Heijunka

Definition

Heijunka is the practice of leveling production by both volume and product mix over a fixed period of time. Rather than building products in the sequence that customer orders arrive — which creates wild swings in demand on people, equipment, and suppliers — a heijunka system takes total demand over a period (typically a day or a shift) and distributes it into a repeatable, mixed-model pattern.

The purpose is to eliminate mura (unevenness), which Toyota considers the root cause of both muda (waste) and muri (overburden). When production is uneven, some periods are overloaded and others are idle. Equipment is oversized for peaks, people are stressed then bored, and suppliers receive erratic orders they cannot efficiently fulfill. Heijunka removes this variation at the source.

Japanese Origin

The word heijunka (平準化) is composed of three characters:

• 平 (hei) — flat, level, even
• 準 (jun) — standard, level, criterion
• 化 (ka) — the suffix meaning “-ization” or “to make into”

The literal meaning is “the act of making level and standard” — equalization. In ordinary Japanese, the word is not exclusive to manufacturing; it can refer to leveling anything (prices, workloads, terrain). Within TPS, it specifically means leveling the production schedule so that every time period looks as similar as possible to every other time period.

The concept addresses what Toyota calls mura (ムラ, unevenness) — one of the three enemies of efficiency alongside muda (waste) and muri (overburden). Taiichi Ohno argued that mura was the most fundamental of the three: unevenness creates both waste and overburden as inevitable consequences.

History at Toyota

The roots of heijunka trace back to Kiichiro Toyoda, who articulated the just-in-time concept in the 1930s. His 1937 pamphlet on JIT production implicitly required leveling: you cannot deliver the right parts in the right quantity at the right time if the production schedule swings wildly from day to day.

1950s — After near-bankruptcy in 1950 and a painful workforce reduction, Toyota began rebuilding its production system under the direction of Taiichi Ohno. As Ohno developed kanban-based pull systems in the machining shops at the Honsha Plant, he discovered that pull could not function when downstream demand was erratic. A surge of orders for one model would drain kanban for specific parts, while other parts sat idle. This experience made leveling a prerequisite — not an optional refinement — for pull production.

1960s — As kanban was adopted company-wide (completed by 1963 according to Toyota’s 75-year corporate history), leveling became systematized. Toyota’s history records that “production items and production volumes were equalized” to prevent individual processes from being burdened with excess personnel or equipment. The heijunka box emerged during this period as a physical scheduling device on the shop floor.

1965–1970s — When Toyota extended kanban to its supplier base (from 1965 onward), heijunka became even more critical. Suppliers could not respond to wildly fluctuating orders. Toyota’s leveling of its own schedule was, in effect, a prerequisite for asking suppliers to deliver just-in-time. Former North American President Atsushi Niimi later explained that Toyota is fundamentally a build-to-stock replenishment system, not a build-to-order system — and leveling is what makes that replenishment efficient.

2004 — Art Smalley published Creating Level Pull through the Lean Enterprise Institute, the Shingo Publication Award-winning workbook that provides a step-by-step method for implementing leveled pull production based on his experience at Toyota’s Kamigo engine plant.

How It Actually Works

Heijunka operates across multiple planning horizons, not just the shop floor. Most descriptions skip the upstream planning layers and jump straight to the daily level, which gives a misleading picture.

Planning Levels at Toyota

Annual sales plan — Toyota develops an annual forecast of total volume and model mix by region. This drives facility planning, supplier capacity agreements, and workforce planning. It sets the broad parameters within which leveling occurs.

Quarterly plan — The annual plan is refined quarterly based on updated demand signals. Supplier commitments are adjusted and major capacity decisions are confirmed.

Monthly production plan — Each month, Toyota finalizes the production volume and model mix for each plant. This is the level at which takt time is set and staffing is adjusted. When the monthly plan changes, standardized work is updated across every affected line — operators may gain or lose work elements, and the number of workers on a line may change.

Daily/shift-level sequencing — This is where leveling becomes operationally visible. The day’s production is sequenced to distribute models as evenly as possible.

At Toyota’s Assembly Plants

Inside Toyota’s vehicle assembly plants, daily sequencing and leveling is managed by the Assembly Line Control (ALC) system — a computerized production control system, not a physical heijunka box. The ALC system generates the leveled production sequence, manages the order-to-line assignment, and communicates build instructions electronically to each station along the line.

Example: Suppose the plant needs to produce 480 vehicles in a shift: 240 of Model A, 120 of Model B, 80 of Model C, and 40 of Model D. Instead of running all A’s, then all B’s (batch production), the ALC system distributes them in a repeating mixed pattern: A-A-B-A-C-A-B-A-D. Every upstream process and supplier sees a steady, predictable demand pattern rather than surges and droughts.

This software-driven approach has been Toyota’s practice for decades. The leveling logic is the same — distribute every model as evenly as possible across the shift — but the execution is computerized at the assembly level.

