Heijunka (平準化, "the act of making level and standard") is the leveling of production volume and product mix over a fixed period of time. The records below define the term, trace it from Kiichiro Toyoda's just-in-time concept through Ohno's pull experiments to its place as a foundation of the Toyota Production System, set out how it actually works across planning horizons, and separate Toyota's practice from the version most lean textbooks describe.
Heijunka is the leveling of both production volume and product mix over a fixed period
Heijunka is the practice of leveling production by both volume and product mix over a fixed period of time, typically a day or a shift. Rather than building products in the sequence customer orders arrive, a heijunka system takes total demand over the period and distributes it into a repeatable, mixed-model pattern. The aim is a production plan in which every time period looks as similar as possible to every other.
Toyota Motor Corporation, Toyota Production System; Art Smalley, Art of Lean — artoflean.com/reference/heijunka
The kanji 平準化 breaks into "flat," "standard," and "-ization"
The word heijunka is composed of three characters: 平 (hei, flat/level/even), 準 (jun, standard/level/criterion), and 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 prices, workloads, or terrain.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Heijunka is the production response to variation, not a pretense that demand is flat
Heijunka is not a slogan for pretending demand is level. It is a production and planning response to variation, constrained by the organization's position in the supply chain, customer behavior, product mix, capacity, lead time, and ability to influence demand. It absorbs, shapes, and responds to variation within system limits rather than denying that variation exists.
Art Smalley, Art of Lean (lei-vault jit/heijunka skill) — artoflean.com/reference/heijunka
Leveling has three dimensions: quantity, type, and sequence
Leveling operates in three distinct dimensions. Quantity leveling smooths total volume over a practical period so daily or shift-level takt logic can operate. Type leveling reduces lot sizes and spreads product types more evenly rather than batching one type for long periods. Sequence leveling produces in a planned sequence that supports downstream work, material availability, and customer need.
Art Smalley, Art of Lean (lei-vault jit/heijunka skill) — artoflean.com/reference/heijunka
Volume mura and mix mura are distinct, and heijunka addresses both
Volume mura is unevenness in total production over time; mix mura is unevenness in the product mix even when total daily volume is stable, such as running all Model A in the morning and all Model B in the afternoon. Both cause surges in the specific resources each model needs. Heijunka addresses both — leveling the total quantity and the product sequence — which is why mix leveling matters even after volume is stable.
Lean Enterprise Institute Lean Lexicon; Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/mura
Heijunka traces to Kiichiro Toyoda's just-in-time concept of the 1930s
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, and the idea began to be put into practice from 1938 as the Koromo plant came on line: parts cannot be delivered in the right quantity at the right time if the production schedule swings wildly from day to day. Leveling was latent in the original JIT idea before it was named and systematized.
Toyota Motor Corporation, Toyota 75-Year History — artoflean.com/reference/heijunka
Ohno discovered leveling was a prerequisite for pull, not an optional refinement
As Ohno developed kanban-based pull in the Honsha Plant machining shops in the 1950s, he discovered that pull could not function when downstream demand was erratic. A surge of orders for one model would drain the kanban for specific parts while other parts sat idle. This experience made leveling a prerequisite — not an option — for pull production.
Toyota Motor Corporation, Toyota 75-Year History; Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Ohno reached leveling through trial and error with the supermarket idea
Ohno took the kanban concept from the American supermarket he saw on a 1956 U.S. visit: the customer takes what is needed and the store replenishes only what was withdrawn. The first major problem was that when the following process withdrew a large amount of the same part at once, it threw the preceding process into disorder. After much trial and error, the answer was leveled production.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Toyota's "dekansho production" was the month-end-push problem leveling solved
Before leveling, assembly could never be done in the first half of the month; parts arrived irregularly and had to be assembled in a rush at month-end. Ohno called this "dekansho production," a pun on a folk song about sleeping half the year away. The first internal task was to average and level the month-end push, then to take leveled production outward to cooperating firms.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Leveling became systematized as kanban spread company-wide by the early 1960s
As kanban was adopted across Toyota — completed company-wide by 1963, per Toyota's corporate history — leveling became systematized. Toyota's corporate history records that production items and production volumes were equalized to keep 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.
