Lean teams use pull because it caps inventory, shortens lead time and makes problems visible. Where possible they go further, to continuous flow, where items move one at a time from step to step without waiting.
Where it comes from
Pull is part of just in time in the Toyota Production System. Taiichi Ohno described how the American supermarket, where shelves are restocked only as customers take goods, inspired Toyota's kanban system.
At a glance
Push vs pull
Swipe sideways to see the whole diagram.
How to apply Pull system
- 1
Map the flow
Draw the value stream and mark where items wait. Look for steps that could be joined into continuous flow, one piece at a time.
- 2
Create flow where you can
Place connected steps next to each other and balance the work to takt time, so items move without waiting.
- 3
Pull where you cannot flow
Between steps that cannot flow, add a supermarket with set quantities, a FIFO lane with a maximum length, or a CONWIP limit on the whole loop.
- 4
Size the loops
Set the number of kanban cards or the maximum quantity from demand, replenishment time and a safety margin, and write it down.
- 5
Schedule one point
Send the customer schedule to a single pacemaker process. Everything upstream is pulled by it.
- 6
Reduce the buffers
Once the loop is stable, remove cards one at a time to expose the next problem, and fix it.
Worked example (illustrative)
A window maker used to cut frame profiles on its saw to a weekly schedule, so about 3 days of cut parts sat in front of assembly, often in lengths nobody needed that week. It now keeps a small supermarket beside assembly, with 2 bins of each fast-moving profile. When assembly empties a bin, the bin goes back to the saw as the signal to cut more. Cut parts waiting fall to about half a day, and the saw cuts only what assembly has used.
Common mistakes
- Adding kanban cards without sizing them, so the loop holds as much stock as the old push system.
- Scheduling several processes instead of one pacemaker, so push and pull fight each other.
- Starting pull on an unstable process, where breakdowns empty the supermarket and stop the line.
When Pull system is not the right tool
Pull needs fairly steady demand and reliable processes. For one-off custom work, a sequenced pull with FIFO lanes or a CONWIP limit often works better than supermarkets.
Supermarkets, FIFO lanes and CONWIP
A supermarket holds a small, fixed stock of each part. When the next step takes one, the empty space or card tells the supplying step to make another.
A FIFO lane holds a maximum number of items in order. When it is full, the step feeding it stops. CONWIP caps total work in progress across a loop, releasing new work only when finished work leaves.
Sizing a pull loop: a worked example
Here is an illustrative pull loop in a hospital ward's supply room. The ward uses about 40 packs of dressings a day. Stores take 2 days to refill, and the team adds a 25% safety factor. Each box holds 10 packs and carries a kanban card.
Number of cards: 40 × 2 × 1.25 ÷ 10 = 10. When staff open a box, its card goes in a post box, and stores refill one box for each card. The loop never holds more than 10 × 10 = 100 packs, and the 25% margin covers days when use runs above 40.
The kanban calculator runs the same sum with your own numbers. Recheck the count when demand or the refill time changes.
Push vs pull: who authorizes the next job
The real difference is what gives permission to start work:
| Point | Push | Pull |
|---|---|---|
| Authorization | A schedule releases work | A downstream need authorizes work or a refill |
| Link between steps | Output moves on even if the next step is slower | Supply responds to the consuming step |
| Inventory | Stock builds up between separate schedules | Refill rules or caps limit release |
| When the buffer is full | Production may carry on to plan | The upstream step waits for a free slot |
Authorization
A schedule releases work
Pull: A downstream need authorizes work or a refill
Link between steps
Output moves on even if the next step is slower
Pull: Supply responds to the consuming step
Inventory
Stock builds up between separate schedules
Pull: Refill rules or caps limit release
When the buffer is full
Production may carry on to plan
Pull: The upstream step waits for a free slot
Pull still needs planning for quantities, mix, refill timing and capacity. Keep planning and release apart in writing: forecasts help buy material and plan staffing, but only the agreed signal lets a container or job start. An order-driven shop is not automatically pull either; if every order can be released at once, queues have no limit.
The five steps of a refill loop
A working loop connects use, authorization, refill and confirmation:
- 1
Consume
Assembly takes an approved container of parts.
- 2
Signal
The empty container, card or scan asks for a replacement.
- 3
Authorize
The supplying step accepts the signal under the agreed rules.
- 4
Refill
It makes the set quantity and confirms quality.
- 5
Return
Material handling brings the replacement to its marked location and closes the loop.
↻ Step 5 closes the loop; the next withdrawal starts it again
Over longer distances, separate move and make signals keep the two jobs clear. Before a pilot, name an owner for every way the loop can break: a missing card, a damaged container, rejected parts, a late refill or a demand spike. Otherwise operators improvise at the moment the rules matter most.
