故事 · 丰田超市与那张卡片
Origin · Toyota's Supermarket and a Single Card
二战后,丰田的 大野耐一 从美国超市得到启发:货架空了,店员才补货 —— 顾客"拉动"补给,而非仓库往前"推"。
他把这套搬进车间,发明了"看板(Kanban)":每个零件箱挂一张卡,箱子被取走、卡片回流,就是给上游的生产许可。
没有卡,绝不开工。于是系统在制品被牢牢锁在"卡的总张数 × 每箱容量"之内。
张数太少会断料,太多则又回到堆库存的老路 —— 所以丰田用 K = (D × L × (1 + S)) / C 算出恰好够用的卡数,
让在制品上限既能扛住补货等待与波动,又不至于浪费。
After WWII, Toyota's Taiichi Ohno took his cue from American supermarkets: clerks only restocked when a shelf went empty — customers pulled the replenishment, the warehouse didn't push it forward.
He carried that idea onto the shop floor and invented the kanban: every bin of parts gets a card; when the bin is taken and the card flows back upstream, it acts as the production license to make more.
No card, no work. WIP is now locked inside "total cards × pieces per bin".
Too few cards starves the line; too many drags the plant back into pile-it-up inventory — so Toyota uses K = (D × L × (1 + S)) / C to compute exactly the right number,
so the WIP ceiling absorbs replenishment delays and variation without wasting cash on excess stock.
1 四个旋钮如何拼出看板数 K How Four Dials Compose the Kanban Count K
K = —2 卡片循环:在制品被锁在 K × C 以内 The Card Loop: WIP Locked Within K × C
满箱挂卡在下游待用,被取走后空卡回流到上游当作生产令。 系统里同时存在的箱子永远 ≤ K 张 —— 这就是 WIP 的硬上限。 Full bins sit downstream with cards attached, waiting to be consumed. Once a bin is taken, the empty card flows back upstream as a production order. The number of bins alive in the system can never exceed K cards — that's the hard WIP ceiling.
3 现实里的看板卡数 Kanban Card Counts in the Real World
装配线补料:线边料盒挂卡,盒空回卡,仓库按卡备料 —— 卡数定死线边库存上限,杜绝堆积。
Assembly-line replenishment: bins at the line carry cards; an empty bin returns the card to the warehouse, which then refills it. Card count fixes the line-side inventory cap and stops piles from forming.
看板软件墙:每列设 WIP 上限(= 卡位数),满了就不许再拉新卡,逼团队先完成再开新活。
Software kanban board: every column has a WIP limit (= the number of card slots). Once it's full, no new card can be pulled in — forcing the team to finish before starting anything new.
超市货架:补货点 = D×L,安全库存 = D×L×S,看板把这套"两箱法 / 双箱卡"直接物化成卡片。
Supermarket shelf: reorder point = D×L, safety stock = D×L×S — kanban turns this classic two-bin system into a physical card you can hold.
需求骤变:旺季 D 上升要增卡,淡季 D 下降要撤卡 —— 卡数不是一次定终身,要随节拍周期复算。
Sudden demand shifts: when D climbs in peak season add cards, when it drops in the slow season pull cards out — K is never set once and forever, you re-compute it every takt review.
一句话In One Line
看板数公式其实就一句话:补货等待期内会消耗多少,就备多少卡,再加点安全裕度。
分子 D×L 是前置时间需求,(1+S) 应对波动与故障,分母 C 把件数换算成卡数。
最关键的领悟是:卡数 = 在制品上限。你不是在"计划库存",而是在设一个不可逾越的天花板 ——
一旦卡都在用、没有空卡回流,上游就必须停工,问题立刻暴露而不是被库存掩盖。这正是拉动系统逼出持续改善的机制:
想缩短前置时间、降库存,就主动抽卡,让系统在更紧的约束下自我进化。
The kanban formula is really one sentence: stock enough cards to cover what you'll consume during replenishment, plus a touch of safety.
The numerator D × L is lead-time demand, (1 + S) absorbs variation and breakdowns, and the denominator C converts pieces into cards.
The deepest insight: K = the WIP ceiling. You're not "planning inventory" — you're setting a hard cap nobody is allowed to cross.
Once every card is in use and none have flowed back empty, the upstream must stop — problems surface instantly instead of being buried under stock. This is the mechanism that forces a pull system to keep improving:
if you want shorter lead times and less inventory, pull a card out — and let the system evolve under a tighter constraint.
常见误用Common Mistakes
卡数定一次就再不复算。D、L 会随季节与改善变化,要定期按节拍周期重算 K,多了撤卡、少了增卡。
Setting K once and never revisiting it. D and L drift with seasons and improvement work — re-compute K every takt review, pull cards out when there are too many, add them in when there aren't enough.
安全系数 S 设得越大越安心。S 越大库存越高、问题越被掩盖;先降波动与前置时间,再压 S,而非用库存对冲。
Cranking S up because it feels safer. A larger S means more inventory and more problems hidden under stock. Reduce variation and lead time first, then squeeze S down — don't hedge with inventory.
需求剧烈起伏的产品也强行上看板。看板适合相对稳定、重复性需求;高度定制或波动极大时该用其他拉动 / 计划方式。
Forcing kanban onto wildly fluctuating demand. Kanban fits relatively stable, repetitive demand. For highly customized or very volatile products, use a different pull or scheduling approach.