「全国宠物医院大概几万家，我们做到千分之一就是几百万营收，明年就能打平。」每个字都真诚，每个字都在关掉他的兴趣。「打平」在他耳朵里等于「上限已经写好了」——他不需要一家能打平的公司，他需要一家可能翻二十倍的公司。
「排班只是我们进门的那把钥匙：它拿到的是每家医院的人力排布和到店时段。现在 40 家在付费，续费的 38 家。同一套数据往下走是耗材补货和薪酬结算，那两块的客单价是排班的十倍以上。所以我们赌的不是排班这门生意，是成为这几万家小机构的后台——第一个能证伪它的节点是明年 Q2：如果补货模块在现有客户里跑不到三成渗透，这条路径就是假的。」
EN Venture returns are brutally skewed: most deals return less than the money put in, and a fund's entire result comes from a couple of outliers. That single fact reshapes the room. The investor is not asking "can this work" — they are asking "if it works, how big does it get", because a company that reliably nets a few million is, for them, the wrong product, however good a business it is. Their error costs are asymmetric too: passing on a winner costs them one line in a memo, while backing a company that quietly plateaus ties up capital and attention for a decade. So a pitch built on prudence — small team, break-even next year, low burn — reads to them as a ceiling, not as discipline. Give them a credible today (paying customers, retention, unit economics you can defend) and then one path to something much larger, with the milestone that would falsify it named out loud. And if your business genuinely is a good, steady, mid-sized business, the honest move is not to dress it up as a rocket — it's to go find money shaped like your business.
机制是基金的结构，不是投资人的性格。他的钱来自有期限的基金，业绩由整支组合决定，而组合的回报被极少数赢家主宰——这就把他的两类错误做成了不对称：错过一个赢家只损失一次机会，投中一个不温不火的公司却要占住资金和席位十年。于是理性的做法是宁可错过很多，也要专挑「万一巨大」的那一类。你讲得越稳，他越确信天花板已经写好；你讲得越夸张而无证据，他越确信你自己也不知道要去哪。真正过关的是第三种：今天可信、上限敞开。
边界与失效场景：幂律逻辑适用于要「基金级回报」的机构钱。天使个人、战略投资方、产业基金的目标函数不同（有人图学习，有人图供应链协同），硬套「必须做到百亿」反而失焦。另外别把「敞开上限」做成编故事：你给的每一个大数字，都要能倒推回一个今天能查的小数字，否则第一轮尽调就碎。
"Scheduling is the wedge, not the business." —— 用切入点区分眼下与终局。
"Here's what would prove me wrong, and by when." —— 主动给出证伪节点。
"That's a solid business, but it isn't venture-shaped — I'd rather say so." —— 诚实承认资金形状不匹配。
「我们帮 [具体的谁] 把 [某件事] 从 [现在的状态] 变成 [新的状态]。」——第一句就交付定位，不要用「我们是一个致力于赋能行业的平台」开场。
「现在有 [数字] 在付费，其中 [数字] 续了第二年。」——把可核对的东西提到最前面。
「我今天想跟你确认的是 [一个具体问题]，其余的都在材料里。」——给这场会一个明确目标，别让它变成漫谈。
EN A deck is the skeleton of a conversation, not a document. One question per slide, answered, then move on. Two well-replicated findings govern the room. First, redundancy: when the same words appear on screen and come out of your mouth at once, comprehension drops — reading and listening compete for the same channel, so a slide crowded with your script effectively mutes you. Put the sentence in your mouth and one image or one number on the slide. Second, thin slices: strangers form remarkably stable impressions within seconds, and those early impressions steer how everything afterwards is interpreted — a strong first thirty seconds doesn't win the deal, but a weak one makes the next twenty minutes an uphill argument. Practically, that means the first line delivers the positioning rather than warming up to it, the checkable numbers come early rather than being saved for the end, and the meeting has a stated goal. Be honest about the status of the famous "10 slides / 20 minutes / 30-point font" rule: it's seasoned practitioner advice, not an experimental result.