Toyota Motor Corporation, Toyota 75-Year History — artoflean.com/reference/heijunka
Extending kanban to suppliers from the mid-1960s made leveling more critical
When Toyota extended kanban to its supplier base from 1965 onward, heijunka became even more critical, because 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. Sending erratic orders to suppliers and then blaming them for delivery failures is usually a sign the customer failed to level its own schedule first.
Toyota Motor Corporation, Toyota 75-Year History — artoflean.com/reference/heijunka
Sugimori's 1977 paper documented leveling as a published prerequisite for JIT
The 1977 paper by Sugimori, Kusunoki, Cho, and Uchikawa — production-control engineers at Toyota — states the rule directly: if the quantity withdrawn by subsequent processes varies considerably, processes and subcontractors must keep peak capacity or excess inventory at all times. Therefore, to make just-in-time possible, the prerequisite is to level production at the final assembly line. The degree of this leveling is determined by top management.
Sugimori, Kusunoki, Cho & Uchikawa, "Toyota production system and Kanban system," Int. J. Production Research (1977) — artoflean.com/reference/heijunka
The degree of leveling at Toyota is a management decision, not a shop-floor one
Sugimori's paper states plainly that the degree of leveling is determined by top managers. Sugiura's Global Ten material likewise shows production volumes as a president-level decision tied to domestic sales, export volume, and trade-friction considerations, with monthly model volumes leveled at assembly lines and special handling for export vehicles that had to be loaded on ships in time. Leveling is a real planning system reaching to the top of the company, not merely a box on the floor.
Sugimori, Kusunoki, Cho & Uchikawa (1977); Mikio Sugiura, Global Ten — artoflean.com/reference/heijunka
A leveling buffer is sized as cycle stock plus buffer stock plus safety stock
The finished-goods buffer that lets production stay level while demand varies is sized from three explicit components: cycle stock (normal replenishment over the order interval), buffer stock (absorbs demand swings), and safety stock (covers internal disruptions). Calculating these deliberately — rather than holding arbitrary inventory — is what makes leveling against uneven demand work.
Art Smalley, Creating Level Pull (Lean Enterprise Institute, 2004) — artoflean.com/reference/heijunka
Heijunka operates across multiple planning horizons, not just the shop floor
Leveling is not only a shop-floor device. At Toyota it operates across an annual sales plan that sets facility, supplier, and workforce parameters; a quarterly refinement of that plan; a monthly production plan that finalizes volume and model mix per plant and sets takt time; and daily or shift-level sequencing where leveling becomes operationally visible. Most descriptions skip the upstream layers and jump straight to the daily level, which gives a misleading picture.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Toyota's information system sends the leveled sequence plan to one place only
Toyota issues a January "preliminary notice" and a February "confirmation" of March production down to model, type, and detail, with the same notice and confirmation sent to outside cooperators at the same times. The daily schedule is then finely worked out with leveling thoroughly woven in, and the further-leveled sequence plan is sent to a single location: the head of the final assembly line. Ohno called this a great feature of the Toyota information system.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Monthly volume averaged over working days sets the daily leveled quantity
Sugimori's account describes the method precisely: production per day is averaged by taking the number of vehicles in the monthly production schedule, classified by specification, and dividing by the number of working days. For the daily sequence, the cycle time of each specification vehicle is calculated so that each specification appears at its own cycle time, with different specification vehicles ordered to follow one another.
Sugimori, Kusunoki, Cho & Uchikawa (1977) — artoflean.com/reference/heijunka
A leveled mixed sequence interleaves models rather than batching them
Instead of running all of Model A, then all of Model B, a leveled plant distributes models in a repeating mixed pattern. For a shift of 480 vehicles split 240 A, 120 B, 80 C, and 40 D, the sequence interleaves them — for example A-A-B-A-C-A-B-A-D — so every upstream process and supplier sees a steady, predictable demand pattern rather than surges and droughts.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Quantity leveling can come early; type and sequence leveling usually develop later
Quantity leveling may be a precondition for basic daily takt logic and is often needed early to establish a workable daily volume. Type and sequence leveling typically develop over time as changeover, stability, and scheduling capability improve. Treating leveling as a single all-or-nothing step misreads how the capability is actually built.
Art Smalley, Art of Lean (lei-vault jit/heijunka skill) — artoflean.com/reference/heijunka
Leveling requires a calculated finished-goods buffer, not zero inventory
Leveling requires decoupling production from the exact sequence of customer orders, which a small, calculated finished-goods inventory makes possible by absorbing daily demand variation while the plant runs a level schedule. This buffer is not waste; it is the minimum inventory needed to enable leveling. Companies that equate lean with zero inventory remove this buffer and then cannot level — Ohno himself noted that inventory below what the system needs is as problematic as excess.
