一个产品可以有很多用户，但依然不是一个 business。真正重要的不是有多少人来，而是他们为什么来、来了以后能获得什么、以及他们为什么愿意为此付费。
最近我在重新思考一个很基础的问题：business 的目的到底是什么？
一个很朴素的答案是：把产品或者服务卖出去。
但这句话背后其实有很多更底层的问题：
- 我们到底在卖什么？
- 用户为什么要买？
- 用户买完以后获得了什么结果？
- 这个结果为什么值得付费？
很多时候，我们在这些问题还没有想清楚的时候，就已经开始开发产品 feature 了。做更多页面、更多按钮、更多流程、更多体验优化。开发 feature 很容易让人产生进展感，因为它看得见、摸得着，也可以被展示。
但问题是：feature 不等于 value。
如果用户不知道自己为什么需要这个东西，如果我们也不知道自己到底在卖什么，那么再多 feature 也只是复杂度。
有用户，不代表商业成立
一个产品有一万个人来用，听起来很不错。但如果这一万个人只是来看看、玩一玩、凑个热闹，而我们不知道他们到底想买什么，也不知道他们为什么会留下来，那这一万个人的商业价值可能并不高。
真正有价值的不是“有人来”，而是：
- 用户有明确的问题；
- 我们提供了一个清晰的解决方案；
- 用户理解这个解决方案的价值；
- 用户愿意为这个价值付费；
- 付费以后，用户获得了可感知的结果。
所以，做 business 最重要的一个问题不是：
我们还能做什么功能？
而是：
用户为什么会付钱？
这个问题如果没有答案，产品越复杂，反而越危险。因为复杂度会掩盖商业模型的不清晰。
Feature Trap
我觉得创业者很容易陷入一个陷阱：当商业问题没有被回答时，就用产品开发来逃避。
比如：
- 用户不转化，那是不是 onboarding 不够好？
- 用户不留存，那是不是 community 功能不够强？
- 用户不付费，那是不是需要更多高级功能？
- 用户不活跃，那是不是 UI 不够顺？
这些问题都可能是真的。但它们不是第一性问题。
第一性问题应该是：
这个产品创造的经济价值是什么？
如果这个问题没想清楚，优化用户体验可能只是让用户更顺滑地使用一个他们不会付费的东西。
用一个 AI-native trading 项目做例子
我最近在做一个 AI-native trading 相关的项目。表面上看，我们可以思考很多产品问题：
- 平台体验怎么做得更好？
- 用户来了以后应该看到什么？
- AI agent 怎么展示？
- leaderboard 怎么设计？
- 研究和回测功能怎么做？
- 怎么让用户更愿意参与？
这些都重要，但它们不是最先要回答的问题。
更重要的问题是：
- 用户进入这个系统以后可以干嘛？
- 这个行为为用户创造什么价值？
- 这个行为为平台创造什么价值？
- 用户为什么会为这个价值付费？
如果一个 AI trading 产品只是看起来很酷，它可能会吸引一些好奇的人。但好奇心不一定等于付费意愿。
所以我现在更倾向于从另一个角度理解这类产品：
它的核心价值不应该只是“多一个交易工具”，而应该是帮助用户把市场想法变成可以被验证、追踪和复盘的东西。
换句话说，我真正关心的不是平台有多热闹，而是它能不能承载一个更清楚的价值链：从想法，到研究，到验证，到持续反馈。
这件事的价值不在于“又多了一个交易平台”，而在于：
我们能不能让一个模糊的判断，变成一个可以被系统化检验的假设？
我们到底在卖什么？
如果我们说自己只是在卖“回测服务”，这个定位其实很弱。因为市场上已经有很多免费或者低成本的工具。用户也可以用各种成熟平台和 open-source framework 自己做测试。
所以问题不是：
我们能不能做回测？
而是：
我们提供的东西，为什么不是一个普通回测工具？
我现在的答案是：更有价值的部分不只是工具，而是从一个模糊 trading idea 到 tested strategy 的研究过程。
用户可能一开始只有一个很模糊的想法：
如果某个资产连续下跌后反弹，是不是有 edge？
或者：
某个宏观事件发生以后，哪些资产更容易出现趋势？
传统工具给用户的是一把锤子。用户还需要自己知道怎么设计实验、怎么找数据、怎么写代码、怎么评估结果、怎么避免过拟合、怎么追踪结果。
但我更感兴趣的是一个完整的研究流程：
- 把模糊想法变成可测试 hypothesis；
- 找到相关数据；
- 生成实验和回测；
- 做参数和风险分析；
- 输出可以被复盘的结论；
- 持续追踪这个结论在新数据里的表现。
这和普通 backtesting tool 的区别在于：工具只是能力，研究流程才是结果。
用户真正想买的，也许不是“我可以点一个按钮跑回测”，而是：
我想知道我的交易想法到底靠不靠谱。
Platform 只是价值交换的载体
这也是我最近对 AI-native platform 的一个新理解。
一个 platform 本身不是 business。真正的 business 是 platform 促成的价值交换。
对一个 AI-native research system 来说，这个价值交换可能是：
- 用户提出一个市场假设；
- 系统帮助用户研究和验证；
- 结果变成一个可以被追踪和复盘的结论；
- 用户因此获得更高质量的反馈和决策依据。
这时候，平台不是单独存在的产品。它是研究、验证、展示和反馈的载体。
这个定位比“AI trading platform”更清楚。因为它回答了用户到底获得什么 transformation：从一个无法验证的想法，变成一个可以被研究、测试和追踪的判断。
付费渠道是现实检验
这也是为什么我现在觉得，支付系统不只是一个技术任务。
表面上，支付系统只是收钱。实际上，它是一个现实检验：
用户是否愿意为我们定义的 value exchange 付费？
如果用户说这个东西很酷，但不愿意付费，那说明价值还不够清楚，或者用户不是正确用户。
如果用户愿意付费，但支付路径很混乱，说明我们在商业交付上没有闭环。
如果用户付了费，但不知道自己买了什么，说明 offer 没有定义清楚。
所以在做更多功能之前，我更想先验证一件小事：
是否有人愿意为“把一个想法变成一个被研究、测试、解释过的结果”付费。
重点不是价格本身，也不是产品包装，而是验证：这个 transformation 对用户到底重不重要。
我现在的判断
我越来越觉得，早期创业最危险的事情，不是产品做得不够多，而是没有想清楚自己到底在卖什么。
用户不会为 feature 付费。用户为结果付费。
用户不会为平台愿景付费。用户为自己获得的 transformation 付费。
