在多伦多公共图书馆的信息之海

---

产品经理如何做出正确的产品决策

by  Brandon Chu （shopify 产品总监）

翻译：Kevin嚼薯片

虽然产品经理可能不会在一线去构建产品，但他们还是会为团队带来一样东西：产品决策。

这些决策可能包括任何方面：小到的文档里的一条线的拷贝摆放，大到一个新功能的MVP应该是什么。

产品经理做的产品决策是为了给团队成员排除障碍，让产品得以继续开发。他们不需要亲自做每一个决策，但他们需要确保每一个决策都有人去做——无论是他们自己、他们的团队成员、还是利益相关者。

产品经理的任务是对坑不确定性，这是我们工作的独特之处，因为我们往往是整个公司中了解最多用户场景的。通常，产品经理需要做的最重要的决策包括：

1. 团队存在的原因（愿景及期望的影响力）

2. 实现愿景的具体方法（策略）

3. 要做的事情，以及不做的事情（项目优先级）

4. 能构建的最小价值产品（MVP）

因为我们之前已经讨论过产品决策本身，并且介绍过一些失败决策的事例，在这篇文章中，我们将关注普遍的问题：对于产品经理来说，如何成为好的决策者？

当一个好的决策者

直觉上，做对很多事情。与此同时，也有一些结果在我们控制之外。

也许你正在开发一个iOS应用，然后苹果改变了他们的服务条款，这样你的应用程序就被下架。也许是一条法律通过，使你的服务在最大的用户市场上变成了非法的。

这些情况都是突然发生的，很难预测，会击毁你的任何决定。在这样的环境下，如何客观地评定什么是好决策？

我认为做一个好的决策者意味着以下两件事：

1. 用正确的信息量去做决策

2. 尽快做出决策

使用正确的信息量做出正确的决策

决定一个决策的重要性，是你需要做出的最重要的决策。对于那些在生活中做出决策的人来说，理解一个决策什么时候重要与否，是最为关键的技能。

在《亚马逊(Amazon)2016年的股东信》中，杰夫·贝佐斯(Jeff Bezos)用他所谓的“第一类”和“第二类”决策来描述这一概念。

第一类决策是不可逆转的，你必须非常小心地做出决定。

第二类决策就像走过一道门——如果你不喜欢这个决定，你可以退回去。

贝佐斯在信中所指出的是，一个决策会影响着你实现一件事情需要付出的努力。你可能99%的时候都是正确的，但如果你在那1%的时候错了，你就不是一个有效的决策者。所以当成本很高时，你应该更努力地做出正确的决策。

虽然我不认为“第一类”和“第二类”决策是相对的，但我确实认为它们代表了一个确定“决策重要性”的框架的图谱：

把你决策的每一个维度放在这个图表上，你就能看到它的重要性

这个图表框架将决策的重要性分解为三个维度：

1. 决策导致的资源投入

2. 决策正向结果的影响

3. 决策负面结果的影响（这对你思考更有帮助，但子维度可能是不太一样的）

这里的想法是沿着上述的图表框架去分解一个决策，以确定整体的重要性。我们举三个例子。

首先，把主页上的“注册”由原来的文本链接变更为一个按钮。

这显然是一个低重要性的决策。如果你错了，很容易回退。负面结果是影响了在变化期的用户。而正向结果是微小渐进的。

注意，这并不意味着它不值得做，这意味着这个决策做不做并不是很重要，我们稍后会看到，这会影响我们去做决策的速度。

另一个例子：变更公司的logo

这是明显是一个非常重要的决策。它可能不需要太多的资源来执行，但对公司的品牌和客基础的影响（正向和负面）都是巨大的。还记得吗，优步改变了他们的logo后，没人能找到这个应用。还有AirBnB改了logo后，总有人觉得看到生殖器的形状。

还有一个例子：“登月计划”，为公司打开一个新的市场

这个特别的决策是中偏低的重要性。重要的是它会消耗巨大的物质资源，但即使失败也没关系。

对于那些已经达到规模并正在寻找新增长方式的科技公司来说，这是一个常见的决策类型。高额投资+失败后的低负面影响+成功后的高正向风险/回报比率，这对那些有资源的公司非常有吸引力，以至于他们经常创建专门的团队来重复试验这些项目。像Facebook 8号楼，和谷歌X。