Art Smalley, Art of Lean; Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Three prerequisites enable leveling: finished-goods buffer, short changeover, reliable processes
Mix leveling has three practical prerequisites. A small finished-goods buffer decouples production from the raw order sequence. Short changeover times — typically via SMED — allow producing every model every day. Reliable processes are needed because a leveled schedule has no slack: an equipment breakdown during a leveled sequence disrupts the entire day's plan. Basic stability in uptime, quality, and attendance must be in place first.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Shorter changeover is the enabler that makes smaller lots and mix leveling practical
Shorter changeover makes smaller lots practical, and smaller lots make JIT easier by reducing waiting, improving responsiveness, supporting type and sequence leveling, and exposing problems sooner. Quantity leveling may be possible before advanced type and sequence leveling, but type and sequence leveling usually require progressively smaller lots, which often require changeover improvement first.
Art Smalley, Art of Lean (lei-vault jit/changeover-and-lot-size skill) — artoflean.com/reference/smed
A practical starting sequence: level volume first, then attack changeover, then level mix
The recommended path is to begin with volume leveling — producing the same total quantity each day — then calculate actual inventory requirements (cycle, buffer, safety) and compare to what is carried, then reduce changeover times on constraint processes to enable more frequent product changes. When changeovers are short enough, begin leveling the mix within each day, using a heijunka box to pace withdrawal at a fixed pitch.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Do not demand mixed-model leveling while changeover stays long and unstable
A leveling program must not run ahead of the capability that supports it. Do not demand mixed-model leveling while leaving changeover long and unstable; conversely, do not treat long changeover as a permanent law without studying the actual work. Mix leveling cannot be applied where changeover, quality, material supply, or equipment stability cannot yet support it.
Art Smalley, Art of Lean (lei-vault jit/changeover-and-lot-size and jit/heijunka skills) — artoflean.com/reference/heijunka
A leveled plan needs an abnormality-response rule for when it cannot be followed
Treating the leveled plan as decoration while expediting runs the real system is a core failure mode. A genuine leveling system includes response rules: what happens when the leveled plan cannot be followed, who responds, and how the plan is recovered. Without an abnormality response, leveling is theater and the real scheduling reverts to firefighting.
Art Smalley, Art of Lean (lei-vault jit/heijunka skill) — artoflean.com/reference/abnormality-management
The physical heijunka box is a supplier and small-operation tool, not how Toyota assembly works
The physical heijunka box (平準化ボックス) — a wall-mounted device with rows for product types and columns for time increments — is the classic leveling tool for component production, supplier operations, and smaller-scale processes where a computerized system is unnecessary. Kanban cards are placed in its slots and a material handler withdraws them at a fixed pitch, pacing work. This is the version most lean textbooks describe, but it represents supplier-level practice, not Toyota's assembly plants.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
The heijunka box acts as a physical pacemaker converting daily orders into timed mixed withdrawals
In a supplier setting, the heijunka box converts a batch of daily orders into a timed sequence of small, mixed withdrawals. A material handler or mizusumashi (water spider) withdraws kanban cards from the box at a fixed interval — the pitch, the time to produce one container of parts — and delivers them to the line. The mechanism matters less than the principle: distribute every model evenly across the production period.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Toyota's own assembly plants level by computer (ALC), not by a physical heijunka box
Inside Toyota's vehicle assembly plants, daily sequencing and leveling is managed by the Assembly Line Control (ALC) system, a computerized production-control system rather than a physical heijunka box. ALC generates the leveled sequence, manages order-to-line assignment, and communicates build instructions electronically to each station. This software-driven approach has been Toyota's assembly-level practice for decades.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Leveling volume without leveling mix is the most common failure
Most companies that attempt heijunka address only volume. Volume leveling produces the same total quantity each day but may still run long batches of each product type, so upstream processes still see wildly fluctuating demand by part number, and most of the benefit is lost. Mix leveling — producing every product type every day in a repeating pattern — is far harder, requiring short changeovers and flexible workers, but it is mix leveling that truly smooths demand upstream. Toyota does both.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Leveling does not mean freezing demand or hiding it behind an average
A leveled plan still adjusts to the market. Sugimori's account states that even after the monthly schedule is set, Toyota makes changes among specification vehicles on daily orders and revises the monthly schedule to reduce the shock of market fluctuation. Leveling to an average that ignores real daily or seasonal demand patterns is a failure mode, not the method — heijunka shapes variation within a range, it does not pretend variation away.
