: a worked week
An illustrative week: customers need 600 units over five days, but not evenly. The leveled plan makes 120 a day and starts with 60 units of finished stock:
Mon
0
Tue
30
Wed
0
Thu
60
Fri
60
Bottom row: finished stock left after shipping.
The buffer absorbs the swings while the line runs at one steady rate. Leveling does not mean zero inventory; it means a small, deliberate buffer instead of the large stock that long batches create. Check due dates by product too: if all of one product ships on Tuesday, a leveled plan needs stock or an earlier start to cover it.
Mix leveling: a repeating pattern
Suppose demand splits 50% A, 30% B and 20% C. A ten-slot pattern keeps that ratio without long runs of any one product:
Leveled
Batched instead
Five A, three B and two C, repeated through the day. Decide what one letter means, one unit or one container of ten, and write it on the board; mixing the two up is a common scheduling error.
Unit counts are only part of the picture. If C takes twice as long at one station, two C in a row can overload it even when the hourly total looks steady. Toyota sequences heavy and light jobs alternately for the same reason, so check work content at each station, not only the ratio.
Takt time and pitch
Two numbers turn the leveled plan into a rhythm the floor can follow:
Takt time
Available time divided by customer demand. With 420 minutes and demand of 210 units, takt is 2 minutes per unit.
Pitch
Takt time multiplied by the pack-out quantity. With 10 units to a container, pitch is 20 minutes: one container of work is released every 20 minutes.
Both describe the pace demand requires, not the pace the process can hold. Compare them with real cycle times, changeovers and equipment availability before committing. The takt time calculator works out the first number.
Reading a heijunka box
A heijunka box has a row for each product and a column for each pitch interval. Kanban cards in the slots say what to make and when. An illustrative two-hour window, where each card releases one container of ten and each column is 20 minutes:
This window releases 30 units of A, 20 of B and 10 of C. A material handler takes the cards from one column per interval, so work reaches the pacemaker process in small, steady steps. Post a short legend next to the box:
- Card meaning: the product and quantity each card stands for.
- Who releases: the person who pulls each column on time.
- Late or blocked cards: where they go, so the gap stays visible.
- Missed slots: who resolves one before more work is released.
Don't rearrange cards after the fact to make the box look tidy. A missed slot left in place is the evidence for the next improvement discussion.
Can the line run the pattern? A changeover check
An illustrative check: 420 available minutes a day, 200 units at 1.8 minutes each, and a 20-container pattern with 19 product changes.
Changeovers at 4 minutes
436 of 420 min · 16 over
Changeovers at 2 minutes
398 of 420 min · fits, 22 to spare
At 4 minutes per changeover the pattern needs 436 minutes and cannot run. The options are shorter changeovers, a different sequence or a longer interval between runs of each product. At 2 minutes it fits with 22 minutes to spare, which may or may not cover other losses; only watching the line tells you. Telling operators to follow heijunka does not remove the constraint.
Planning a heijunka pilot
Keep the first trial small enough to watch closely:
- 1
Choose one product family. A group that shares a route through the process. List what is excluded and why, or exceptions will quietly use the pilot's capacity.
- 2
Build the demand picture. Quantities, mix, due dates and unusual orders. Note whether the data is actual orders, a forecast or replenishment of what was used.
- 3
Test the pattern on paper. Run and changeover minutes against available time, plus due dates by product, not only weekly totals.
- 4
Set the buffer and the cards. What stock covers timing differences, and what each card authorizes, in what quantity, to where.
- 5
Run and record. Planned against actual by interval. When a slot is missed, record the reason instead of moving the card.
- 6
Improve one obstacle at a time. If missing tooling keeps breaking the sequence, fix tooling readiness rather than redrawing the board.
Agree one exception rule before starting, such as a named supervisor approving any sequence change and recording the work it displaced. It keeps the cost of urgent insertions visible.
Measuring whether leveling works
Judge the pilot on customers and on the process together:
| Measure | What to record | What it tells you |
|---|---|---|
| Delivery performance | Orders completed by the promised date | Whether customers benefit |
| Schedule attainment | Cards completed in their planned interval | Whether the pattern is feasible |
| Inventory by product | Finished goods and WIP, item by item | Shortages hidden inside a healthy total |
| Changeover time | Actual minutes against the planning figure | A key capacity assumption |
| Exceptions | Expedites and sequence changes, with causes | Where the plan loses stability |
| Replenishment interval | Actual time between runs of each product | Whether the intended frequency happens |
Delivery performance
Orders completed by the promised date
What it tells you: Whether customers benefit
Schedule attainment
Cards completed in their planned interval
What it tells you: Whether the pattern is feasible
Inventory by product
Finished goods and WIP, item by item
What it tells you: Shortages hidden inside a healthy total
Changeover time
Actual minutes against the planning figure
What it tells you: A key capacity assumption
Exceptions
Expedites and sequence changes, with causes
What it tells you: Where the plan loses stability
Replenishment interval
Actual time between runs of each product
What it tells you: Whether the intended frequency happens
If total stock falls but one product keeps missing deliveries, the pilot is not a success yet. Check that product's demand timing, buffer and run frequency. Hitting the pattern only through repeated firefighting is a feasibility problem, not good compliance.
Heijunka misconceptions
Common misreadings:
- 1
Leveling means zero stock. A small, managed buffer bridges production timing and customer demand.
- 2
A repeating pattern is always feasible. The number of products and the changeover time limit how often each can run.
- 3
The board is the improvement. The work is removing what stops the pattern from running.
- 4
The pattern is fixed forever. Level against a chosen demand picture, and revisit it when demand shifts for good, not with every order.
- 5
Every product must fit one sequence. In high-mix work, level the workload by product group or by hours of work instead.
| Concept | Main concern | What it adds |
|---|---|---|
| Heijunka | Volume and mix over time | A steady, repeating schedule |
| Just-in-time | The right item, time and amount | Links production and supply to real need |
| Kanban | Permission to make or move | Signals what to replenish; heijunka paces those signals |
| Line balancing | Work spread across stations | Checks each station can hold the pace |
Guides on this topic
- Line balancing: how to split work so a line meets taktLine balancing explained: takt and cycle time, balance efficiency, the data to collect, yamazumi charts and assignment rules, a worked four-station example, workstation checks, software questions and a pilot plan.Read the guide
- Muri: how to spot and reduce overburden in leanMuri explained: what overburden means, how it differs from muda and mura, the common causes, an observation checklist, six ways to reduce it, examples from factories to software, and how to measure progress without hiding overload.Read the guide
- How to do value stream mapping, step by stepHow to do value stream mapping step by step: walk the process, record data, draw the current state by hand, design the future state, then add software.Read the guide
- One-piece flow: how single-unit flow cuts waiting and WIPOne-piece flow explained: a worked comparison with batches, takt and cycle time, the benefits to check, production and transfer batches, flow and pull, where it fits, the conditions it needs, a pilot plan and how to measure it.Read the guide
Sources
Books and papers show their publication year. Web pages show the month we last checked them.
- 1.Art Smalley, Creating Level Pull (Lean Enterprise Institute, 2004)
- 2.Jeffrey K. Liker, The Toyota Way: 14 Management Principles from the World's Greatest Manufacturer (McGraw-Hill, 2004)
- 3.Yasuhiro Monden, Toyota Production System: Practical Approach to Production Management (Industrial Engineering and Management Press, 1983)