所以在开发更多功能之前，我想先回答一个问题：
用户买完以后，世界发生了什么变化？
对这个方向来说，我现在的答案是：
用户从“我有一个交易想法”，变成“我有一个经过研究、测试、解释，并且可以被持续追踪的判断”。
如果这个 transformation 足够重要，business 就有机会成立。
如果这个 transformation 不重要，那么再多用户、再多 feature、再漂亮的 UI，都不能让它变成一个真正的 business。
这可能是我最近对产品和商业最重要的一个提醒：
在问“我们还能做什么”之前，先问“用户为什么会买”。
A product can have lots of users and still not be a business. What matters isn’t how many people show up, but why they come, what they get once they’re there, and why they’d be willing to pay for it.
Lately I’ve been rethinking a very basic question: what is a business actually for?
The plain answer is: to sell a product or a service.
But underneath that sentence there are a lot of more fundamental questions:
- What are we really selling?
- Why would users buy it?
- What result do users get once they’ve bought it?
- Why is that result worth paying for?
A lot of the time, we start building product features before we’ve thought these questions through. We add more pages, more buttons, more flows, more tweaks to the experience. Building features easily gives you a sense of progress, because it’s visible, tangible, and something you can show people.
The problem is: features are not value.
If users don’t know why they need the thing, and we don’t know what we’re really selling either, then more features just add complexity.
Having users doesn’t mean the business works
Ten thousand people using a product sounds great. But if those ten thousand people are just looking around, playing with it, or joining in because everyone else is, and we don’t know what they actually want to buy or why they’d stay, then their commercial value may not be very high.
What’s really valuable isn’t “people showing up.” It’s that:
- users have a clear problem;
- we offer a clear solution;
- users understand the value of that solution;
- users are willing to pay for that value;
- after paying, users get a result they can actually notice.
So the most important question in building a business isn’t:
What other features can we build?
It’s:
Why would users pay?
If this question has no answer, the more complex the product gets, the more dangerous it becomes, because complexity hides an unclear business model.
The feature trap
I think founders fall into a trap easily: when the business question hasn’t been answered, they escape into product development.
For example:
- Users aren’t converting, so maybe onboarding isn’t good enough?
- Users aren’t sticking around, so maybe the community features aren’t strong enough?
- Users aren’t paying, so maybe we need more premium features?
- Users aren’t active, so maybe the UI isn’t smooth enough?
Any of these could be true. But none of them is the first-principles question.
The first-principles question should be:
What economic value does this product create?
