由来 · 丰田的两层改善
Origin · Toyota's Two Layers of Kaizen
丰田的改善从来不是"谁有空就改改身边"的随机活动。它有清晰的两层结构:
流改善(Flow Kaizen)由管理层主导,站在整条价值流上看 —— 瓶颈在哪、库存堆在哪、节拍对不对,优化的是"水流本身";
点改善(Process Kaizen)由班组现场推进,打磨单个工序的动作、布局、换模,优化的是"某块石头"。
Mike Rother 在《丰田改善套路(Toyota Kata)》里反复强调:顺序不能反。先看清整条流,再去改具体的点 ——
否则你会非常勤奋地,在错误的地方,做着完全正确的改善。
At Toyota, kaizen was never "whoever's free tinkers with whatever's nearby." It has a clear two-layer structure:
Flow Kaizen is led by management, standing over the whole value stream — where's the bottleneck, where does WIP pile up, is takt right — optimizing the current itself;
Process Kaizen is run by the team on the shop floor, polishing motion, layout, and changeover at one station — optimizing a single stone.
Mike Rother hammers it home in Toyota Kata: you cannot reverse the order. See the whole flow first, then improve specific points —
otherwise you'll work very hard, in the wrong place, doing exactly the right kaizen.
1 左边踩坑,右边看清:同一条线,两种改法 Same Line, Two Kaizen Approaches
拖滑块对比2 谁的努力换来了真产出 Whose Effort Actually Bought Output
两条进度条:点改善常常推半天却纹丝不动(改的不是瓶颈),流改善把瓶颈抬起来,整条线立刻跟着涨 —— 这就是"先流后点"的全部道理。 Two progress bars: process kaizen often pushes for hours and the needle doesn't move (wrong target); flow kaizen lifts the bottleneck and the whole line climbs with it — that's the entire logic of "flow first, point second."
3 现实里的两层改善 The Two Layers in the Real World
改善层次:管理层画价值流定方向(流改善),班组做现场改善(点改善),两层各司其职、上下衔接。
Strategic vs tactical: management maps the value stream and sets direction (flow kaizen); teams do shop-floor improvement (process kaizen). Two layers, each with its own job, locked together.
先价值流后工序:先用 VSM 找到瓶颈和断点,再决定在哪几个工序投点改善 —— 顺序反了就白忙。
Value stream before station: VSM first — find the bottleneck and the breaks — then decide which stations get process kaizen. Flip the order and the effort burns for nothing.
避免局部优化:某工位效率冠军,整线却没提速 —— 经典局部最优,与 TOC 约束理论同一个道理。
Avoid local optima: one station wins efficiency-of-the-month, the line doesn't move — textbook local optimum, the same lesson the Theory of Constraints keeps repeating.
呼应 TOC 约束:非瓶颈的产能是"假产能";只有抬高约束(瓶颈),系统产出才会增长。
Echoes TOC: capacity at a non-constraint is phantom capacity. System output only grows when you elevate the constraint itself.
一句话In One Line
一条产线的产出,由最慢的那道工序(瓶颈)封顶。在非瓶颈上做再漂亮的点改善,
也只是把在制品堆得更高,产出一点没多 —— 这就是为什么"先流后点"是铁律。
流改善负责回答"该改哪里",点改善负责回答"怎么改好"。没有流的视角,所有点改善都是在赌运气。
这也正是约束理论(TOC)的核心:管好那一个约束,胜过优化其余九个。
A line's output is capped by its slowest station — the bottleneck. The most beautiful process kaizen on a non-bottleneck
just piles WIP higher and adds zero output — which is why "flow first, point second" is iron law.
Flow kaizen answers where to improve; process kaizen answers how. Without the flow view, every process kaizen is a coin flip.
This is the same heartbeat as the Theory of Constraints: managing the one constraint beats optimizing the other nine.
常见误用Common Mistakes
哪里好改改哪里(只做点改善)。先用价值流找瓶颈,再投点改善;先流后点。
Improving whatever's easiest to improve (process kaizen only). Map the value stream first, find the bottleneck, then invest process kaizen — flow first, point second.
用单工序效率当成绩考核。看整线产出和前置时间,别奖励局部最优。
Scoring people on single-station efficiency. Measure line throughput and lead time — don't reward local optima.
把流改善当成一次性大项目。流改善定方向、点改善持续落地,两层循环往复。
Treating flow kaizen as a one-shot mega-project. Flow kaizen sets direction; process kaizen executes daily. The two layers loop, forever.