故事 · 一位工程师对「打分排名」的反叛
Origin Story · One Engineer's Revolt Against Decimal Scoring
1980 年代,英国工程师 Stuart Pugh 受够了团队评选方案时的乱象:每个人拿着一张「各项打 1~10 分」的表,
争论「这项到底是 7 分还是 8 分」,最后加权一算,谁也不服。Pugh 看穿了问题 ——
人对「绝对分数」的判断极不可靠,但对「这个比那个好还是差」却相当敏锐。
于是他把打分表砍到只剩三个符号:＋、0、−,全部相对一个共同基准来比。
没有了虚假的小数点,争论瞬间聚焦:「这个方案在这条准则上,到底比基准好还是差?」答案非黑即白。
更关键的是,Pugh 矩阵不是用来一锤定音的 —— 它是个收敛引擎:
看完一轮,把各方案的强项嫁接出新方案、砍掉一直挨打的弱者,换个更强的当基准,再比一轮。
几轮下来,方案越来越少、越来越好,最后剩下的那个,是被反复捶打验证过的最优概念。
In the 1980s, British engineer Stuart Pugh had had enough of how teams chose between design concepts. Everyone showed up with a "score each criterion 1–10" sheet,
argued whether a row "deserved a 7 or an 8," then weighted-summed the result — and nobody trusted the winner. Pugh saw straight through it:
humans are unreliable at judging absolute scores, but quite sharp at judging "is A better or worse than B".
So he cut the scoring sheet down to three marks — ＋, 0, − — all relative to one shared datum.
With the fake decimals gone, the argument snapped into focus: "On this criterion, is this concept better than the datum, or worse?" The answer is binary.
And the deeper point: a Pugh matrix isn't meant to crown a winner in one shot — it's a convergence engine.
After each round you graft the strengths of different concepts into new hybrids, kill off the perennial losers, promote a stronger concept to be the new datum, and run another round.
A few cycles in, the field shrinks and sharpens, and what remains has been beaten on from every direction — the best concept, earned the hard way.
1 可交互 Pugh 矩阵:点格子,看净分 Interactive Pugh Matrix: Click Cells, Watch Net Scores
点格子试试Click a cell2 Pugh 不是评分,是「收敛引擎」 Pugh Isn't a Scoring Sheet — It's a Convergence Engine
① 选基准:挑一个大家熟悉的方案(常是现有产品或最稳的概念)当参照系。
① Pick the datum: choose a concept everyone knows well (often the incumbent product or the most stable proposal) as the reference.
② 逐条比:每个方案 × 每条准则,只判「比基准好 ＋ / 相当 0 / 差 −」。
② Compare cell by cell: every concept × every criterion, judged only as "better than datum (+) / same (0) / worse (−)".
③ 杂交强项:把不同方案各自的 ＋ 嫁接到一起,造出更强的新方案。
③ Hybridise the strengths: graft the +'s from different concepts together to engineer a stronger new candidate.
④ 换基准再比:淘汰弱者,用上轮赢家当新基准,再跑一轮,直到收敛。
④ New datum, new round: drop the losers, promote last round's winner to be the new datum, and run the matrix again — until it converges.
3 现实里的 Pugh 选择 Pugh Selection in the Wild
概念选择:产品早期有多个架构方案,用 Pugh 快速筛掉明显劣势的,聚焦少数候选。
Concept selection: early-stage products with several architecture options — Pugh quickly filters out the obvious losers and leaves a short list to focus on.
相对评估:不纠结绝对分数,只问「相对基准好还是差」,决策更快也更稳健。
Relative evaluation: stop fighting over absolute scores. The only question is "better or worse than the datum?" — decisions land faster and hold up better.
迭代收敛:跑多轮、换基准,让方案一轮比一轮强,而非一次性拍板。
Iterative convergence: run several rounds, swap the datum, let each round produce a stronger field — instead of pretending one matrix is the verdict.
方案杂交:A 的电池布局 + B 的散热结构,杂交出谁都比不过的「混血」新概念。
Concept hybridisation: A's battery layout + B's thermal architecture → a "crossbreed" concept that nothing else in the field can touch.
一句话In One Line
Pugh 矩阵的天才之处,是它主动放弃了精确。它不问「这个方案值几分」,只问「它比基准好还是差」 ——
因为人脑判断相对优劣,远比判断绝对分数可靠。砍掉小数点,争论就从「7 分还是 8 分」回到了「到底哪个更好」。
但更深的智慧在第二层:它根本不是用来选「现有方案里最好的」,而是用来「造出一个更好的」。
看清谁在哪条准则上强,就把这些强项嫁接成新方案;换个基准再比,让标准水涨船高。
所以同一份矩阵,换个基准、跑下一轮,胜负就可能洗牌 —— 这不是 bug,正是它的设计:
最优概念不是被「选」出来的,是被一轮轮比较「逼」出来的。
The genius of the Pugh matrix is that it deliberately gives up precision. It refuses to ask "how many points is this concept worth?" and asks only "is it better or worse than the datum?" —
because the human brain judges relative merit far more reliably than absolute scores. Strip away the decimals and the argument shifts from "7 or 8?" back to "which one is actually better?"
The deeper lesson is the second move: Pugh isn't really about picking the best of what you already have — it's about engineering something better than any of them.
Once you can see who wins on which criterion, you graft those strengths into a hybrid concept; swap in a stronger datum, run the matrix again, and the bar rises with every round.
That's why the same matrix can reshuffle its winners when the datum changes — not a bug, the whole point:
the best concept isn't "selected"; it's beaten into existence, round after round.
常见误用Common Mistakes
偷偷给 ＋/− 加权打成小数。Pugh 的力量正在于只用三档符号,加了小数就退回它要逃离的虚假精确。
Sneaking weights onto ＋/− and turning them into decimals. The power of Pugh is the three-symbol discipline — once decimals creep back in, you've slid right back into the false precision the method was built to escape.
跑一轮就拍板定案。它是收敛引擎,要杂交强项、换基准、多轮迭代,越跑越优。
Calling the verdict after a single round. It's a convergence engine — hybridise the strengths, switch the datum, iterate. Each round should leave the field stronger than the last.
选个谁都不熟的方案当基准。基准要是团队公认、有共识的参照,比较才有共同语言。
Picking an obscure concept as the datum. The datum has to be a reference the team genuinely knows and agrees on — otherwise the comparisons don't share a common language.