Sugimori, Kusunoki, Cho & Uchikawa (1977); Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Installing a heijunka box without volume, type, sequence, pitch, and response logic is artifact copying
A common failure mode is artifact copying: installing a heijunka box without clear volume, type, sequence, pitch, and response logic behind it. Other failure modes include leveling to an average that hides real variation, expecting lower-tier suppliers to absorb variation the OEM refuses to level, claiming type leveling is impossible without studying changeover, and treating the leveled plan as decoration while expediting runs the real system.
Art Smalley, Art of Lean (lei-vault jit/heijunka skill) — artoflean.com/reference/heijunka
A company's leveling ability depends on its position in the supply chain
Toyota has some control over sales and production planning because it sits at the top of its production pyramid. A Tier 2 supplier, a seasonal business, or an organization in an industry that does not level demand may struggle to apply Toyota-style JIT math directly. That does not make JIT thinking useless: the organization must shorten order-to-delivery lead time, reduce waste, design appropriate buffers, and level what it can control rather than copying Toyota artifacts.
Art Smalley, Art of Lean (lei-vault jit/heijunka skill) — artoflean.com/reference/heijunka
Seasonal and promotion-driven demand may require consciously designed buffers
Seasonal products, outdoor gardening supplies, promotion-driven retail, healthcare demand spikes, and long-lead supply chains may require capacity, inventory, or planning buffers. The real question is whether those buffers are consciously designed or merely accumulated through poor synchronization. Heijunka thinking applies even where Toyota's automotive leveling math does not, by forcing buffers to be deliberate rather than accidental.
Art Smalley, Art of Lean (lei-vault jit/heijunka skill) — artoflean.com/reference/heijunka
Leveling, not building to order, is what makes Toyota's replenishment system efficient
Toyota's standard production system is replenishment-based — build-to-stock with a leveled schedule — not build-to-order. Former North American president Atsushi Niimi explained that Toyota is fundamentally a build-to-stock replenishment system, and leveling is what makes that replenishment efficient. Many Western companies switch to build-to-order, fail to maintain short lead times, and then blame lean thinking when delivery suffers — confusing two entirely different production models.
Art Smalley, Art of Lean — artoflean.com/reference/pull-system
Toyota's Tsutsumi plant alternated Corona and Carina on the same line rather than batching by half-day
Ohno gives a concrete leveling example: at the Tsutsumi plant, two lines made the Corona, Carina, and Celica. On one line Corona and Carina alternated unit by unit rather than running Corona all morning and Carina all afternoon, specifically to maintain leveling. This is leveling of type and sequence on a real assembly line, not an abstract scheduling rule.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
A worked Corona example: 5,000 sedans, 2,500 hardtops, 2,500 wagons leveled to a daily mix
Ohno's worked example: to make 10,000 Coronas over 20 working days — 5,000 sedans, 2,500 hardtops, and 2,500 wagons — means 250 sedans, 125 hardtops, and 125 wagons per day. On the line this runs a sedan every other unit and a hardtop and wagon roughly every third, minimizing both lot size and production variation. This is mix leveling reduced to an exact daily sequence.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Leveling forced Toyota's press changeover from hours to minutes
Leveling was at first a harsh demand on the press shop, where stamping as continuously as possible from one die had long been common sense. To match the leveled variety, presses had to perform frequent changeovers, and the watchword became "small lots, swift changeover." Toyota's internal press changeover, which took 2 to 3 hours in the late 1940s, was cut to about 3 minutes by the late 1960s.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Sugimori documented a 10-minute press setup against Western times of 4 to 6 hours
Sugimori's 1977 paper reports that Toyota reduced setup time on an 800-ton press line stamping hoods and fenders to 10 minutes, down from about 1 hour, with improvements made since 1971. The paper's comparison table shows Western plants at the time taking 4 to 6 hours for the same class of setup, with Toyota changing over 3 times a day to a 1-day-use lot size versus competitors' 10-day or 1-month lots.
