New to this topic?
We recommend reading these guides first to get the most out of this one:
Push
Forecast → Produce
Pull
Consume → Replenish
#1
Overproduction = Push's Gift
WIP ↓
Pull's Guarantee

The Core Difference

Every production system is either push, pull, or a hybrid. The distinction is simple but the consequences are enormous:

PushPull
TriggerSchedule / forecast tells you to produceActual consumption signals you to replenish
Question"What does the plan say to make?""What has been consumed?"
WIP ControlNone — WIP is a consequence of the scheduleBuilt-in — WIP is capped by number of kanban cards
OverproductionCommon — forecast errors, batch logic, "just in case"Structurally prevented — no signal, no production
Lead TimeLong — driven by WIP accumulationShort — WIP is low, material flows fast
InventoryHigh — buffers everywhere to cover forecast uncertaintyLow — only enough to cover replenishment time

Little's Law Makes It Concrete

Lead Time = WIP ÷ Throughput. Push systems accumulate WIP because there is no mechanism to limit it. More WIP = longer lead time. Pull systems cap WIP, which directly caps lead time. Use the WIP calculator to see the math.

How Push Systems Work

In a push system, a central planning function (typically MRP/ERP) generates a production schedule based on forecasts, customer orders, and inventory targets. Each work center receives a schedule telling it what to produce and when. Material is pushed to the next step whether it is ready for it or not.

MRP Schedule
push →
Cut
push →
WIP 📦📦📦
push →
Weld
push →
WIP 📦📦
push →
Paint
Push: MRP drives each step. WIP accumulates between steps because there is no limit mechanism.

How Pull Systems Work

In a pull system, nothing is produced until the downstream process (or customer) consumes something. That consumption triggers a signal (kanban) to the upstream process: "I used one, make one." The system is self-regulating — production automatically matches consumption.

Cut
← kanban
Supermarket
pull →
Weld
← kanban
Supermarket
pull →
Paint
Pull: downstream consumption triggers upstream production. Kanban signals flow backward. WIP is capped.

The Hybrid Reality

Most real-world operations are a hybrid. Few factories are 100% pull (some long-lead raw materials still need forecast-based ordering) or 100% push. The practical question is: where in your value stream can you switch from push to pull?

StrategyPush PortionPull PortionBest For
Make-to-Stock (MTS)Forecast drives production to FG bufferCustomer pulls from FG bufferHigh-volume, stable demand, short customer lead time expectations
Make-to-Order (MTO)Long-lead materials ordered to forecastFinal production triggered by customer orderCustom products, variable demand
Assemble-to-Order (ATO)Components built to forecast / pullFinal assembly triggered by customer orderConfigurable products (e.g., electronics, vehicles)
One boundary, three positions Push and pull are the two ends nobody occupies of a single line through one value stream. Make-to-stock, assemble-to-order and make-to-order are three positions of the same boundary — and the five-step transition is that boundary walking upstream.

The page's hybrid table and its transition list, drawn on one ordinal axis. The stream is the page's own sequence of stations: step 4 gives "assembly pulls from machining, machining pulls from raw material store", step 2 gives the pacemaker as "typically final assembly or packaging", and the table supplies "FG buffer", "Components" and the customer. The three cuts are the table read as positions rather than as species — make-to-stock has "forecast drives production to FG buffer" and "customer pulls from FG buffer", so the cut is at the FG buffer; assemble-to-order has "components built to forecast" against "final assembly triggered by customer order", so the cut is at the component buffer; make-to-order has "long-lead materials ordered to forecast" against "final production triggered by customer order", so the cut is at raw material. The two ghost lanes are the page's own sentence that "few factories are 100% pull ... or 100% push". The counts are arithmetic on this axis and nothing else: the five stream positions before the customer are split by the cut into 5 and 0, 3 and 2, 1 and 4 — the same five stages counted twice, which is why one bracket shortens exactly as the other lengthens. The axis is a PLACE in the value stream, not a duration. This page carries no worked example and no magnitudes, so no lead time, inventory value or day count appears here, and the lane order is where the cut sits, not how long anything takes.

