起源故事 · QS-9000 与客户视角 Origin Story · QS-9000 and the Customer's Lens
1989 年,克莱斯勒 / 福特 / 通用三家在制定 QS-9000 时发现一个尴尬:工厂用"控制图组内 σ"算出来的 Cp 漂亮极了, 客户拿到几个月的累积数据再算,结果完全不一样 —— 中间差了一个"过程随时间在偷偷漂"。为了平息争吵,三家联合引入了 Pp / Ppk, 强制用 σ_overall(全部数据合并)替代 σ_within,让供应商必须报"长期客户视角"。从此 Pp / Ppk 才有了独立身份 —— 它们不是新公式,是同一个 Cp / Cpk 换了把 σ 尺子,专门用来照妖。 In 1989, Chrysler, Ford, and GM were drafting QS-9000 and ran into the same awkward gap every supplier knew about: Cp built from a control chart's within-subgroup σ looked beautiful, but the customer would crunch a few months of accumulated data and get something far worse — the missing piece was the process quietly drifting over time. To end the argument, the Big Three jointly introduced Pp / Ppk, forcing the use of σ_overall (pooled from all data) instead of σ_within, so suppliers finally had to report the customer's long-term truth. That's how Pp / Ppk earned their own identity — not as new formulas, but as the same Cp / Cpk swapped onto a longer ruler, purpose-built to expose hidden drift.

1 一周数据:组内瘦,全合胖 One Week of Data: Lean Per Day, Fat When Pooled

无漂移

2 Cp 与 Pp 双柱 · 稳定性体温计 Cp vs Pp Bars · The Stability Thermometer

Cp = Pp → 完美稳定,SPC 价值低;Cp ≫ Pp → 过程在偷漂,必须查刀具磨损 / 季节 / 班次差。 Cp = Pp → rock steady, SPC adds little; Cp ≫ Pp → the process is drifting on you — go hunt tool wear, season, or shift-to-shift differences.

3 现实里的 Pp Pp in the Real World

试生产 vs 量产:试生产 1 天 Cp=1.67,量产 3 个月 Pp=1.2 → 过程在飘,必须查"长期变量"(温湿度 / 批次 / 班次)。 Pilot vs mass production: one-day pilot Cp = 1.67, three-month mass-production Pp = 1.2 → the process is drifting. Hunt the long-term variables (temperature, humidity, batch, shift).
全自动稳定设备:Cp≈Pp,差距 < 5% → SPC 价值低,可以拉长抽样周期省成本。 Stable automated equipment: Cp ≈ Pp with under 5 % gap → SPC adds little value; stretch the sampling interval and save inspection cost.
季节温度依赖过程:夏冬温差大,σ_between 大,Pp ≪ Cp → 必须上恒温车间或季节性参数补偿。 Season-sensitive process: big summer-to-winter swing pumps σ_between up, so Pp ≪ Cp → either climate-control the line or compensate parameters by season.
客户报告:客户买的是几个月的平均产品,必须给 Pp;只给 Cp = 报喜不报忧,PPAP 会被打回。 Customer reporting: the customer buys months of average product, so Pp is mandatory. Sending Cp only is cherry-picking — PPAP will be rejected.
一句话In One Line
Cp 是"今天瞬间最佳",Pp 是"长期客户视角"。 它们的差距 = 过程在偷漂的代价(刀具磨损、季节、班次差、原料批次)。 差距大 → SPC 价值高,必须找出 σ_between 的来源并锁住;差距小 → 你的工艺已经很稳,可以减少 SPC 抽样频率省成本。 公式上只差一把 σ 尺子,含义上一个是体检报告、一个是长期病历,缺一不可。 Cp is the "best snapshot of today", Pp is the "customer's long-term reality". Their gap is the price the process pays for drifting — tool wear, season, shift, raw-material batch. A wide gap means SPC has real value: trace σ_between to its source and pin it down. A tiny gap means your process is genuinely steady, so you can stretch sampling and save cost. One formula, two rulers. One is a snapshot physical, the other is a long-term medical record — neither is optional.
常见误用Common Mistakes
Cp 和 Pp 当一回事互相替换。Cp 用 σ_within(控制图 R̄/d₂),Pp 用 σ_overall(全部数据合并 STDEV)。完全两个尺子。 Treating Cp and Pp as interchangeable. Cp uses σ_within (control-chart R̄/d₂); Pp uses σ_overall (pooled STDEV of all data). Two different rulers, not the same thing.
不收集时间序列,只算一个合并 σ。必须分子组采样(每天 / 每班次一组),才能拆 σ_within 和 σ_between。 No subgrouped time series — just one bulk σ. Sample in subgroups (one per day or per shift). Only then can you decompose σ_within and σ_between.
报告只给 Cp 不给 Pp。客户要的是 Pp / Ppk;PPAP 文件必须双报。只给 Cp = 报喜不报忧,会被审计打回。 Reporting Cp only, hiding Pp. Customers want Pp / Ppk; PPAP requires both. Cp-only is cherry-picking and the audit will throw it back.

Pp 整体 σ