Sugimori, Kusunoki, Cho & Uchikawa (1977) — artoflean.com/reference/smed
Toyota's working-asset turnover ran far above Western competitors in Sugimori's 1977 data
Sugimori's 1977 paper reports Toyota's turnover ratio of working assets at 41 in 1960, 66 in 1965, and 63 in 1970, against major U.S. competitors holding steady at roughly 5 to 13 over the same period. Sugimori attributes this asset efficiency to the Toyota production system as a whole — the minimal inventory that pull-based, small-lot production permits — of which leveled production is one part.
Sugimori, Kusunoki, Cho & Uchikawa (1977) — artoflean.com/reference/inventory
Toyota's final car plants leveled enormous mixed volume — over 200,000 cars a month
Ohno noted that Toyota's finished-car lines turned out over 200,000 cars a month in countless varieties, with market diversification appearing directly as product diversification that thinned the old mass-production effect. Leveling was Toyota's means of holding cost down under exactly the high-variety conditions that defeat batch mass production. This is the scale at which Toyota's leveling system operated.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
The purpose of heijunka is to eliminate mura, the root cause of muda and muri
The purpose of leveling 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 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.
Toyota Motor Corporation, Toyota Production System — artoflean.com/reference/mura
Mura, not the customer's demand, is what heijunka attacks
Customer demand naturally varies. Mura is not the customer's unevenness — it is the production system's failure to absorb that unevenness. Heijunka does not attempt to make customers buy evenly; it creates an internal leveling mechanism so external variation does not propagate into the factory. The trade-off is a small increase in finished-goods inventory in exchange for eliminating mura throughout the system.
Taiichi Ohno, Toyota Production System (1978); Lean Enterprise Institute Lean Lexicon — artoflean.com/reference/mura
In general, mura drives muri and muda — though not always in one fixed order
The three categories are connected, and a common path runs mura to muri to muda, though the order is not always strict. Mura (unevenness) in the schedule creates peaks and valleys. During peaks, people and machines are pushed beyond capacity, which is muri (overburden); muri then often causes breakdowns, defects, safety incidents, and rework, which is muda. During valleys, people and machines sit idle, which is also muda. This is the general reason Toyota addresses mura first through heijunka rather than attacking visible waste directly.
Art Smalley, Art of Lean, building on Ohno's muda/mura/muri framework — artoflean.com/reference/muri
Heijunka is the countermeasure to mura, and reducing mura relieves muri and muda in turn
Heijunka is aimed directly at mura — unevenness in volume and mix. Because much of the muri (overburden) and muda (waste) on the floor is generated by that unevenness, leveling reduces them as downstream effects rather than attacking each separately: smoother demand means people and machines are neither overburdened at the peaks nor idle in the valleys, and upstream processes and suppliers see a stable signal. The leverage is in treating mura as the root rather than chasing the waste it produces.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
"Lower the peaks, shallow the valleys" — Ohno's image for leveling
To uphold the rule that the preceding process produces only what the following withdrew, every process must be staffed and equipped to produce when and as needed. The greater the variation in withdrawals, the more excess people and equipment the preceding process must hold, and variation at the final process propagates upstream. Ohno's countermeasure was to knock down the production peaks at the final car line and shallow the valleys, calming the surface of the flow — production leveling.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Unevenness amplifies as it travels upstream, which is why leveling starts at final assembly
Sugimori's paper notes that in a multistage industry the demand for a part becomes progressively more erratic the further the process is removed from the original finished-goods demand, forcing preceding processes to hold surplus capacity and making overproduction more likely. Leveling at the final assembly line — the one process that knows true timing and quantity — keeps this amplification from starting, so leveling there levels all the subsequent-withdrawal processes too.
Sugimori, Kusunoki, Cho & Uchikawa (1977) — artoflean.com/reference/heijunka
Staffing to peak demand is a structural muda that leveling removes
Organizations that do not address mura end up staffing and equipping for the worst-case surge, creating permanent overcapacity during normal periods. This is a structural form of muda, invisible because it is built into the operating model rather than appearing as a visible shop-floor waste. Leveling removes the peaks that force peak staffing, so the same total volume is produced with less equipment and a more stable headcount.
Lean Enterprise Institute Lean Lexicon; Sugimori et al. (1977) — artoflean.com/reference/mura
Leveled production lets Toyota run with less equipment and a more stable workforce
Sugimori's paper states that leveled production operates with smaller production changes than scheduled production, so it can be done with less equipment capacity and a more stable number of workers — a point the paper notes is especially important for Japanese companies with lifetime employment. Leveling thus has a direct workforce-stability payoff, not only an inventory one.