这个图表框架，及它的延伸框架，将帮助你创建比较决策的基础，并将它的重要性展示给每个人看。根据公司的情况去调整它，并认识到定义决策的重要性，可以帮助你做决策。所以需要首先建立这个框架。

一旦你知道了决策的重要性，你就应该调整为此而花费的时间。

我有一个看法，那就是好的决策者也是最迅速做出决策的人。以下是几点原因：

1. 重要性越低的决策，你做决策所需要获取的信息就越少。

2. 收集信息遵循“帕累托原则”，这意味着你可以很容易地获得80%的信息，但是获得最后的20%需要付出很大的努力。

3. 大多数的决策都偏向不重要。

让我们来一个一个分析。

1. 重要性越低的决策，你做决策所需要获取的信息就越少。

产品，尤其是许多人使用的产品，是一个复杂的系统，在那里预测决策的所有结果几乎是不可能的，即使你有100%的可用信息。由于我们无法完全预测结果，这意味着所有的决策都是一场赌博。

这赌博基于我们对正向结果的信心，而信心基于我们所拥有的信息量。我们获得的信息越多，我们的信心水平就越高。

“信息”在这个背景下可以是与决策相关的。例如之前的注册按钮示例，我们可以研究其他站点的按钮模式，以获得一个积极的正向结果。

然而，关键问题是：在做出决定之前，应该积累多少信息量？

不管有意与否，很少有人在做出决策之前会收集到100%的可用信息。最终的结果是，在某一时刻，每一个决策者都下意识地认为：“我百分之七十肯定增加了这个绿色的按钮，会增加转化。”  然后他们就做了。

这揭示了一个人实际上是如何做出决定的：首先，他们评估他们的信心水平(70%)，其次，他们定义了做出这个决定所需的信心阈值。

这个信心阈值基于这个决策的重要性。你可能愿意有60%信心的情况下添加一个注册按钮，它会产生一个积极的结果，但你不会愿意只有60%信心的情况下改变公司的logo。后者是一个更重要的决定，因此它的信心阈值应该更高。

总结一下：

- 你对决策结果预测的信心程度取决于你获得多少信息量

- 一个决策越重要，信心阈值就越高，因此你需要的信息量就越多

我们可以在下面的模型中将这些点图形化：

这揭示了一些关于决策的违反常理的现象：你的目标不是总是做出正确的决定，而是根据决策的重要程度去投入合适的时间。

这是有好坏两面的。坏的一面是，你可能最终会做出错误的决策，但它们可能并不重要。好的一面是，你最终会做出更多决策，这意味着你会创造更多价值。

会增加多少决策？这就是下一节的内容。

2. 收集信息遵循“帕累托原则”

收集35%和收集70%的信息，在付出上有什么不同？它不是2倍的关系——大多数时候你已经有了大量相关信息：关于过去用户行为的数据，以前的研究数据，公司以前的项目等等。

帕累托原则在这里非常适用，你通常可以付出很小的成本就能得到80%的信息，而获得100%的信息则需要付出很大的成本。

假设你是Facebook的产品经理，在决定是否要进入P2P市场。正如Craiglist,Letgo，和其他一些人所证明的，你知道这成功机会很大，所以你开始在Facebook上开始这一个需求。

你评估它的重要性，并得出“中等重要性”，因为需要资源投资，以及在Facebook上进行P2P交易可能的潜在法律风险。

你会收集哪些信息来做这个决策？

也许你会首先看的是Facebook群组里现有的用户行为。即使没有特定的功能，但用户就已经在群组里这样做了吗？你可以分析多个群组和发布在上面的内容，以确定有多少用户已经在进行P2P交易。

通过查看群组中的文章，你也可以弄清楚正在做的交易是什么，用了什么样的图片组合，以及没有支付功能时交易是怎么进行的。为了解决法律风险，你也可以与法律团队一起研究这个对公司的影响。所有的这些信息都是现成的。

你可能不太容易弄明白的是：为什么人们选择使用Facebook而不是现有的其他产品？为了找到答案，你可能需要与用户直接进行定性研究，这将花费大量时间。

这真的需要吗？这额外的时间值得花吗？

可能不值得。已有的信息已经说明了80%的可能性用户是否想要这个功能，如果公司能够接受法律风险的影响。虽然弄明白用户使用Facebook来进进行P2P交易是设计这个功能的重要信息，但实际上并不是做决策时真正需要的。

