Heijunka Guide

Step-by-step instructions on how to implement heijunka

Why is production leveling important?

Adjusting capacity is difficult due to:

  • Costs (e.g., hiring/layoff costs for labor)
  • Risks (e.g., liquidity risk of buying equipment if actual volumes are lower than planned)
  • Limitations (e.g., 30- to 90-day terms for Temp labor, or no budget to invest in new equipment)
  • Delays (e.g., long lead times for talent acquisition or equipment delivery)


Reasons for leveling:

  • Eliminate waste of over-burden (e.g., overtime, postponing preventive maintenance) and under-burden (e.g., waiting, unnecessary activities)
  • Avoid purchasing equipment to meet demand peaks
  • Level the demand on supplier companies and processes 
  • The belief it’s easier to manage a leveled workload


Leveling resolves these difficulties at the price of increased inventory and/or longer lead times


Key Point: Leveling is a lean workaround, not a lean principle


Production leveling terms and definitions

Production Leveling is a set of techniques used to avoid the peaks and valleys in production and therefore, capacity:

  1. Determine leveling period: how often you can adjust capacity (e.g., monthly)
  2. Take forecasted demand (black line)
    fluctuating black line representing  demand levels for a manufacturing business
  3. Set capacity to meet average daily demand during one or more leveling periods and produce the same quantity every day of the leveling period
    heijunka leveling production capacity manufacturing lean kaizen


Production Smoothing is the “spreading out” of production batches to make a given product as often as possible, using the smallest batch size possible (ideally batch size 1):

  1. Take the average daily demand by end product
    heijunka bar one to measure average daily demand
  2. Split it into the smallest possible batches (taking into account changeover time and capacity)
  3. heijunka bar two
  4. Spread out the batches as evenly as possible: this is your daily production schedule
    heijunka bar three


Heijunka: the term used in the Toyota Production System for both leveling and smoothing (note: many people have different definitions for these words).


Lot size (or Batch size): the number of units of a given product type that are produced in one production run  (i.e., in between changeovers).


Leveling period: the length of the planning period during which the daily workload will remain at a fixed value (e.g., one month).


EPEI: Every Part Every Interval. The time interval (usually measured in days) it takes to “cycle” through all the product types within a leveled schedule (e.g., 5 days)


Restricted OEE (OEEr): OEE calculated with the changeover times removed from the Planned Runtime and used for calculating EPEI. OEEr= Theoretical time for good parts produced / Planned Runtime minus Changeover times. 


Pacemaker: the single process step in a lean value stream that receives a leveled production schedule (or more precisely, production signals) 


Kanban: a signal coming from the downstream (customer) process to “pull” materials (and therefore, trigger work) from the upstream (supplier) process.


Heijunka box (leveling box): for “smoothing” the kanbans sent to the pacemaker.


Pitch: the rhythm at which the heijunka box sends out signals (e.g., 30 minutes). It is not only the “pulse” of information flowing in the value stream, but also the “pulse” of materials flowing.


Supermarket: a highly standardized inventory serving a process or process step, with assigned locations for the materials and maximum/minimum quantities.


The Bullwhip Effect

When delays (long lead times) and silos (isolated and incomplete information) exist in a system, it results in reinforcing feedback loops. In the supply chain, this is called the bullwhip effect: a small change in demand from final consumers gets amplified as it spreads upstream.

graph showing heijunka bullwhip effect experienced by customer retailer wholesaler distributor factory


One common reason why leveling is adopted is the belief it reduces the bullwhip effect. This is only partially true.


Five tools for leveling production

Leveling isn’t just a change in planning—it’s achieved by reducing lead times (reduces bullwhip effect and increases schedule slack) and using inventory buffers (absorbs daily demand variation)


Here are the Five tools for leveling production and how they work:


Flow cells and Supermarkets

Flow cells reduce lead time thanks to one-piece-flow. Supermarkets buffer for variation in demand and supply.


Quick changeovers & small batch sizes

With faster changeovers we run more but smaller batches without capacity loss, hence lower cycle stock and shorter lead times.


Small containers & more frequent transport

Small containers can be filled/moved more often, thus reducing material wait time after production, and therefore lead time. Small containers increase flexibility. Remember that the kanban formula has “number of parts per container” in the denominator!


Heijunka (leveling) box with Kanban

Spreading production and transport of material removes “bumpiness” of demand and of inventory.


Multi-skilled labor

Operators trained to perform many jobs can be moved as needed, providing added flexibility. This is an area where on-demand labor can be a game-changer.


New reality: Production leveling is unnecessary, but we still need production smoothing

Leveling was one of the earliest methods used by Toyota and has remained unquestioned until now.


The reason leveling was created is that perfectly matching capacity to daily changes in demand was unreasonable:

  • Companies didn’t have the required labor flexibility
  • Lean organizations can’t expect deep engagement of their workforce in continuous improvement if they lay off workers—even if it happens rarely
  • Temp labor isn’t flexible enough to accommodate daily adjustments


The result is that we minimize the under- and over-staffing, but imperfectly, and at the cost of higher inventory and longer lead times.


