如果说lean startup讲的是方法论，那么这本就是方法

定位嘛，就是产品经理入门级读物了吧。不过也不用每章都读。

前面的一些方式方法的问题搞清楚，后面分开商业模型，每一类可以选取着读。

数据需要定性还是定量？Initially, you're looking for qualitative data. 在产品还没有庞大的稳定的用户群体时，定性分析是关键。

好数据应该是什么样子的呢？作者定出了一些vanity metrics, they are up-to-the-right and shed no light on anything. 好的数据首先是干净的，然后是可比的，以百分比的形式，衡量增长变化，其次要按照时间划分好，便于明晰某个时间点上，产品的的变化，所导致的用户行为上的变化。理想的情况下，随着产品设计的改变，数据上的统计方法也是一起设计好的，并且数据的结果应该有什么样的产品策略改变，也是清楚的。

但是，让团队内的每一个人，对于产品应该做的事情达成一致就已经很耗费时间了，更别说获取数据的干净程度，开发上能不能有无限的人力，都很难说。假设，干净的数据取来了，也并不意味着得到了因果关系。Usually causations are simple one-to-one relationships. Many factors conspire to cause something. ... You prove causality by finding a correlation, then running an experiment in which you control the other variables and measure the difference. 谁都知道因果关系好，可是现有的科学知识，都只是针对自然世界的因果关系推导。市场，用户，产品这三个都是变数没有特定规律，只能谈经验的东西。作者书里说，"unknown unknowns" are where the magic lives. They lead down plenty of wrong paths, and hopefully toward some kind of "eureka!" moment when the idea falls into place. 看来产品的魅力（也即，从业者们一切痛苦的根源）就在于这个"unknown unknowns"了。为了验证因果关系，测试的方法呢，有：segments, cohorts, A/ B testing, and multivariate analysis.

“You might think that people will play your multiplayer game, only to discover that they’re using you as a photo upload service. Unlikely? That’s how Flickr got started.”

一切都要依靠数据么？“Humans do inspiration; machines do validation.” 用数据的方式可以帮我们找到局部最优，产品经理则需要找到全局最优。过度地依赖数据，很有可能会失去上帝视角。

lines in the sand？书中提到对于数据的统计结果，是要有预期的，并不能拿来数据看一看就做决策。怎么画这道线呢？
1）当40%的用户说，没有你他们会很失望，产品可以进入扩张期。
2）创业之前，至少找15个人，谈谈你的想法，再冲出门去找投资。

太多需求和问题，先做哪个后做哪个？作者用了一个squeeze toy来比喻，按照more stuff，more people，more often，more money，more efficient，从上往下挤。尤其在产品初期，将精力汇聚在一个目标上非常关键。

"UGC is all about turning visitors into creators." 想像一下明天tumblr flickr等等都不再有任何内容更新，大家仅仅根据已有的内容进行互动，就不难发现UGC的关键真的是内容创建，不是别的任何。怎么赚钱呢？“The UGC business might focus on user contribution above all else, but it still pays its bills with advertising most of the time.” “商业模式”是个有魔力的词汇，好像很多人都喜欢创造新的赚钱方法，只要UGC能通过广告赚到足够多的钱，何必为了创新而创新呢。书里面有一个business model flip book，供读者遍历每一种商业模式，看看怎么赚钱适合自己。

我要怎么用数据呢？对于创业公司，数据并不是在探索新机会，提示功能改进，而是检测自己的假设。“In a startup, your business model — and proof that your assumptions are reasonably accurate — is far more important than your business plan. ”看来TDD不仅仅只是程序管理方法，也适用于公司本身。“Deciding what business you’re in is usually quite easy. Deciding on the stage you’re at is complicated. This is where founders tend to lie to themselves. They believe they’re further along than they really are.”

可以参考的内容也不仅限于数字。用户访谈的方法书里面也有介绍。如果公司里面没有专门的用户研究部门，可以看看书里的用研技巧介绍。人们对话时会有认同别人的倾向，我们通过对方的穿着、谈吐，洞悉对方的身份，说话的时候会照顾这种“身份”，更何况直接表明来意的用户访谈。促使用户说出真相也有一些小tricks：1）问相反的问题，看她是不是会说出你想听的话。2）让他付出过多，看他什么时候放弃。

书里对产品工作的不同frameworks做了总结，着重介绍了lean framework，以及何时才能从一个phase移动到下一个phase：

所以说此书还是最适合当工具书。每个phase都有一些干货，实际操作中，同一时间一个产品只会有一个焦点（如果有多个的话，则与书里面提出的OMTM原则相悖，所以真的不用通读此书）。

具体论点不再赘述，只做摘抄：
Empathy:
如前

Stickiness Phase:
Don’t drive new traffic until you know you can turn that extra attention into engagement.
Expect to go through many iterations of your MVP before it’s time to shift your focus to customer acquisition.

Virality Phase:
viral coefficient

Revenue Phase:
Most people’s first instinct when things aren’t going incredibly well is to build more features.
the likelihood that any one feature is going to suddenly solve your customers’ problems is very small.
You’re moving from proving you have the right product to proving you have a real business.
more stuff to more people for more money more often more efficiently
In the Revenue stage, you need to figure out which “more” increases your revenues per engaged customer the most:

Scale Phase:
If the Revenue stage was about proving a business, the Scale stage is about proving a market.
Porter observed that firms with a large market share (Apple, Costco, Amazon) were often profitable, but so were those with a small market share (the coffee shop). The problem was companies that were neither small nor large. He termed this the “hole in the middle” problem — the challenge facing firms that are too big to adopt a niche strategy efficiently, but too small to compete on cost or scale.
Scaling is good if it brings in incremental revenue, but you have to watch for a decrease in engagement, a gradual saturation of the initial market, or a rising cost of customer acquisition. Changes in churn, segmented by channels, show whether you’re growing your most important asset — your customers — or hemorrhaging attention as you scale.

最后一个部分，才是全书的重点：how good is “good”? 同样，也分了不同的商业模式来讨论。

当然用数据分析产品，也离不开产品上每个人的思维模式，数据是客观的，但是解读它的人仍然是主观的。前面细细碎碎一大堆，最后结尾回到了一些经验式的忠告和建议。

PS: "Our Lean Analytics stages suggest an order to the metrics you should focus on. The stages won’t apply perfectly to everyone. We’ll probably get yelled at for being so prescriptive — in fact, we already have, as we’ve tested the material for the book online and in events. That’s OK; we have thick skins." 噗嗤

回头看来，这本的确非常prescriptive，