Transitioning from Push to Pull

Stabilize the process firstPull exposes problems ruthlessly. If equipment is unreliable, quality is erratic, or changeovers are long, fix those with TPM, poka-yoke, and SMED before removing the WIP buffers that hide them.
Start at the pacemaker processThe pacemaker is the process closest to the customer that sets the rhythm. Implement pull here first — typically final assembly or packaging. Use a heijunka box to level the schedule.
Install supermarkets between processesWhere you cannot create continuous flow, install a kanban-controlled supermarket. Downstream pulls from the supermarket; upstream replenishes it. The supermarket is a controlled buffer — not the WIP chaos of push.
Extend pull upstreamOnce the pacemaker is stable in pull, extend the system backward: assembly pulls from machining, machining pulls from raw material store. Each step gets its own kanban loop.
Reduce kanban cards over timeAs processes improve (shorter changeovers, better quality, more reliable equipment), reduce the number of kanban cards. This lowers WIP, shortens lead time, and exposes the next problem to solve. See kanban.
✅ Pull Advantages
  • WIP structurally limited — cannot explode
  • Lead time short and predictable
  • Overproduction prevented by design
  • Problems exposed immediately (no inventory to hide behind)
  • Self-regulating — adjusts to demand changes automatically
❌ Push Pitfalls
  • WIP grows unchecked during disruptions
  • Lead time long and variable
  • Overproduction is the default behavior
  • Problems hidden under inventory for weeks
  • Requires constant manual schedule adjustments

🎯 Key Takeaway

Push means producing to a plan. Pull means producing to consumption. Pull is structurally superior because it limits WIP, prevents overproduction, and shortens lead time by design. But it requires stable processes as a foundation. Start by stabilizing your bottleneck, then install pull at the pacemaker process, and extend upstream over time. The transition from push to pull is the transition from managing chaos to engineering flow.

Interactive Demo

Compare push vs pull production systems. See how WIP, lead time, and overproduction differ.

⚡
Try It Yourself
Push vs Pull Production
▼
Compare push (forecast-driven, make-to-stock) with pull (demand-driven, make-to-order). Watch how WIP accumulates differently in each system.
Material pushed forward based on forecast
0
Raw Material
→
0
Fabrication
→
0
Assembly
→
0
Shipping
Completed: 0 units | Time: 0 ticks
Push: Each stage produces based on forecast and pushes output downstream. WIP accumulates at slower stages, causing longer lead times and overproduction waste.
0 units
Total WIP
0.0 ticks
Avg Lead Time
0 events
Overproduction
0
Completed
Ready for the full knowledge check? Test your understanding with guided scenarios and data export.
PROTake the Pro Knowledge Check →
Free forever · Every feature included

Stop reading, start modeling

Model your process flow, run simulations, optimize staffing with TOC math, and test your knowledge with 107 interactive checks — all in one platform.

Open Workbench → Pro Knowledge Check

Take this to a room

The running order

For planners and production control. They should leave able to say which parts of their system are push, which are pull, and what has to be stable before pull can start.

7 beats · 9 min
  1. 1

    One question tells you which you have

    Does production start because a plan said so, or because something was consumed?

    • Push asks: what does the plan say to make?
    • Pull asks: what has been consumed?
    • Everything else about the two systems follows from that difference.

    Ask the room Which question does our floor answer each morning?

  2. 2

    Push has no WIP brake

    That is the structural problem, not the forecasting.

    • In push, WIP is a consequence of the schedule - nothing caps it.
    • In pull, WIP is capped by the number of cards. It cannot explode.
    • Lead time equals WIP over throughput, so capping WIP caps lead time.
  3. 3

    Overproduction is push's gift

    It is not a discipline failure. Push produces it structurally.

    • Forecast error, batch logic, and just-in-case all point the same way.
    • In pull there is no signal, so there is no production. It is prevented by design.
    • This is why overproduction is waste number one.
  4. 4

    Problems hide under inventory

    The buffer that makes push feel safe is the same buffer that keeps you from seeing anything.

    • A quality problem can run for weeks behind a pile.
    • Pull exposes problems immediately, which is uncomfortable and correct.
    • That is also why pull needs stability first.

    Ask the room What problem here is currently being absorbed by a pile?

  5. 5

    Everyone is a hybrid

    Nobody is purely one or the other, and pretending otherwise makes the conversation useless.

    • Make-to-stock: forecast drives production to a buffer, the customer pulls from it.
    • Make-to-order: long-lead material to forecast, final production on the order.
    • Assemble-to-order: components to forecast, final assembly on the order.
    • The question is where the boundary sits, not which system you are.
  6. 6

    The transition, in order

    Five steps, and doing them out of order is the standard way this fails.

    • Stabilise first - TPM, error-proofing, SMED. Pull removes the buffers that were hiding those.
    • Start at the pacemaker, the process closest to the customer.
    • Put supermarkets where you cannot make continuous flow.
    • Extend upstream one loop at a time, then reduce cards.
  7. 7

    What we do next

    Find the pacemaker and stabilise it, then put one supermarket in.

    • Name the process that sets our rhythm today.
    • Check its reliability, quality and changeover time honestly.
    • One kanban loop between it and the step upstream.
    • Reduce cards only once that loop is boring.

    Ask the room Is our pacemaker stable enough to remove its buffer?