Sugimori, Kusunoki, Cho & Uchikawa (1977) — artoflean.com/reference/heijunka
TPS was built for high-variety, low-volume conditions that batch mass production cannot serve
Ohno framed TPS, and leveling within it, as the answer to Japan's postwar market constraint of high-variety, low-volume production. The Maxcy-Silberston curve — cost falling sharply with volume — held in the high-growth era, but the era of maximizing lot size to chase that effect had ended, and that method generates every kind of waste. Leveling lets a plant make cheaply at high variety and low volume, where building large batches by model cannot.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Leveling was fundamental to Ohno's TPS, though the "house" diagram came later
Ohno did not draw the TPS "house" — he did not work in diagrams. But in building the Toyota Production System through the 1950s and 1960s he treated leveling as fundamental: without it, just-in-time and jidoka cannot hold. The familiar "house" diagram, with heijunka as the floor beneath the two pillars, is a later depiction codified by others at Toyota. It captures the role leveling actually played in Ohno's practice, even though the image itself is not his.
Taiichi Ohno, Toyota Production System (1978), on leveling's role in TPS; the TPS "house" diagram is a later codification — artoflean.com/reference/tps-house
Heijunka is what makes pull systems and kanban actually function
Leveling is the precondition that lets pull systems and kanban work: a pull system run against wildly uneven production oscillates and fails. By smoothing the demand each upstream process sees, leveling lets kanban hold a stable number of cards and lets the supermarket absorb withdrawal without surging. Most Western JIT failures trace to attempting pull and flow without leveled production.
Art Smalley, Art of Lean; Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/kanban
Kanban performs fine adjustment automatically within a leveled range
Ohno noted that because no detailed plan is given to each line in advance, the next model is unknown until a kanban comes off, yet processes simply make what the kanban specifies even when the actual ratio reverses an expected mix. This automatic fine adjustment within the leveled range is kanban's great merit; leveling sets the band within which kanban absorbs day-to-day variation without anyone issuing change orders.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/kanban
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 becomes a new situation and quality suffers. Leveling provides the stable, repeating pattern of work content that lets standardized work hold across a shift.
Art Smalley, Art of Lean — artoflean.com/reference/heijunka
Toyota's "small lots, swift changeover" watchword exists to serve leveling
The reversed-thinking watchword "small lots, swift changeover" arose directly from leveling. As the final assembly process shrank lots to level the mix, the upstream press shop had to follow with frequent changeovers rather than long uninterrupted runs from one die. Changeover reduction is therefore not a trophy in itself but an enabler of the leveled mix.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/smed
"Dedicated equipment with generality added" lets leveling and product variety coexist
Ohno held that market diversification and production leveling cannot harmonize from the start; the key equipment measure to reconcile them is dedicated equipment with generality added. If one Corolla line can assemble sedan or coupe in any order, mix leveling becomes possible. By adding minimal equipment and jigs, a dedicated, low-cost process gains the flexibility leveling needs without surrendering the volume effect.
Taiichi Ohno, Toyota Production System (1978) — artoflean.com/reference/heijunka
Heijunka, takt time, flow, and pull form one integrated JIT chain
Leveling is one link in JIT's operating chain: stability, then flow, then takt time, then pull, then leveling as the system matures. The elements are interdependent — takt relates pace to demand, flow moves work continuously, pull authorizes from real need, and leveling smooths quantity, type, and sequence so the others can operate. Removing any one degrades the system; leveling cannot be bolted on without the stability and flow beneath it.
Art Smalley, Art of Lean (lei-vault jit/overview skill) — artoflean.com/reference/just-in-time
Level production before establishing flow and pull, but after stability
In the JIT implementation sequence, stabilize first — reliable equipment, consistent quality, standardized work — then level production, because leveling creates the stable demand pattern everything else requires. Only then establish flow where possible, implement pull where flow is not yet possible, synchronize to takt, and continuously reduce buffers. Skipping heijunka is the most common cause of Western JIT failure, because it amplifies demand variation instead of damping it.
Toyota Motor Corporation, Toyota Production System; Art Smalley, Art of Lean — artoflean.com/reference/just-in-time