If that isn’t clear, improving the user experience may only help users move more smoothly through something they’ll never pay for.
An AI-native trading project as an example
I’ve recently been working on a project in AI-native trading. On the surface, there are lots of product questions we could think about:
- How do we make the platform experience better?
- What should users see when they arrive?
- How should the AI agents be presented?
- How should the leaderboard be designed?
- How should the research and backtesting features work?
- How do we make users more willing to take part?
These all matter, but they aren’t the first questions to answer.
The more important questions are:
- What can users do once they’re in the system?
- What value does that create for the user?
- What value does it create for the platform?
- Why would users pay for that value?
If an AI trading product just looks cool, it may attract some curious people. But curiosity doesn’t necessarily mean willingness to pay.
So these days I lean toward seeing this kind of product from a different angle:
Its core value shouldn’t just be “one more trading tool.” It should be helping users turn market ideas into something that can be verified, tracked, and reviewed.
In other words, what I really care about isn’t how lively the platform is. It’s whether it can carry a clearer value chain: from idea, to research, to validation, to ongoing feedback.
The value of this isn’t “yet another trading platform.” It’s in this:
Can we turn a vague hunch into a hypothesis that can be tested systematically?
What are we really selling?
If we say we’re just selling “backtesting as a service,” that positioning is actually weak. There are already plenty of free or cheap tools on the market, and users can run their own tests on established platforms and open-source frameworks.
So the question isn’t:
Can we do backtesting?
It’s:
Why isn’t what we offer just an ordinary backtesting tool?
My answer right now is that the more valuable part isn’t just the tool. It’s the research process that takes a vague trading idea to a tested strategy.
A user might start out with nothing more than a vague idea:
If an asset rebounds after falling several days in a row, is there an edge?
Or:
After a particular macro event, which assets are more likely to trend?
Traditional tools hand users a hammer. Users still have to know how to design the experiment, find the data, write the code, evaluate the results, avoid overfitting, and track the results over time.
What interests me more is the whole research process:
- turn a vague idea into a testable hypothesis;
- find the relevant data;
- generate the experiments and backtests;
- analyze the parameters and the risk;
- produce a conclusion that can be reviewed;
- keep tracking how that conclusion holds up on new data.
The difference from an ordinary backtesting tool is that a tool is only a capability, and the research process is the result.
What users really want to buy may not be “I can click a button and run a backtest.” It may be:
I want to know whether my trading idea actually holds up.
A platform is just the vehicle for exchanging value
That’s also something new I’ve come to understand about AI-native platforms.
A platform by itself isn’t a business. The real business is the exchange of value the platform makes possible.
For an AI-native research system, that exchange of value might be:
- a user proposes a market hypothesis;
- the system helps them research and validate it;
- the result becomes a conclusion that can be tracked and reviewed;
- the user gets better feedback and a better basis for decisions.
At that point, the platform isn’t a standalone product. It’s the vehicle for research, validation, presentation, and feedback.
That positioning is clearer than “AI trading platform,” because it answers what transformation the user actually gets: from an idea they can’t verify to a judgment that can be researched, tested, and tracked.
Payment is a reality check
That’s also why I now think a payment system isn’t just a technical task.
On the surface, a payment system just collects money. In practice, it’s a reality check:
Are users willing to pay for the value exchange we’ve defined?
If users say it’s cool but won’t pay, the value isn’t clear enough yet, or they aren’t the right users.
If users are willing to pay but the path to paying is a mess, we haven’t closed the loop on delivering the business.
If users pay but don’t know what they bought, the offer wasn’t defined clearly.
So before building more features, I’d rather validate one small thing first:
Whether anyone is willing to pay for “turning an idea into a result that has been researched, tested, and explained.”
The point isn’t the price itself, or the packaging. It’s finding out whether this transformation really matters to users.
Where I’ve landed for now
More and more, I think the most dangerous thing in an early-stage startup isn’t building too little product. It’s not being clear about what you’re actually selling.
Users don’t pay for features. Users pay for results.
Users don’t pay for a platform’s vision. Users pay for the transformation they get.
So before building more features, I want to answer one question first:
After users buy, what has changed in the world?
For this direction, my answer right now is:
Users go from “I have a trading idea” to “I have a judgment that has been researched, tested, and explained, and that I can keep tracking.”
If that transformation matters enough, there’s a chance for a real business.
If it doesn’t, then no number of users, no amount of features, and no amount of UI polish will turn it into a real business.
This may be the most important reminder I’ve had lately about products and business:
Before asking “what else can we build,” ask “why would users buy.”