Welcome back to my channel! since I posted a real match interview question from lift. People told me that it's more helpful than simply sharing framework. I hear you in today's video, let's go over another real matrix interview question. The question is, from this book be cold and conquer is a book about product management interviews. I have found it also very helpful for data science interviews. The original question is on page 124 under the headline, what metrics will look at to evaluate success of a product. The question is, liking is testing a new feature, asking a new user to upload their profile photo during the sign of face. Currently, a new user is asked to upload a profile photo after the sign of process.
欢迎回到我的频道！因为我发布了一个来自Lift的真实比赛面试问题。人们告诉我，这比简单地共享框架更有帮助。我在今天的视频中听到了你的声音，让我们来看看另一个真正的矩阵面试问题。问题是，从这本书来看，《冷酷与征服》是一本关于产品管理面试的书。我发现它对数据科学也很有帮助。采访。最初的问题是在第124页的标题下，“评估产品成功的标准是什么？”问题是，Liking正在测试一项新功能，要求新用户在签名期间上传他们的个人资料照片。目前，新用户在登录流程后会被要求上传个人资料照片。
What metrics do you look at to evaluate success of the future? This is a typical kind of natural questions that can appear in a data science interview. And we have covered a few steps to answer this kind of question in a previous video. When it says testing, most of the time, it refers to every testing, which is one of data scientists, core confidences. In reality, stakeholders rely on the scientists to design every test, analyze the results, and make data informed decisions. The book provides a simple answer, but I don't think it reflects the real interview scenario, because the candidate does not clarify the function of the future. And the candidates comes up with many metrics, which are not very feasible in any tests.
你用什么标准来评估未来的成功？这是数据科学面试中可能出现的一种典型的自然问题。在之前的视频中，我们已经介绍了回答这类问题的几个步骤。当它说测试时，大多数时候，它指的是每一次测试，这是数据科学家的核心信心之一。在现实中，利益相关者依靠科学家来设计每一项测试，分析结果，并根据数据做出明智的决策。这本书提供了一个简单的答案，但我不认为它反映了真实的面试场景，因为候选人没有阐明未来的功能。候选人提出了许多指标，这些指标在任何测试中都不太可行。
I will provide another way to answer this question and I will expand the question to make it more like a data size interview question. Let's get started. Li ping is testing a new feature, asking a new user to upload their profile photo during the sign of face. Currently, a new user is asked to our the profile photo after sign a process. What matrix will you look at to invite the success of the future? Before I begin, I like to make sure I understand the picture. It's ugly, profile photos, optional or ministry is optional. So for the new feature, users could choose to not upload a photo during the sign of process but upload it upwards. Correct? What is the goal of lunching the new feature? What do you think would be the goal? In the current sign of process? It's possible that users forget to upload profile photos up the sign up. The new feature moves up loading photos into the sign process and asking more users to do it. And that's the function of the future. In terms of the business goal, I think profile photo is one important indicator of the profile complaints and more Complete or profile, the more active the user will be after finishing sign up.
我将提供另一种方法来回答这个问题，我将扩展这个问题，使它更像一个数据大小的面试问题。我们开始吧。李平正在测试一项新功能，要求新用户在签名时上传他们的头像。目前，新用户在签名过程后会被要求提供我们的个人资料照片。你会看什么矩阵来邀请未来的成功？在我开始之前，我想确保我理解了这幅图。这是丑陋的，个人资料照片，可选或部是可选的。因此，对于新功能，用户可以选择在签名过程中不上传照片，而是向上上传。正确的？推出新功能的目标是什么？你认为目标是什么？在当前的进程标志中？用户可能忘记在注册时上传个人资料照片。新功能将照片加载到签名过程中，并要求更多的用户这样做。这就是未来的功能。就商业目标而言，我认为个人资料照片是个人资料投诉的一个重要指标，个人资料越完整，用户在完成注册后就会越活跃。
So I think the business goal is to improve user engagement by having more users of their photos. It's also possible that the goal is to detect fraud. Adding a profile photo can add credibility to a new account. I a user with a profile photo is less likely to be a fraudulent user compared with user without a profile photo. To help us maintain focus, i'll use improving engagement as a goal that make sense continue. Give me a second to collect my thoughts. There are three metrics. I would consider. I have two success metrics and one that will metric. The first metric is to major use the engagement, because that is the goal of the teacher. We could use the average number of investment per user per day or the number of post made per user per day. I want to see that the new languages users are more engaged than those in the control group.
