查了一圈发现「代理记账」搜索量最大，于是写了一篇《代理记账全解析》。这个词下面站着的是同行、学生和找工作的人——真正会付钱的人不这么搜，他搜的是今天卡住他的那件事。
五个人都问过同一句：「我个体户和公司哪个更划算？」于是写《月流水 8 万以下，个体户还是公司：一张表算完》，把两边的税、社保、注销成本全列出来。这句话只有正在做这个决定的人会打出来，而正在做这个决定的人就是要买的人。
1. 翻最近 10 次真实对话（销售电话、聊天记录、客服工单），把对方的原话抄下来，别顺手改写成书面语——原话就是标题。
2. 数频次：出现三次以上的问题，一条一篇；只出现过一次的先攒着。
3. 越觉得「这也要写？」越要写。你的「太基础」正好是他的「终于有人讲人话」。
4. 标题别翻译成行话：他不会搜「税务架构选择」，他搜「个体户和公司哪个划算」。术语是给同行看的，同行不买。
EN Don't start from what has search volume; start from your own message history. Which paragraph have you explained to three different people this month? That paragraph is your article. It has already been tested on real humans — you watched it land — and every time you deliver it live you pay full price again, so writing it down buys out the repetition once and for all. The hard part is that you will dismiss it as too basic to be worth writing, and that judgment is almost always wrong: once you know something, you can no longer estimate how much someone who doesn't know it is missing. So: read your last ten real conversations, copy the other person's words verbatim rather than rewriting them into industry language, count which questions came up three times or more, and write those. Their phrasing is your headline. Jargon speaks to peers, and peers don't buy.
机制两层。一是杠杆：讲一次的成本是线性的，写一次是固定成本换近零边际成本——内容替代的是你的重复劳动，不是广告。二是纠偏：你选题会系统性选错，因为回不到「不知道」的状态，这个偏差叫知识的诅咒。频次统计的作用就是把选择权从你的直觉手里拿走。
边界：这两条只说明「你会选错话题」，不说明「重复三次的问题一定值得写」——频次是便宜的筛子，不是保证。另有一层幸存者偏差：你的样本是已经找上门的人，所以每隔一阵要往名单外面问一轮（怎么问才不被误导，见 The Mom Test 那期）。
「自己剪片子太费时间了，交给专业团队，效率提升 3 倍。」——他的第一反应不是「有道理」，是「你当然这么说」。而「效率提升 3 倍」这种没出处的数字，只会把他刚建立的一点信任一起抹掉。
《一个人剪一期播客的完整流程：我的时间轴、快捷键、和三个最费时间的地方》——每一步花多少分钟都写出来，把「口型对齐要 40 分钟」那段讲透。读的人有两种：一种照着做成了，他会记得是谁教的；另一种看到「40 分钟」那行就停住了——他自己得出了结论，而这个结论没被打折。
1. 步骤给全，包括你收费做的那一步。你怕的那一步恰恰是最有说服力的一步——它证明你真的每天在做这件事。
2. 每步标上时间或钱。成本是那个让他自己算账的东西，不是你的说服句。
3. 写清坑在哪。「这里我踩过，第二次才发现是因为……」比任何资历介绍都好用。
4. 结尾不要转折成推销。留一句「不想自己弄的话我做这个」就够了，把要价压到最小、且不要求交换——这跟冷邮件那期的先给后要是同一件事。
EN Educational content isn't there to inform; it's there to get the conclusion out of the reader's own mouth. Write "you should outsource this" and it has to clear a defence — he knows you sell that, so anything favourable to you gets discounted on arrival. Write the full method instead — step one, step two, here are the three places that eat your afternoon — and he runs it in his head and reaches the conclusion himself, undiscounted and better remembered. Which is why the method has to be complete, including the part you charge for: a half-hidden tutorial is the worst version, since it neither helps anyone nor hides that you're fishing. Put a time or a cost on every step, because the cost is what makes him do the arithmetic; say where you personally got stuck; and don't pivot into a pitch at the end — one line saying you do this for people is enough. The readers who follow the method successfully were never going to pay you anyway, and now they repeat what you said.