将其归纳为一个模型，我们可以说，收集信息的时间（成本）遵循帕累托原则：

请注意，这条曲线与我们在前一节信息量于决策重要性的曲线是基本相同的。

将这两条曲线合并在一起，我们可以用做决策速度和重要程度去做函数建模。

这个整合模型揭示了一些重要的东西。你的决策速度和其重要性不是线性相关的，它们是指数相关的。因此错误评估一个决策的重要性意味着你会浪费大量时间来收集多余的信息。

3. 大多数决定并不重要

下面的模型展示了，决策重要性遵循了一个逆幂定律，即绝大多数决定并不重要。

为了节省时间，我断言这个模型是普遍的，直观的，真实的。

我们整天在产品公司做决策——但它们并不都是同等重要的，坦率地说，大多数都是微不足道的。我们应该使用逗号，还是不用？我们应该推广给5%的用户，还是10% ？是开午餐会议，还是赶紧吃完午饭然后在15点开会？

放宽心吧！

让我们回顾一下这三个结论：

1. 重要性越低的决策，你做决策所需要获取的信息就越少。

2. 收集信息遵循“帕累托原则”，这意味着你可以很容易地获得80%的信息，但是获得最后的20%需要付出很大的努力。

3. 大多数的决策都偏向不重要。

因此，绝大多数决策应该迅速做出。整合最后两个模型，我们能清楚地证明这一点。

好的决策者，是能快速做决策的人。

做一个优秀的决策者

在实践中应用这些原理和框架非常困难。没有人喜欢犯错，而且当一个产品经理犯错会被公开和放大，同时让人受到职场伤害。

当你明白做决策是你作为产品经理的职责和产出时，你就能抑制恐惧，专注于决策所带来的影响。

如果你是产品经理们的上司，要认识到谨慎和总是正确的产品经理通常不会创造很大的影响力产品。影响力大往往是由那些能迅速做出很多决策的人所驱动的，而且对于重要程度高的会做出正确的决策。

按钮的决策应该快速即时做出。Facebook上进行P2P交易的决策需要一些时间来思考，但即使你错也并不是世界末日。

你应该把你的时间花在logo的决策上。

欢迎各位 “打赏” ，支持我继续推送产品运营相关好文章。并留下邮箱，我会赠送最新的《产品经理能力模型地图》。

---

Making Good Decisions as a Product Manager

by  Brandon Chu

While product managers may not build the actual product, they do produce something very tangible for a team:decisions.

July 23 update: in retrospect and from feedback, this framework applies to any role, not just product management.

These decisions can be about anything: small ones like a line of copy in the docs, to big ones like what the MVP of a new feature should be.

The decisions PMs make are the ones that unblock their team so they can continue to build. They don’t need to make every decision, but they are responsible for ensuring a decision gets made — whether by them, their team, or their stakeholders.

Product managers are the hedge against indecision, and it’s uniquely our job because we tend to have the most context in a company. Typically, the most important decisions that a PM makes are, in order:

1. Why a team exists (vision and the impact they aspire to create)

2. What the general approach is to accomplish the vision (strategy)

3. What to build now, and what not to build (project prioritization)

4. What the smallest thing that can be built is that achieves the impact (MVP)

Since we’ve already discussed these decisions before, and observed a really bad one, in this post we’re going to focus on something more general:What does it mean for a PM to be a good decision maker?

Being a Good Decision Maker

Intuitively,being right a lot matters.At the same time, and in contrast, there are often factors outside of our control that determine outcomes.

Maybe you’re building an iOS app and then Apple changes their terms of service so your app is in breach. Maybe a law passes that makes your on-demand services illegal in your biggest markets.

These situations all happen in a flash, are hard to predict, and can negate any good decision you otherwise made. In such an environment, how can one objectively analyze their decisions?

I propose that being a good decision maker means doing two things:

1. Making decisions using the right amount of information

2. Making decisions as quickly as possible

Making good decisions using the right amount of information

Deciding how important a decision is, is the most important decision you can make. For people that make decisions for a living, understanding when one is really important vs. not-that-important isthemost critical skill.

In Amazon’s 2016 shareholder letter, Jeff Bezos touches on this concept with what he calls type 1 and type 2 decisions.

