中文精读
范皓宇的产品完整性：数据之外，要看人与产品怎样一起生活
42章经单期精读 · 产品设计与长期体验
嘉宾：范皓宇，理想汽车高级副总裁（SVP），身份依当期介绍｜主持：曲凯
发布：2025年10月11日｜时长：82分21秒｜整理：2026年10月18日
官方节目与收听入口
一个有用的功能，为什么还没有成为好产品
范皓宇自述做产品与设计约25年，经历工业设计、手机、游戏、自媒体、操作系统和汽车。他把贯穿这些行业的能力概括为观察、理解、组合和压缩：看见更多生活要素，再把复杂条件变成用户能够直接感受到的方案。
本期按官方约定主要谈产品与人生，不能当成理想汽车具体业务披露。城市、代际、品味及AI未来的讨论，是2025年嘉宾与主持人的观察和假说，未形成代表性调查。本文关注这些观点怎样改变产品评审，同时保留主持人对品味可能出错、超前市场的质疑。
1. 一个不倒翁指纹仪：把场景里的小事放进同一个设计
节目线索：13:56–21:28。 范皓宇最喜欢的早期作品，是研究所办公室的指纹识别仪。桌面堆满材料，小设备容易被盖住；他做成水滴形不倒翁，人翻动材料时它会摇出来，平时又能把玩。找到设备与办公室的小乐趣，在同一形态里得到回应。
这个例子解释他所谓“压缩”：不是先列十条功能，再按顺序完成，而是寻找一个能同时满足多个条件的方案。做游戏、做内容和做硬件，都可以先收集素材、连接关系，经过思考再输出。对匆忙进入A/B测试，他担心的是粗糙假设把选择范围先限制住，而非所有实验都无价值。
他还谈到追逐刺激的循环：用户适应后要更强刺激，供给者不断加速，双方可能更累。这是从创作者交流中形成的解释，不能当普遍因果研究。编辑推论：在点击和停留之外，补问用户使用后是否更容易完成事情、是否认可这段时间，以及持续更新是否制造了额外负担。
2. 同样叫情绪价值，父母和子女可能看见不同的东西
节目线索：26:15–36:07。 曲凯用父母拆解披萨成本作例子：面粉加肉，为何卖那么贵？他想说明实用主义。范皓宇提出反例：上一代也会请朋友吃饭、喝点好酒，也珍惜老物件与照片，情绪价值并未缺席，只是附着在不同意义上。
这段分歧比“年轻人重情绪、老人重实用”的标签更有用。用户在日常省钱与特殊时刻花钱，并不矛盾。看一个人的某次消费，未必能推断其全部价值观；同一个功能在家人、工作和独处情境下也可能有不同意义。
范皓宇还用不同城市的机会、关系网络与服务供给作观察框架。它提醒设计者自己的生活未覆盖所有人，却不能证明一线人必然孤独、低线人必然安全。节目转述的调查比例没有报告可核，本文不将它作为社会事实。编辑推论：让目标用户解释一次真实选择、谁在场、为何值得，比按年龄和城市直接分配动机更可靠。
3. 品味与数据冲突：判断要具体，也要承担结果
节目线索：39:42–47:46。 范皓宇把taste理解为区隔、风格和内容表达。一个产品被看见，不只靠功能清单，还靠持续的选择：提供什么体验、保持什么态度、什么不做。高价或漂亮，并非品味的充分定义。
当数据与感觉冲突，他更相信深入使用和接近用户后的感受。他指出，指标选择与分类本身包含主观判断；像熟练厨师察觉味道不对，产品人也可能先感到版本偏离，而仪表盘尚未报警。
曲凯没有让这个观点轻易过关：品味可能错，也可能比市场早；即使某种准则下好，商业结果也可能不好。范皓宇承认必须承担判断后果。编辑推论：把“感觉不对”写成具体假设——哪个场景、哪类人、哪段体验受损——再补用户原话、失败案例和长周期分群。指标不能代表全部体验，直觉也不能因此免除复核责任。
4. “既要又要”需要现实底座，软件同样有改不动的东西
节目线索：47:47–57:15。 范皓宇认为理想主义要有商业生存：成本、可用、技术、易理解和长期意义应同时考虑。这是设计寻找更好组合的任务，并不是无限加功能、取消资源取舍。到“有用、有趣只能选一个”的追问，他倾向砍掉一些事，保留能兼具两者的方案。
主持人认为硬件难改、反馈慢，嘉宾提醒软件也有架构和用户习惯。软件像家里的物件，换位置会影响生活；硬件则更像房子本身。软件能回滚，不意味着每次改变无代价。入口、记录与既有用法一旦形成，迁移和解释也是交付的一部分。
找需求时，他不主张直接问技术名词。问卖烤红薯的人要不要AI助理，可能得到“不需要”；对方真正想要的是卖得更好。先有可被推翻的假设，再接近用户，验证问题、方案和接受程度，收入支持下一轮尝试。他把双方比作朋友：可以交流、反对、互相理解，但这不能替企业免除服务承诺和商业责任。
5. “四个声音”：把评审从第一次购买拉到换代
节目线索：59:54–01:02:00。 这套说法是范皓宇的体验隐喻，拟声词在转写中不稳定；值得保留的是四个阶段，而非强造口号。
