故事 · 川喜田二郎与「让数据说话」
Origin Story · Jiro Kawakita and "Let the Data Speak"
1960 年代,日本人类学家川喜田二郎在喜马拉雅做田野调查,带回成箱杂乱的观察记录。
他没有按既有理论分类,而是把每条记录写在卡片上,反复阅读、凭直觉把感觉相关的卡片聚到一起,
再给每一堆起个名字。他发现:当你不预设框架,真正的模式反而会自己浮现。
这套方法以他名字的首字母命名 —— KJ 法,后来成了设计思维、用户研究里整理定性数据的标配。
它的反直觉之处在于:聪明的分类,是发现出来的,不是设计出来的。
In the 1960s, Japanese anthropologist Jiro Kawakita came back from Himalayan fieldwork with crates of tangled observation notes.
Instead of forcing them into existing theories, he wrote each observation on a card and kept re-reading them, intuitively pulling related cards together,
then naming each pile only at the end. His finding: when you refuse to set the frame in advance, the real patterns surface on their own.
The method took the initials of his name — the KJ method — and became the default way to wrangle qualitative data in design thinking and user research.
The counterintuitive lesson: smart categories are discovered, not designed.
1 亲和画布:把像的便签拖到一起 Affinity Board: Drag Similar Notes Together
拖动 / 自动聚类Drag / Auto-cluster2 两种思路:涌现 vs 硬塞 Two Mindsets: Emergence vs Forcing
自下而上(KJ)Bottom-up (KJ)
先看便签像不像,凑近成堆,最后才给堆命名。分类是数据自己长出来的,常冒出你没想到的新主题。
First check if notes feel related, cluster them, and only at the end give each pile a name. Categories grow out of the data — and unexpected themes routinely show up.
自上而下(硬分类)Top-down (forcing)
先定好几个框,再把便签往里塞。框是你的成见,塞不进的真问题会被忽略或硬扭,主题永远跳不出你的预设。
Lock in a few buckets first, then cram notes in. The buckets are your biases — any real issue that doesn't fit gets ignored or contorted, and the themes never escape your preset.
3 现实里的亲和图 Affinity Mapping in the Real World
用户访谈整理:几十场访谈的碎片金句,用亲和图聚出真正反复出现的痛点主题。
User interview synthesis: dozens of interview snippets get affinity-mapped into the pain themes that genuinely repeat.
想法归纳:头脑风暴产出的几百个点子,归纳成几个可执行的方向,而非一锅粥。
Idea synthesis: hundreds of raw brainstorm ideas roll up into a handful of actionable directions instead of one giant blob.
团队共识:大家一起动手贴、一起聚,分类是共同涌现的,比一个人拍板更容易达成共识。
Team consensus: everyone sticks and clusters together; categories emerge collectively, which lands consensus far easier than one person dictating.
定性数据编码:用户研究、社会调查里给开放式回答打主题标签,KJ 是入门级编码法。
Qualitative coding: for open-ended responses in user research and social surveys, KJ is the entry-level thematic coding method.
一句话In One Line
人脑天生爱用现成的盒子装新东西 —— 这快,但也意味着你永远只能看到你已经会的那些分类。
亲和图刻意推迟命名,逼你先和原始数据厮磨:哪两张便签其实在说同一件事？这一堆到底该叫什么？
当主题从下往上长出来时,你常会撞见一个谁都没料到的真问题。
记住:最好的洞察,往往藏在你原本那套分类框装不下的地方。
The brain loves to drop new things into ready-made boxes — fast, but it also means you only ever see the categories you already know.
Affinity mapping deliberately delays naming, forcing you to wrestle with the raw data first: which two notes are actually saying the same thing? What does this pile really deserve to be called?
When themes grow upward from the data, you regularly stumble onto a real problem no one expected.
Remember: the best insights tend to hide exactly where your old category frame couldn't fit them.
常见误用Common Mistakes
开局就先定好分类框再贴便签。先聚后命名,让主题自下而上涌现,别让成见框死结果。
Locking in categories before anyone sticks a note up. Cluster first, name later — let themes emerge bottom-up instead of letting your biases dictate the outcome.
一个人闷头分,分完通知团队。全员一起动手聚,过程本身就是建立共识。
One person sorting alone, then announcing the result. Cluster as a group — the act of clustering together is the consensus-building.
聚成几堆就完事,不命名不提炼。每堆都要起个能说清主题的名字,亲和图的价值在那个名字里。
Stopping at "we made piles" — never naming or distilling them. Every cluster needs a label that captures its theme; the value of an affinity map lives inside those names.