起源故事 · 那个走得最慢的胖男孩
Origin Story · The Slowest Kid in the Hike
《目标》里,厂长罗哥带童子军徒步,队伍越走越散。他发现:最胖、走得最慢的赫比排在中间时,
后面的人被堵成一团,前面的人则把队伍越拉越长。让前面的孩子拼命快走(局部最优),
只会让队伍更散、行李堆在赫比身后。直到罗哥把赫比调到队首、帮他减负、全队迁就他的速度,
队伍才整齐前进。工厂也一样:让非瓶颈工序拼命满负荷,只会制造库存;全局速度永远由瓶颈说了算。
局部最优之和,从来不等于全局最优。
In The Goal, plant manager Alex leads a Boy Scout hike that keeps fraying. He notices: when Herbie — the heaviest, slowest kid — walks in the middle,
the back end bunches up behind him while the front end stretches out of sight. Telling the kids in front to walk faster (the local optimum)
just stretches the column further and stacks gear behind Herbie. Only when Alex moves Herbie to the front, redistributes his pack, and makes the whole troop match his pace
does the line move together. A factory works the same way: maxing out the non-bottleneck stations just manufactures inventory; global pace is always set by the constraint.
The sum of local optima never equals the global optimum.
1 四道工序,产出卡在最慢那道 Four Stages, Output Capped by the Slowest
瓶颈:喷涂Bottleneck: Painting每道工序是一个管口,宽度=它的处理能力。物料从左流到右,瓶颈(最窄管口)决定了流出量。把上游工序拉快,物料只会在瓶颈前积成红色的 WIP。 Each stage is a pipe; width = its throughput capacity. Material flows left to right, and the bottleneck (narrowest pipe) decides what comes out. Speed up the upstream stages and material just piles up as red WIP in front of the bottleneck.
2 拉快非约束:产出不动,库存暴涨 Speed Up Non-Constraints: Output Flat, Inventory Spikes
蓝条=整厂成品产出(被瓶颈封顶,怎么拉非约束都不长);橙条=累积 WIP(非约束越快、堆得越高)。只有把瓶颈本身提速,蓝条才会上涨。 Blue bar = finished output (capped by the bottleneck, so pushing non-constraints does nothing); orange bar = accumulated WIP (the faster the non-constraints, the higher it climbs). Only speeding up the bottleneck itself moves the blue bar.
3 现实里的局部 vs 全局 Local vs Global in the Real World
KPI 陷阱:按工位"开机率/产量"考核,逼着每台机器满负荷,结果库存与交期双双恶化。
KPI trap: Pay each station on uptime or unit count and you push every machine to max — inventory swells and lead times slip at the same time.
部门墙:采购压价、生产冲量、销售压货,各部门 KPI 都达标,公司却被一堆呆滞库存拖垮。
Silo walls: Purchasing chases low PPV, production chases volume, sales chases sell-in — every department hits its KPI while the company drowns in slow-moving inventory.
聚焦瓶颈:与其全员加速,不如先识别约束、给约束减负提速 —— 一处突破,全局受益。
Focus on the constraint: Don't accelerate everyone — identify the constraint, offload it, speed it up. One breakthrough lifts the whole system.
软件交付:开发拼命产出需求,测试/上线是瓶颈 —— 待测积压暴增,价值却没更快交付。
Software delivery: Dev churns out features while QA / release is the bottleneck — the test queue balloons and customer value still ships at the old pace.
一句话In One Line
一条链的强度,由最弱的那环决定;让其它环节更强,链子并不会更结实,只会更重。
系统的产出由约束决定,所以在非约束上提效,几乎全部变成库存与浪费。
真正的杠杆只有一个:识别约束 → 全力压榨约束 → 让其它工序迁就约束,
而不是逼每个部门各自满分。把考核从"局部利用率"改成"全局产出",
才是从忙乱走向高效的关键 —— 下一页把这套思路搬到项目管理:关键链。
A chain is only as strong as its weakest link; reinforcing the other links doesn't make it stronger — only heavier.
System output is set by the constraint, so efficiency gains on non-constraints turn almost entirely into inventory and waste.
There is exactly one real lever: identify the constraint → exploit it ruthlessly → make every other stage subordinate to it,
instead of demanding a perfect score from every department. Move the KPI from "local utilization" to "global throughput"
and you trade chaos for flow. Next page: carry the same idea into project management — the critical chain.
常见误用Common Mistakes
按工位利用率考核,逼人人满负荷。用全局产出考核;非约束本就该有空闲,满负荷只生产库存。
Paying each station on utilization and demanding 100% everywhere. Pay on global throughput. Non-constraints are supposed to have idle time — running them flat-out only manufactures inventory.
到处改善、平均用力。先识别约束,把改善资源集中砸在瓶颈上,杠杆最大。
Spreading improvement effort evenly across the whole plant. Find the constraint first, then concentrate improvement dollars on the bottleneck — that's where leverage lives.
把局部成绩相加当成整体业绩。局部最优之和 ≠ 全局最优;只有约束的改善才转化为全局收益。
Summing local scores and calling it overall performance. Sum of local optima ≠ global optimum. Only improvement at the constraint converts into system-level gains.