第一，在一堆相似产品里看见不同，愿意靠近。第二，初用兴奋而顺畅，他以买到iPhone后装应用作例子。第三，用了数月仍发现新能力，例如后来装上新的玩法。第四，换新设备时，旧生活、资料与习惯顺利回来。惊喜不是每次更强刺激，而可能是少一次重新设置的麻烦。
编辑推论：这为长期产品补齐四类证据：发现、首次价值、持续探索、迁移。广告点击高只能说明第一段的一部分；新用户演示不能代表几个月后的持续价值；没有真实迁移，不能宣布第四段完成。“做到几个声音就能卖”的话适合作为鼓励，不是销售预测公式。
6. 有趣来自共同使用，AI也要求更完整的交付责任
节目线索：01:06:22–01:20:18。 打火机被把玩、扇子被题字、电脑被贴贴纸，说明人会把贴近自己的物件变成个人表达。范皓宇不愿把有趣机械拆成若干情绪刺激档位，更愿从生活的小片段中找创意，再与用户磨合。
这种机制可以启发参与式设计，却不能保证每件物品都会给每个人带来意义。节目中的物理比喻与未具名心理实验，不作为科学证据。编辑推论：观察用户自己加了什么、怎样分享、哪些玩法持续存在，往往比先规定“这里必须感动”更有帮助；用户也应能拒绝某种互动。
尾声回到组织。范皓宇讲到面试设计师时，对方用AI展示了含账号注册的完整工具，区别于只交付图标或需求文档。他认为AI会推动端到端工作，即对从需求到结果的整段任务负责。这不等于所有专业协作消失，也不意味着一个人无条件包办全部。最终问题、动作授权、失败处理与结果验证，仍要有明确负责人。
用四个阶段做一次产品回访
以下为编辑练习，不是理想汽车的公开评测方案。找不同使用阶段的人，而非只请新用户打分：
- 请第一次接触者说清哪里不同、预期会得到什么，再观察首次真实任务是否匹配承诺。
- 请使用三个月的人演示最近一次主动发现的新价值，分辨真实能力与新鲜感。
- 请换机、换账号或版本迁移的人重做任务，记录数据、习惯和设置怎样回来。
- 把“有品味”“不好用”都改写成一个具体片段，并列成本、可靠性与用户解释；决定保留方向、修改实现或停止。
这适合长期服务、创作或生活产品。一次性刚需工具可以有更窄的目标，有趣未必是购买必要条件；不要为了套方法给所有工具加人格。
回听导航与关联阅读
- 13:56：不倒翁指纹仪。
- 18:20：数据反馈与刺激循环。
- 32:26：代际实用主义的争论。
- 42:03：品味与数据冲突。
- 53:43：从假设接近真实需求。
- 59:54：四阶段的体验审阅。
- 01:17:47：AI与完整交付组织。
相关主题：产品方法论、长期体验、用户研究。可比较Albert的评估与经营框架，以及个人模型怎样理解个人。
来源边界：完整阅读所提供转写、旧精读和官方当期介绍后重组，未逐句复听。姓名与职务按官方介绍；不扩写转写含混履历。城市分层、消费与心理解释为个人观点；AI对话比例与角色变化是未来设想；未据此核实理想具体产品实现或经营结果。片尾歌曲不作访谈观点。
English edition
Fan Haoyu on Product Completeness: Look Beyond Metrics to How People Live with Products
42 Zhangjing episode analysis · Product design and lasting experience
Guest: Fan Haoyu, senior vice president of Li Auto, as introduced at the time | Host: Qu Kai
Published: October 11, 2025 | Duration: 82 minutes 21 seconds | Prepared: October 18, 2026
Official episode and listening link
Why a useful feature may still fall short of a good product
Fan describes approximately 25 years in products and design across industrial design, phones, games, media, operating systems and cars. He identifies a common capability: observe, understand, combine and compress, turning numerous aspects of life into a coherent solution people can directly experience.