With Veryable, matching capacity to actual demand on a daily basis is no longer unreasonable. The rationale for leveling disappears. You must continue to use the 5 tools, but no longer need to level production. 


A corollary is that planning is less important when you can adjust your labor to the actual needs, including unplanned needs such as last-minute orders, compensating for defective production, etc.   


Leveling creates inventory. Smoothing reduces inventory. Specifically, smoothing cuts the cycle stock in between Work Centers.

graph showing heijunka smoothing, which reduces inventory and cuts the cycle stock

Every time we reduce changeover time, we use the freed-up time to do more changeovers, thus reducing the average inventory level.

Note that smoothing works best with extremely short changeover times — therefore it often is applied in assembly operations, where changing from one product to another is immediate or measured in seconds.

This in turn creates a quasi-continuous demand for all end product components.

You go from this:

graph showing heijunka smoothing, which reduces inventory and cuts the cycle stock

To this:

When one-piece-flow exists between two processes, supply and demand have the same slope and cycle stock can be eliminated.

When one-piece-flow exists between two processes, supply and demand have the same slope and cycle stock can be eliminated.


Setting batch size with EPEI

What is the best production batch size? The conventional answer is:


EPEI = Time required to complete all changeovers / Time available per day for changeovers


The Lean answer is: as small as possible.


This is where EPEI comes in. EPEI stands for Every Part, Every Interval. With EPEI, we can calculate the smallest possible batch size, taking into consideration OEE losses, changeover times, planned runtime, and product demand.


EPEI = Time required to complete all changeovers / Time available per day for changeovers


For example, if we produce 6 products and the changeover time is 1 hour, the time required to complete all changeovers and “cycle” through the products is 6 hours.


If planned runtime is 21 hours per day, and 18 hours is the pure production time required to meet the daily demand of all 6 products, that leaves 21-18=3 hours available per day for changeovers. EPEI = 6/3 = 2 days, meaning it takes two days to cycle through the six products.


Small containers are key to smoothing

We know from Value Stream Mapping that lead time is 95-99% caused by inventory (i.e., material waiting). Production lead time has various components which are shown in this diagram from the perspective of an average part:

heijunka production lead time

  • Queue time is determined by the production batch size (material is waiting for parts “ahead in the line” to be processed).
  • Value-Add time is determined by the processing time.
  • Wait time is determined by the transport batch size (waiting for the container to be filled, then waiting for a forklift to pick it up).
  • Move time is determined by the distance to the next Work Center and the speed at which the material is moving.

SMED and EPEI are the tools used to reduce queue time. Small containers are the main tool used to reduce wait time. Another is Mixed-Load Transport (tug trains), which increases the frequency of transport. Cell and Focused Factory layout is the tool used to reduce move time. 


Heijunka boxes

The heijunka (or leveling) box stores instructions for one or more pacemakers in the form of kanban cards (cards are product-specific, and in the example shown here, the pacemakers are customer-specific). 

kanban cards are product specific, Each column corresponds to a time

Each column corresponds to a time (e.g., every 10m)


The material handler will remove all cards at the designated time and hand them over to the various cells. Each cell receives a number of cards, indicating what they must produce (and how many) until the next card(s) arrive(s).


A pacemaker is usually the flow cell at the end of the process, but it could also be the finished goods supermarket, in which case the kanban card is a “withdrawal” card instructing the material handler to move parts from the supermarket to Shipping.


Pacemakers are used only when the entire value stream has been changed to flow cells/flow lines and pull systems, and no process step remains that is managed through push.


Procedures for production smoothing

For each Work Center in the value stream:

  1. Use the EPEI calculator to determine the smallest possible batch size.
  2. Compare to a target state, e.g. a target lead time.
  3. Determine what needs to be improved (it could be one or more of the following):
    1. Target changeover time
    2. Target OEEr
    3. Target Cycle Time
    4. Target Planned Runtime
  4. Perform kaizen to reach target(s), then use EPEI to set the new batch size.
  5. Adjust the ERP/MRP parameters and planning procedures.
  6. Create supermarkets sized on the new parameters.
  7. Determine container size and quantity for each loop between Work Centers.
  8. Adjust plant logistics for frequent move between Work Centers (see Milkrun Guide).
  9. Create a Heijunka box. Normally, the box works with a kanban system, however, it can also function with any kind of signal telling the Work Center(s) what to produce and how many.


What’s Next?

As you can see, heijunka entails a lot of activities and the journey can take a long time.


Start with SMED. Reducing production batch size is the most important thing you can do if you’re working with machines, after ensuring machine reliability with TPM (Total Productive Maintenance).


If it’s not clear where to start with SMED, run a Value Stream Mapping event to understand what determines the lead time in your value stream and what the key issues are (see the VSM section).


Like many things in Lean, it is a continuous process…improvement knows no end!


Need help implementing Heijunka? Contact the Lean Center of Excellence!

Fill out this quick form and we'll get back to you within a couple of days with more information on how you can implement Heijunka on the shop floor.