自己生成的结论比听来的结论更牢、也更能改变行为。两条机制叠在一起：一是生成本身加深了编码（自己想出来的东西记得更好）；二是它绕开了心理逆反——当人感到「你在推我往哪走」，他会本能地往反方向用力，而这个力对你自己产生的念头不发作。
边界：这条打法在「知道了也做不了」的品类最安全（要设备、要时间、要资质、要担风险）；在「看完就能自己做」的品类，你确实会失去一部分本来会付钱的人——这是它真实的成本。判断标准：如果你的全部价值就是「知道那三步」，那你卖的东西本来就不牢。
1. 把内容绑在触发场景上，别绑在产品名上。他不会突然想起「那家代理记账」，他会想起「上次社保被卡时看过一篇讲这个的」。所以标题里放那个场景：「换办公室那个月，社保怎么不断」。
2. 拿一个能直接联系他的地址（邮箱、微信），别只留平台关注——平台随时能掐掉你和他之间那条线，而你连他是谁都不知道。
3. 频率宁低勿高，每封只做一件事：给一个他今天能用的东西。做不到就那个月别发。
4. 退出写清楚、一键能退。留下来的人的价值远大于名单长度。
5. 别做「没回就加频次」的自动化。那是把还没进场的人提前赶走，而他本来是明年的客户。
EN When content doesn't produce customers, the usual reason isn't the writing — it's that they aren't buying today. At any given moment only a small slice of a category is actually in a purchase decision; everyone else isn't uninterested, they just haven't hit their trigger yet: the contract hasn't expired, nothing has broken, the budget hasn't landed. So nurturing can't ripen anyone, and shouldn't try. Its job is to keep you among the two or three names he can retrieve on the day he does show up. That shapes everything: tie your content to the situation he'll be standing in, not to your category name — he won't suddenly think "that bookkeeping firm," he'll think "I read something about this when our filings got stuck." Get an address you own rather than a platform follow, since the platform can cut the line between you at any moment. Keep the frequency low and make every send carry one usable thing; skip the month if you have nothing. Make unsubscribing easy, and never escalate frequency because someone didn't reply — that just evicts next year's customer early.
两条机制。一是基础比例：在市场的人本来就少，你看到的低转化有一大半是这个数在起作用，不是你的文案。二是记忆的可及性：买的时候他脑子里只捞得出几个名字，能不能被捞出来取决于他当时的情境跟你当初绑的线索对不对得上。
边界：这条最容易被拿来给「没结果」找借口。区分方法：内容连一次真实对话都没换来过，那不是时点问题，是入口或人群问题（下一节）；时点问题长的是另一副样子——有人来找你了，但说「先了解一下，明年再说」。
1. 标题写成他会打出来的那句问句，能带上限定条件更好（人数、金额、场景）。「月流水 8 万以下」比「小微企业」精确得多。
2. 一篇只回答一个问题，并在开头三行内把答案给了。绕圈子的写法现在连被摘录的机会都没有。
3. 放只有你才有的东西：你自己的一手数据、你经手的具体数字、可核对的过程。二手总结的边际价值已经接近零——任何人都能一秒生成一篇。
4. 每篇留一条能直接找到你的路：一个邮箱，或一句「卡在哪一步可以直接问我」。别只留平台关注。
5. 别把入口全押在一个平台的推荐上，它的规则不归你管。指标也换掉：别数流量和排名，数「这个月产生了几次真实对话」——渠道会变，这个指标不会失效。
EN Doors aren't won by chasing high-volume terms; they're picked up from absurdly specific questions. What someone types has already done your qualifying: "what is X" is a student, "how do I fix X error" or "is A or B cheaper for me" is someone solving a problem right now. Volume and closeness to a sale run in opposite directions — the biggest phrase brings competitors, students and the curious, while the smallest brings five people a month who are actually in pain. So write the question the way he'd type it, add the qualifier (headcount, revenue, situation), answer one question per piece and give the answer in the first three lines. Put in what only you have — your own numbers, your own process, something checkable — because second-hand summaries are now worth roughly nothing when anyone can generate one in a second. Always leave a direct line back to you rather than a platform follow, never bet every door on one platform's recommender, and change your metric from traffic and rankings to real conversations started this month. Channels change; that number doesn't expire.
机制一：搜索是一次自我筛选。他用的词泄露了他在哪个阶段，入口的精度不靠你判断，靠他自己打出来的字——跟资格判定那期是同一件事，只是这次他替你做完了。机制二（要用也要防）：写得越顺口的东西越容易被当成对的。
边界与不确定：搜索结果页直接给答案的比例这几年在上升，靠「排到第一再被点进来」的路子更不稳了——这是我基于当下产品形态的判断，不是研究结论，别拿它当事实。更稳的推论只有一条：越是人人都能生成的内容越不值钱，越是只有你才有的一手信息越可能被引用。
90 分钟，把「我该写点什么」换成「一篇被真人验证过的东西 ＋ 一条能找到你的路」。