The official premise is a conversation about products and life, not disclosure of Li Auto’s business. Discussions of cities, generations, taste and AI’s future remain personal observations and hypotheses from 2025, not representative surveys. This analysis asks how they change product reviews while preserving the host’s challenges about taste being wrong or ahead of its market.
1. A self-righting fingerprint reader: Combine small details from real life
Episode reference: 13:56–21:28. Fan’s favorite early design was a fingerprint reader for a research office. Documents covered desks and could bury a small device. He made it a water-drop-shaped, self-righting object: shifting papers would reveal it through movement, while people could also play with it. Practical discovery and a small office pleasure shared one form.
This illustrates compression. Rather than listing ten features and satisfying them sequentially, find one solution addressing several conditions. Games, content and hardware can all begin with collecting material and connecting relationships before producing an answer. His concern about rushing to A/B tests is that weak hypotheses prematurely restrict alternatives, not that all experiments lack value.
He also describes a stimulation cycle: users adapt, suppliers escalate and both may become tired. This interpretation comes from conversations with creators, not universal causal research. Editorial inference: Alongside clicks and time spent, ask whether users finish tasks more easily, value the time they spent and experience extra burdens from continual changes.
2. Emotional value can mean different things to parents and children
Episode reference: 26:15–36:07. Qu uses parents breaking pizza down into flour and meat to illustrate practical thinking: why should it cost so much? Fan offers a counterexample. Older generations also entertain friends, buy good alcohol and cherish old objects or photographs. Emotional value is present but attached to different meanings.
This disagreement is more useful than labeling young people emotional and older people practical. Saving in everyday life and spending on special occasions need not conflict. One purchase cannot reveal a person’s entire value system. The same feature can mean different things at work, with family or alone.
Fan also uses opportunities, relationships and service availability in different cities as an observational framework. It reminds designers that their lives do not represent everyone, without establishing that big-city residents are inevitably lonely or smaller-city residents secure. A survey percentage retold without its report is not adopted as social fact. Editorial inference: Ask users to explain an actual decision, who was present and why it mattered, instead of assigning motives from age or city.
3. When taste and metrics disagree, make judgments specific and own the consequences
Episode reference: 39:42–47:46. Fan understands taste through distinction, style and content. A product becomes visible through continuing choices about experience, attitude and exclusions, beyond a feature list. Price and visual attractiveness are insufficient definitions.
When metrics conflict with feeling, he trusts experience gained through intensive use and contact with users. He argues that selecting indicators and categories already involves judgment. Like a practiced chef noticing an off flavor, a product designer may sense a version drifting before a dashboard shows an alarm.
Qu challenges this directly: taste can be mistaken or ahead of demand; a product good under one criterion may perform poorly commercially. Fan accepts responsibility for the result. Editorial inference: Turn discomfort into a specific hypothesis identifying the situation, people and damaged experience, then add direct feedback, failure examples and longer-term segmentation. Metrics cannot represent every experience, but intuition still requires scrutiny.
4. Wanting several things requires a practical foundation; software also resists change
Episode reference: 47:47–57:15. Fan grounds idealism in commercial survival. Cost, usefulness, technology, clarity and lasting meaning need simultaneous consideration. Design should seek better combinations, rather than unlimited features without resource choices. Asked to choose usefulness or interest, he would remove some activities and retain combinations offering both.