Which type of pull fits
Match the control method to each item or product family:
| Type | Fits | Watch out for |
|---|---|---|
| Supermarket pull | Repeat items with steady use and manageable storage | Needs stock of every part number held; watch engineering changes and look-alike parts |
| Sequential pull | High variety, where stocking every variant is impractical | Needs short, steady lead times and a set build sequence with one owner |
| Mixed pull | Standard parts refilled, specials built in sequence | Needs one rule for clashes on shared equipment |
Supermarket pull
Repeat items with steady use and manageable storage
Watch out for: Needs stock of every part number held; watch engineering changes and look-alike parts
Sequential pull
High variety, where stocking every variant is impractical
Watch out for: Needs short, steady lead times and a set build sequence with one owner
Mixed pull
Standard parts refilled, specials built in sequence
Watch out for: Needs one rule for clashes on shared equipment
Do not choose by popularity alone. Record why each item is in the supermarket, so a deliberate policy does not turn into habit.
A plan for every part
For each item in the loop, keep one checked record:
Part number and revision
So the right version is stored and refilled.
Container quantity
How many parts each signal represents.
Storage address
The marked place the part lives.
Consuming location
Where it is used.
Supplier
The process or vendor that refills it.
Refill owner
Who collects signals and closes the loop.
Demand and refill time
Observed use and the full loop time, checked at the workplace, not only in the system.
Launching a pull pilot
- 1
Define the pilot and baseline. Name the supplying step, consuming step, product family and leader, and the problem, such as repeat shortages despite high stock. Record delivery performance, stock locations, queue size, refill time and exceptions.
- 2
Choose the control method and release point. Mark on a simple map where work is first authorized and where it just follows the flow.
- 3
Test the rules on paper. Walk through normal use, a full lane, an urgent order, a quality reject and a missing supplier step. Each case needs one answer for who acts and which instruction wins.
- 4
Set quantities and locations. Size from observed use and the full loop time: signal collection, waiting, production, inspection, transport and return, not only run time.
- 5
Train by doing. Have the people who run the loop show consumption, signal return, refill and the full-lane response, including unusual cases.
- 6
Review exceptions daily. Look at shortages, extra stock, late signals, quality holds and unauthorized releases, with one owner and one action for each repeat issue. Remove protection in small, recorded steps.
Pull system mistakes and fixes
- 1
Two sources of instructions. A card says one thing and the schedule another. Name one authorizing signal for each release decision.
- 2
Hidden extra stock. Operators keep a private buffer. Find out why, then fix the process or approve the protection openly.
- 3
No full-lane rule. State the stop rule, who to call and what approved work to do while waiting.
- 4
Urgent orders bypassing the loop. Give exceptions an owner and record what work they displaced.
- 5
One rule for unlike products. Segment products and record why each method fits.
- 6
Digital before the loop works. Prove the physical loop and its exceptions before automating.
- 7
Nobody owns the signals. Assign collection, acknowledgment, refill and closure.
A pull system scorecard
Keep service, quality, stock and discipline on one sheet, with the same definitions before and after launch:
| Measure | What to track | Warning sign |
|---|---|---|
| Delivery | Downstream needs met on time | Falling stock while shortages rise |
| Inventory | Usable stock and WIP against approved limits | Stock moved off the floor, not removed |
| Refill | Typical loop time and late returns | A loop that only works on day shift |
| Quality | Rejected material and how gaps are handled | Bad parts refilled into the supermarket |
| Discipline | Unauthorized releases, overrides and missing signals | A clean chart built on workarounds |
Delivery
Downstream needs met on time
Warning sign: Falling stock while shortages rise
Inventory
Usable stock and WIP against approved limits
Warning sign: Stock moved off the floor, not removed
Refill
Typical loop time and late returns
Warning sign: A loop that only works on day shift
Quality
Rejected material and how gaps are handled
Warning sign: Bad parts refilled into the supermarket
Discipline
Unauthorized releases, overrides and missing signals
Warning sign: A clean chart built on workarounds
When a shortage happens, trace the loop from use through signal, production and delivery. When extra stock appears, check whether demand changed, limits were bypassed or the refill rule needs revising.
Guides on this topic
- Kanban board examples: 15 boards from different settingsKanban board examples for factories, hospitals, offices, software, sales, support, portfolios, classrooms and homes, with columns, WIP limits and lessons.Read the guide
- Kanban project management: principles, practices and metricsKanban project management uses a visual board, WIP limits and flow metrics. Learn Anderson's principles and practices, and how kanban compares with Scrum.Read the guide
- Value stream mapping symbols: the standard VSM icon setValue stream mapping symbols explained: what each VSM icon shows, when to draw it, how to read the timeline, and tips for mapping by hand or in software.Read the guide
- Lean thinking: the 5 principles and the mindsetLean thinking is a way of creating value with less waste. Learn the five principles, the lean mindset, how it differs from lean tools and how managers apply it.Read the guide
Sources
Books and papers show their publication year. Web pages show the month we last checked them.
- 1.Taiichi Ohno, Toyota Production System: Beyond Large-Scale Production (Productivity Press, 1988)
- 2.Art Smalley, Creating Level Pull (Lean Enterprise Institute, 2004)
- 3.James P. Womack and Daniel T. Jones, Lean Thinking: Banish Waste and Create Wealth in Your Corporation (Simon & Schuster, 1996)