At Tier 1 Suppliers and Smaller Operations

The physical heijunka box (平準化ボックス) — a wall-mounted device with rows (one per product type) and columns (one per time increment) — is the classic leveling tool for component production, supplier operations, and smaller-scale processes where a computerized system is not necessary or cost-effective.

In this setting, kanban cards are placed in the box slots. A material handler or “water spider” (mizusumashi) withdraws cards at a fixed interval (the pitch — the time to produce one container of parts) and delivers them to the production line, pacing work. The heijunka box acts as a physical pacemaker, converting a batch of daily orders into a timed sequence of small, mixed withdrawals.

This is the version most commonly described in lean textbooks, but it represents supplier-level practice, not how Toyota’s own assembly plants operate.

Leveling Volume vs. Leveling Mix

There are two dimensions to heijunka, and most companies that attempt it only address one:

• Volume leveling means producing the same total quantity each day (or shift), even if customer orders vary. A finished-goods buffer absorbs daily demand fluctuations. This is the easier of the two.

• Mix leveling means producing every product type every day (or every pitch increment), in a repeating pattern. This is far harder because it requires short changeover times and flexible workers. But it is mix leveling that truly smooths demand on upstream processes and suppliers.

Toyota does both. Most Western companies attempting lean do volume leveling at best and skip mix leveling entirely.

Why Heijunka Is a Prerequisite for JIT

This is the most important and most frequently misunderstood point about heijunka. It is not a nice-to-have optimization. It is the foundation that makes JIT possible.

Without leveling, kanban does not work. If final assembly runs 500 units of Model A on Monday and none on Tuesday, the kanban signals to upstream machining will demand 500 sets of A-specific parts Monday and zero Tuesday. The machining department cannot respond to this — they would need enormous capacity buffers or enormous inventory buffers. Pull collapses into push with extra steps.

Without leveling, standardized work cannot be maintained. If the product mix changes drastically from hour to hour, workers cannot develop the muscle memory and rhythm that standardized work requires. Every hour is a new situation, and quality suffers.

Without leveling, suppliers cannot deliver JIT. Toyota’s daily orders to suppliers follow a leveled pattern — the same mix in roughly the same sequence day after day. This allows suppliers to plan their own production efficiently. Companies that send erratic orders to suppliers and then blame them for delivery failures have usually failed to level their own schedule first.

Taiichi Ohno depicted heijunka as the foundation of the TPS “house” — literally the floor on which the two pillars (JIT and jidoka) stand. This was not metaphorical. Without leveling, the pillars cannot bear weight.

Implementation Guidance

Prerequisites:

• Finished-goods buffer stock. Leveling requires decoupling production from the exact sequence of customer orders. A small, calculated finished-goods inventory absorbs daily demand variation while the plant runs a level schedule. This is not waste — it is the minimum inventory needed to enable leveling. Art Smalley’s work on inventory calculation (cycle stock + buffer stock + safety stock) provides the method.
• Short changeover times. Mix leveling requires frequent changeovers. If changeovers take hours, you cannot produce every model every day. SMED (Single-Minute Exchange of Dies) is typically a prerequisite.
• Reliable processes. A leveled schedule has no slack. If equipment breaks down during a leveled sequence, the entire day’s plan is disrupted. Basic stability (uptime, quality, attendance) must be in place first.

Start here:

• Begin with volume leveling: produce the same total quantity each day
• Calculate your actual inventory requirements (cycle, buffer, safety) and compare to what you carry
• Reduce changeover times on constraint processes to enable more frequent product changes
• When changeovers are short enough, begin leveling the mix within each day
• Build or buy a heijunka box and use it to pace withdrawal at a fixed pitch

Common Mistakes

Leveling volume but not mix. This is the most common failure. A plant produces the same total quantity each day but runs long batches of each product type. Upstream processes still see wildly fluctuating demand by part number. The benefits of leveling are largely lost.

Treating heijunka as optional or advanced. Many lean implementations jump straight to kanban and pull without leveling the schedule first. The pull system then oscillates wildly as demand variation propagates upstream, and the company concludes that “pull doesn’t work here.” It would have worked with a leveled schedule.

Eliminating all finished-goods inventory. Leveling requires a small, calculated finished-goods buffer. Companies that equate “lean” with “zero inventory” remove this buffer and then cannot level. The result is that production chases every order in sequence, creating maximum mura. Taiichi Ohno himself noted that inventory below what the system needs to function is just as problematic as excess.

Building to order and calling it lean. Toyota’s standard production system is replenishment-based (build-to-stock with a leveled schedule), not build-to-order. Many Western companies switch to build-to-order, fail to maintain short lead times, and then blame lean thinking when delivery suffers. As Art Smalley has noted, this confuses two entirely different production models.

Assuming the heijunka box is the only way to level. Many lean implementations focus on the physical heijunka box as if it were the definition of heijunka. The box is a useful tool for suppliers and smaller operations, but Toyota’s own assembly plants use computerized systems (ALC) to achieve the same leveling at scale. The principle — distribute every model evenly across the production period — matters more than the specific mechanism.