The host contrasts hard-to-change hardware with software. Fan points to software architecture and habits. Software resembles objects in someone’s home: moving them disrupts life; hardware resembles the house itself. Rollback does not make changes costless. Established entry points, records and routines make migration and explanation part of delivery.
For research, he discourages asking about technology terms directly. A sweet-potato seller may reject an AI assistant while wanting to sell more successfully. Begin with a falsifiable hypothesis, approach users and test the problem, solution and acceptance; revenue supports the next attempt. He likens the relationship to friendship, allowing disagreement and mutual understanding. That analogy cannot remove service promises or commercial responsibility.
5. Four “sounds”: Extend review from discovery to replacement
Episode reference: 59:54–01:02:00. This is Fan’s experience metaphor. Transcribed exclamations are unreliable, so the useful elements are four stages rather than invented slogans.
First, something stands out among similar products and invites approach. Second, initial use feels exciting and smooth; he gives installing apps on a new iPhone as an example. Third, months later, new capabilities still emerge, such as discovering a new application or activity. Fourth, on replacing the device, the old life, information and habits return smoothly. Delight may mean one fewer setup problem, rather than stronger stimulation.
Editorial inference: Collect evidence for discovery, first value, continued exploration and migration. Advertising clicks cover only part of discovery. A new-user demo cannot establish value months later. Without real migration, the fourth stage remains unproven. Claims that a certain number of sounds predicts sales are encouragement, not a forecasting formula.
6. Interest emerges through shared use, while AI expands responsibility for delivery
Episode reference: 01:06:22–01:20:18. Playing with lighters, writing on fans and adding laptop stickers show people transforming nearby objects into personal expression. Fan prefers discovering small life details and developing them with users to dividing interest into mechanical levels of emotional stimulation.
This can inform participatory design without guaranteeing meaning for everyone. Physical analogies and an unnamed psychology experiment are not treated as scientific evidence. Editorial inference: Observe what users add, share and sustain themselves, rather than first prescribing where they must feel moved. Users should also be able to reject an interaction.
The closing conversation returns to organization. Fan describes a design candidate using AI to demonstrate a complete tool, including account registration, rather than only icons or a requirements document. He expects AI to encourage responsibility across the entire path from need to outcome. That does not eliminate specialist collaboration or require one person to do everything. Defining problems, authorizing actions, handling failure and verifying results still require clear ownership.
Revisit a product across four stages
This is an editorial exercise, not Li Auto’s published evaluation. Recruit people at different stages rather than asking only newcomers to score the product:
- Ask a first-time participant what seems different and what they expect, then observe whether the first real task matches the promise.
- Ask someone with three months’ experience to demonstrate a recently discovered source of value, distinguishing capability from novelty.
- Ask someone who changed devices, accounts or versions to repeat a task, recording how information, habits and settings return.
- Rewrite “tasteful” or “hard to use” as a concrete episode, alongside cost, reliability and user explanations. Decide whether to preserve direction, change implementation or stop.
This suits continuing services, creative products and products embedded in everyday life. A one-off utility can have a narrower purpose. Interest need not be a purchase requirement, and the framework does not require adding personality to every tool.
Listening guide and related reading
- 13:56: The self-righting fingerprint reader.
- 18:20: Metrics and stimulation cycles.
- 32:26: Debate about generational practicality.
- 42:03: Taste conflicting with data.
- 53:43: Approaching real needs through hypotheses.
- 59:54: Reviewing four experience stages.
- 01:17:47: AI and complete delivery organizations.
Related themes: product methods, lasting experience and user research. Compare Albert’s evaluation and business framework and personal models understanding individuals.
Source boundary: Reorganized after reading the complete supplied transcript, prior analysis and official introduction, without sentence-by-sentence audio checking. Names and roles follow official material; unclear career details are not expanded. City distinctions, consumption and psychological explanations remain personal views. Future AI conversation proportions and changing roles are scenarios, not verified adoption. No specific Li Auto implementation or business outcome is established. Closing music is not treated as interview opinion.