因此，我认为商业目标是通过让更多用户使用他们的照片来提高用户参与度。也有可能目标是检测欺诈。添加个人资料照片可以增加新帐户的可信度。与没有个人资料照片的用户相比，有个人资料照片的用户不太可能是欺诈用户。为了帮助我们保持专注，我将把提高参与度作为一个有意义的目标。给我点时间整理一下思绪。有三个指标。我会考虑。我有两个成功指标和一个将指标。第一个指标是主要使用参与度，因为这是教师的目标。我们可以使用每个用户每天的平均投资数量或每个用户每天发布的帖子数量。我想看到新的语言使用者比控制组的人更投入。
The second metric measures whether this feature is useful at all. I will use the percentage of new users who upload their profile photos as a measurement. This metric should be significant, larger than that of the control group. Because otherwise, it indicates that lots of people choose to know their profile photos, which makes feature useless.
第二个指标衡量此功能是否有用。我将使用上传个人资料照片的新用户的百分比作为衡量标准。这个指标应该是显著的，大于对照组。否则，它表明很多人选择知道他们的个人资料照片，这使得功能无用。
Other than the success metric, I suggest we have got your metric to major things that should not degrade in pursuit of this new feature. We could use the percentage of users who drop off in the sign of process. We don't want to see this metric increase at all. It's ideal if the number stay the same or even decreases. Would you like me to come up with more metrics? No, that's good. What if after bonding the test, you see the increase of percentage of people, out of, the profile photos, but the number of people who finish sign up has gone down. I think we need to see the new numbers to make a decision.
除了成功指标之外，我建议我们把你的指标放在主要的事情上，这些事情不应该在追求这个新功能的过程中降级。我们可以使用在进程标志中下降的用户的百分比。我们根本不想看到这个指标增加。如果数量保持不变甚至减少，这是最理想的。你想让我提出更多的指标吗？不，那很好。如果在绑定测试后，你看到个人资料照片中的人的百分比增加了，但完成注册的人数却下降了。我认为我们需要看到新的数字才能做出决定。
Without seeing the numbers is hard to evaluate the business impact. What if the number of invites cents went up by 2 %? But the number of users completely sign up has gone down by 1 %. Both are significant when you say significant. Do you mean we see AP value less than . 05? Okay, it means that engagement has gone up significantly. But the drop in the number of users completely sign up is also significant. This is a tough call. It seems that we make one step forward and one step back. Given that the current objective is to improve engagement, having more engaged users is worth it. In other words, I would recommend lunch in the future. The candidate did a good job clarifying the questions coming up with metrics to measure the success of the future and interacting with interviewer.
在没有看到数字的情况下，很难评估业务影响。如果邀请的数量增加了2%呢？但完全注册的用户数量下降了1%。当你说重要的时候，两者都很重要。你的意思是我们看到AP值小于。05? 好的，这意味着参与度显著提高了。但完全签约用户数量的下降也是显著的。这是一个艰难的决定。似乎我们前进一步，后退一步。考虑到当前的目标是提高参与度，拥有更多参与的用户是值得的。换句话说，我以后会推荐午餐。候选人很好地澄清了问题，提出了衡量未来成功的指标，并与面试官进行了互动。
All of them are important to get the sums up from the interviewer. A few comments I like to share with you. First, there are different metrics you could choose for this question. The question is about testing a new feature rather than building a dashboard to monitor the feature. So you don't need to exhaust them during the interview. For evidence, providing three metrics would be good enough.
所有这些对于从面试官那里得到总结都很重要。我想和你分享一些评论。首先，你可以为这个问题选择不同的衡量标准。这个问题是关于测试一个新功能，而不是构建一个仪表板来监控该功能。所以你不需要在面试中让他们筋疲力尽。作为证据，提供三个指标就足够了。
Secondly, a good metric should be easy to measure so that it's better to provide a time threshold. For example, instead of using the number of invite sent, the candidate uses the average number of invite send per user per day. Lastly, a good metric should also feed the context. If you just simply using a metric framework, such as a popular aarr framework, and try to follow the acquisition, activation, retention, referral, and revenue steps. Without considering the specific use, case with, the metrics you provide may not be very convincing. You may sound robotic during the interview. So there you have it a simple answer to a real magic interview question from Lincoln. Natural questions are open ended. There are different ways to put them. Let me know how you would answer this question. And what do. You think of the simple answer I provided. Any feedback is welcome. Thank you so much for watching this video.
其次，一个好的度量标准应该易于测量，因此最好提供一个时间阈值。例如，候选人不使用发送的邀请数量，而是使用每个用户每天平均发送的邀请数量。最后，一个好的度量标准还应该为上下文提供信息。如果您只是简单地使用度量框架，例如流行的AARR框架，并尝试遵循获取、激活、保留、推荐和收入步骤。如果不考虑具体的使用情况，您提供的指标可能不是很有说服力。在面试过程中，你可能听起来很机械。这就是来自林肯的一个真正神奇的面试问题的简单答案。自然问题是开放式的。有不同的方法来放置它们。让我知道你会如何回答这个问题。然后做什么。你想想我提供的简单答案。欢迎任何反馈。非常感谢您观看此视频。