机制两条，都是加工带宽的问题。一、冗余：同样的文字同时出现在屏幕上和你嘴里，两路信息抢同一条通道，理解反而下降——所以幻灯片留一句话、一张图或一个数字，完整的话由你说。二、薄片印象：人在几秒到几十秒内就形成相当稳定的判断，之后的信息多半被拿去印证这个判断。这不是让你去表演自信，是让你把最有信息量的东西放在最前面：定位、以及一个别人能核对的数字。
边界：并非所有场合都该「少即是多」。给合伙人内部传阅的版本、以及尽调阶段的材料，需要的是完整可查，那一版可以厚；面对面讲的那一版必须薄。两版都要，别互相冒充。还有一点：真正把 deck 讲垮的通常不是页数，是你自己讲到第七页才交代你们究竟是干什么的。
「我们 CTO 是名校计算机硕士，我做过五年运营，我们都非常看好这个方向，执行力很强。」这段话换一家公司照样能用，所以它等于什么都没说。名校和年限是资格证明，不是「为什么偏偏是你们」。
「我在这个行业做过四年运营，被一次排班事故坑过一次赔了六万——所以我知道医院真正怕的不是排不出班，是排错班没人兜。这套东西我们在自己前公司先手工跑了半年，跑到第三个月才发现真正卡住的是换班时的责任交接，而不是排班算法。市面上的产品都在做算法。这个判断是我们花了半年才买到的，别人从零开始也得花半年。」
EN The person across the table cannot verify your ability — they have met you twice. What they can verify is what this has already cost you, and whether that cost would be prohibitive for someone merely pretending. That's the whole logic of a costly signal: it carries information precisely because a low-quality sender couldn't afford it. "We're passionate about this space" is free to say, so it transmits nothing. "Forty paying customers, thirty-eight renewed" is not free, so it does. Rank your evidence by how expensive it would be to fake — enthusiasm, then relevant years, then a working thing you built, then someone paying, then someone paying again and bringing a friend — and lead with the most expensive item you honestly have. The same applies to the "why you" question, which most founders waste on credentials: degrees and years are qualifications, not answers. The answer is the specific expensive mistake you already made, the wrong turn you already took, the thing you now know that a smart newcomer would need six months to learn.
机制是信息不对称下的昂贵信号：当接收方无法直接验证质量时，能传递信息的只有那些「对冒充者来说更贵」的行为。热情、决心、看好某个方向，成本为零，所以在均衡里被完全忽略——不是投资人冷漠，是这些话在数学上不携带信息。反过来，一个已经付过的代价（做出来的东西、有人付的钱、复购、你踩过的具体的坑）无法零成本伪造，所以它才被当真。这也解释了为什么早期融资里「团队」的权重最高：产品会变、市场会变，最难伪造的是这群人已经在这件事上花掉的东西。
边界：昂贵信号也会被伪造——虚报用户数、把试用算成付费、把朋友的订单算成市场验证。这类伪造在尽调期几乎必然暴露，而且损失的不是一轮融资，是你在一个很小的圈子里的名字。另有一条不公平的现实：信号的「贵」是相对个人资源而言的，全职烧半年积蓄做原型，对有家庭负担的人贵得多。这不改变机制，但值得在评价别人时记得。
"We paid for that lesson the hard way." —— 把学费变成资历。
"Forty paying, thirty-eight renewed — happy to walk you through the churn." —— 给数字并主动开放核对。
"Anyone starting today would need six months to learn what we already know." —— 把认知差说成时间差。
「这是我们目前最大的不确定。」——先承认，别绕。绕一次，他后面每句话都会打折。
「我们的应对是 [具体动作]，判断它有没有用的指标是 [具体指标]，时间点是 [具体时间]。」——承认之后必须给出应对，光认不答会掉分。
「而且同一件事往前看是这样：[把它接回增长]。」——把防守问题的落点搬回增长框架。
真不知道的数字：「我不知道，我不想编。下周三之前给你，我会用 [方法] 去算。」——给期限和方法，比给一个假数字强一百倍。
EN Pushback in a pitch is usually diligence, not rejection — someone who has already decided against you says "very interesting, let's keep in touch" and disappears. So the goal in Q&A isn't to win the exchange; it's to show you've already thought about the thing and know what you'd do. Three mechanisms matter. First, ambiguity is worse than risk: people react far more negatively to unknown odds than to known bad odds, so "I don't know yet, here's how I'll find out and by when" beats both a bluff and a dodge. Second, two-sided arguments work — but only the refutational kind. Naming the weakness and answering it raises credibility; naming it and leaving it hanging lands below saying nothing. Third, the frame of the question tends to become the frame of the answer, and defensive questions answered defensively turn the whole meeting into a risk inventory. Answer the thirty seconds honestly, then move the endpoint back to growth. And never invent a number: a fabricated figure survives one meeting and dies in diligence.
机制三条。一、模糊比风险更让人退缩：人对「概率未知」的反应远比对「概率已知但不高」更负面，所以「我还不知道，但我会这样去查、什么时候给你答案」比含糊搪塞或硬编一个数字都强——它把模糊换成了风险。二、双面论证：主动说出弱点并给出应对，比只讲优点更有说服力；但必须带反驳，只承认不应对反而比不提还差。三、框架黏性：问题的框架会顺势变成答案的框架，防守型提问顺着答，整场会被讲成一份风险清单。
边界：重框不等于回避。躲开一个尖锐问题的代价，比答得不完美大得多——你躲一次，他会默认所有没细看的地方都有类似的洞。也别把「承认弱点」做成表演式的谦虚（「我们最大的缺点是太拼了」），那比不承认更伤。这套动作和「异议不是拒绝」那期的结构是同一个：先接住，再回应，最后把话头带回你要去的地方。
"That's our biggest open question, and here's how we're testing it." —— 认 + 答一次给全。
"I don't know — I'd rather not make one up. You'll have it by Wednesday." —— 不编数字，给期限。
"Same thing looking forward, that's actually our fastest channel." —— 把防守落点搬回增长。
九十分钟，四件事。做完你手上会有一份能发出去的东西，而不是一份还在改的东西。