Type 1 decisions are not reversible, and you have to be very careful making them.

Type 2 decisions are like walking through a door — if you don’t like the decision, you can always go back.

What Bezos drives at in the letter is that the impact of a decision guides how much effort you need to put into making it. You can be right 99% of the time, but if you’re wrong the 1% of times when itreally matters,you’re not an effective decision maker. The takeaway is that when the stakes are high, you should work a lot harder at making the right decision.

Although I don’t see Type I and II decisions as binary, I do see them as representing the spectrum for a framework for determining “decision importance”:

Plot your decision against each spectrum to get a gist of the importance of the decision

This framework breaks down decision importance into three dimensions:

1. Resource investment from a decision

2. The overall impact of a positive outcome

3. The impacts of a negative outcome (shown more granularly to help you think it through, but the sub-dimensions could be different for you)

The idea here is to plot a decision along these spectrums in order to determine an overall importance. Let’s take a look at three examples.

First,changing a “Sign Up” text-link on the homepage to a button.

This is clearly a low importance decision. If you’re wrong, it’s very easy to undo. The impact of being wrong is limited only to the number of users affected during the period of change. The impact of being right is only incremental.

Note, this doesn’t mean it’s not worth doing, it means the decision to do it or not isn’t that important, and as we’ll see later that should affect the speed at which we make the decision.

Another example:changing your company’s logo

This is a medium to high importance decision. It might not take a lot of resources to execute, but the impact (both positive and negative) to the company’s brand and customer base is huge. Remember when Uber changed their logo and no one could find the app? Or when AirBnB did and no one could stop seeing genitals ?

One more:a “moonshot” project to open up a new market for the company

This particular decision is medium to low importance. It’s important in that it will consume material resources, but also not important in that failure has no lasting downside.

This is a common decision type for tech companies that have reached scale and are looking for new ways to grow. The high investment + low impact of failure + high impact of success creates a risk/reward ratio that is very attractive to companies that can afford to spend the resources, so much so that they often create dedicated teams to repeatably experiment with these types of projects. Think Facebook’s Building 8, or Google’s X.

This framework, or any variation of it, will create common ground for you to compare decisions and articulate the importance of them to everyone you work with. Tweak it based on your company’s context, and recognize that defining decision importance informs everything else about how you should approach making the decision. So do it first.

Once you decide how important a decision is, you should adjust how long you‘re willing to spend on it.

I have a proposition for you, which is that good decision makers are also the ones who make most of their decisions quickly. To get to that conclusion, we’re going to unbundle the following statements:

1. The less important a decision, the less information you should try to seek to make it.

2. Gathering information follows a Pareto principle, meaning you can get 80% of the information quite easily, but getting the final 20% requires a lot of effort.

3. Most decisions are not important

Let’s go through each of them.

1. The less important a decision, the less information you should try to seek to make it.

Products, especially those used by many people, are complex systems where predicting all the ramifications of a decision are practically impossible, even when you have 100% of the available information.Since we cannot perfectly predict outcomes,this effectively means that all decisions arebetswe are making.

We choose to make these bets based ourconfidencein a positive outcome, and that confidence is based on the amount of information we have. The more information we have as a percentage of all available information about a decision, the higher our confidence level should be.

“Information” in this context can be anything relevant to the decision. Using our previous sign-up button example, we could research button patterns from other sites as information to support a positive outcome.

Here’s the key question, though:How muchinformationshould one be gathering before they make a decision?

Whether consciously or not, few people gather anywhere near 100% of available information before making a decision. What ends up happening is that at some point, every decision maker subconsciously thinks to themselves “I’m ~70% sure adding this green button will increase conversion.”,and then they simply make it.

It’s that line of thought that reveals two things about how decisions actually get made by an individual: first, they assess their confidence level (70%), and secondly, they define a confidence threshold required to make that decision.

That confidence threshold is based on the importance of the decision.You may be willing to add the sign-up button with 60% confidence that it will create a positive outcome, but you wouldn’t change the company’s logo with only 60% confidence. The latter is a more important decision, and therefore the bar for confidence should be higher.

To summarize:

- your confidence level on predicting a decision’s outcome is a function of how much information you have about the decision

- the more important a decision, the higher confidence you require and thus the more information you need

We can visualize these points in the model below:

This reveals something counterintuitive about decision making: your goal shouldn’t be to always make the right decision, it should be to invest the right amount of time in a making a decision relative to its importance.

The ramifications of subscribing to this are two-fold. The downside is that you will end up making morewrongdecisions, but when you do, they’re likely not important. The upside is that you will end up makingmoredecisions, which as a PM means you’re creating moreoutput.

How much more? That’s what the next section is about.

2. Information gathering effort follows the Pareto principle

What’s the difference in effort required for gathering 35% of information versus 70%? It’s not 2x — most of the time you will already have a significant amount of the relevant information close by: data about past user behaviour, previous research studies, previous attempts by the company, etc.

The Pareto principle applies here in that you can usually get ~80% of the information you need with little investment, whereas getting 100% will take a lot of effort.

Say you‘re a PM at Facebook and you’re deciding whether or not to get into the peer to peer buying/selling space. As evidenced by Craiglist, Letgo, and others, you know the opportunity is big so you start exploring the need for such a feature on Facebook.

You assess the importance, and come up with “medium important” because of the resource investment and the potential legal risks of facilitating transactions and commerce on Facebook.

What information would you gather to make this decision?

Well, probably the first place you’d look is the existing user behaviour within Facebook Groups. Are communities of people already doing this without having specific features for it? You can probably analyze the names of groups and the content being posted to them in order to identify how many users are already buying/selling.

By looking at the posts made in the groups, you can also probably figure out what is being sold, the composition of pictures being used, and how a transaction is happening even though there’s no payment features. To address the legal risk, you can also work with your legal team to research the implications to the company. All this information is readily available.

What you probably won’t be able to figure out so easily is:whyare people choosing to use Facebook for this instead of the established marketplaces? To find that out you’ll likely need to do more qualitative research directly with users, which is going to take a lot of time.

Do you need to? Is it worth the extra time?

Probably not. The readily available information provides ~80% of the context for figuring out if people want the feature and if the company can stomach the legal impact. Although knowing why they’re using Facebook for buying/selling will be important knowledge for designing the feature, it’s not really needed to decide whether or not to do it in the first place.

Generalizing this into a model, we can say that the time (effort) to gather information for decisions follows the Pareto principle:

Notice this curve is exactly the same as we showed in the previous section when we showed that the amount of information you need is based on the importance of the decision.

Merging these two curves together, we can actually model the appropriate speed of making a decision as a function of the importance.

The combination of these models reveal something important. The speed of your decision making not linearly correlated to how important you think a decision is, it’sexponentiallycorrelated. Incorrectly assessing a low importance decision as high means you waste a significant amount of time gathering excess information.

3. Most decisions are not important

The model below shows that decision importance follows an inverse power law, where the vast majority of decisions are not important.

To save your time, I’m just going to assert that this model is directionally, intuitively, true.

We make decisions all day working at product companies — they are not all be of equal importance, and frankly most are trivial. Should we use the oxford comma, or not? Should we roll out to 5% of users, or 10%? Lunch meeting, or eat fast and meet up in 15?

Bringing it all home

Let’s review our three statements and the conclusion they form:

1. The less important a decision, the less information you should try to seek to make it.

2. Effort in gathering information follows a Pareto principle. Getting 100% of information is exponentially harder than getting 80%.

3. Most decisions are not important

Therefore, the vast majority of decisions should be made quickly.Overlaying the last two models we covered illustrates this clearly.

Good decision makers, are quick decision makers.

Be a great decision maker

Applying these principles and frameworks are incredibly tough in practice. No one likes to be wrong, and as a PM being wrong is very public and can feel career damaging.

Understanding that decisions are your output as a PM can help you suppress your fears and focus on impact. If you packed boxes for a living, you wouldn’t think packing 3/hr flawlessly is better than a 100/hour with a 2% failure rate. So why think that way about your output?

If you manage PMs, recognize that PMs that are uber careful and always right usually aren’t the ones who are driving the most impact. Impact is driven by those who are driving lots of decisions quickly, and are right when it’s really important to be.

The sign up button decision should be made fast, almost real time. The Facebook buy/sell marketplace decision needs some time to think through, but it’s not the end of the world if you’re wrong.

You should take your time with the logo decision.

英文原文地址：https://blackboxofpm.com/making-good-decisions-as-a-product-manager-c66ddacc9e2b