Analytics DS面试通稿

在2020的末尾结束了第二次跳槽面试。跟上一次大动干戈的转行跳槽比较起来，这次感觉像是只是多开了几十个视频会议。琢磨起来，这些都得益于转行入科技公司的经历（和COVID???）。在这里想写下一点小结，希望可以给想要进入科技公司做analytics（a.k.a. 不做production ML model的data scientist) 或者初入职场的数据同行们一些帮助。有讲的不妥的地方欢迎讨论或指正。

本文不讨论：选择ML还是Analytics，面试题目，ML知识的准备
本文适用于：求职中的学生，初入职场的数据同行们
提示1：为了写作方便，本文会有大篇幅破碎的英文，有语法用法的错误请见谅。
提示2：因为格式问题，另提供了

Data Scientist in Analytics - What are the most demanded skills?
这里推荐一篇关于data scientist career progression的。这里我根据自己的理解做了一些改动：

Level I: execution

• can accomplish what has been assigned to you;
• capable of finding resources to solve the challenges in the projects
• [good to have] deliverables are extendable/reusable
  

Level II: independence

..

• define your project scope;
• start to influence project priorities/product roadmaps
• [good to have] lead interns/onboard new hires
  

Level III (senior): influencer

• recognized as a domain expert
• influence product vision/roadmap/priorities
• own work streams and lead junior scientists
• make decisions together with cross functional leads
  

不难理解，面试想要测试的也就是这些方面。当然HR通常会取面试表现和工作经验的最小值来定level, TAT。把面试类型分类一下：

1. Coding interview (SQL/Python/R) → Execution (basic querying skills)
        Most coding interviews are just simple data wrangling questions. Some companies will test Leetcode/ML questions. Make sure you know what is going to be tested before getting into the interviews.

2. Probability/Stats → Execution (basic experimentation skills)

3. Business Case → Execution & Independence (refer to PM interview skills)

• Business trend analysis: how to interpret a metric moves up/down a lot suddenly
  

                   - Consider internal (ETL, product change, seasonality) vs. external factors (competitor, events)

• Open end product questions
  您好！
  本帖隐藏的内容需要积分高于 188 才可浏览
  您当前积分为 0。
  使用VIP即刻解锁阅读权限或查看其他获取积分的方式

  游客，您好！
  本帖隐藏的内容需要积分高于 188 才可浏览
  您当前积分为 0。
  VIP即刻解锁阅读权限 或 查看其他获取积分的方式
  Unlock interview details and practice with AI
  Curated Interview Questions from Top Companies
  
  p that is excluded from all the new features is the control group when you want to analyze the aggregated results for the experiment series.
• Always document your experiment design, even if it is a simple set up. The key sections include:
  

• Hypothesis
• Metrics：
  

            - Primary Metrics: the one to verify your hypothesis
            - Secondary Metrics: metrics that will be directly impacted in the treatment, such as conversion funnel metrics, marketing engagement, etc.
            - Countermetrics: any trade-offs in this experiment? Usually about cannibalization

• Decision Criteria ：
  

            - Set a target to mark the success of the experiment
            - A decision tree and a guardrail are recommended if there are trade-offs or investment costs: situations that we will stop the experiment right away; situations that we will not launch but continue to iterate

• Keep a list of experiment impacts you analyzed and better to normalize the annual impact to all audiences. It will come back to you once or twice a year when all the cross-functional partners are trying to figure out the impacts they can write in their perf reviews and the standardized experiment impact tracker is also your impact :)
  

其实还有很多感想，但是篇幅已经很长，就姑且写到这里。希望对大家有用，谢谢！

请问“关于Data Challenge的旧文”是哪篇呢？可以把link发一下么

sedm4 发表于 2020-12-21 03:29
请问“关于Data Challenge的旧文”是哪篇呢？可以把link发一下么

看这里

楼主，链接好像不能下载，请给下载权限，谢谢

感谢楼主分享！楼主能稍微再展开讲讲做oppo sizing的方法吗？还有oppo sizing的metric基本都是monetary的metric吗(比如revenue, profit)

很有用的信息！

楼主怎么如此聪明美丽又有智慧

mark yixia

感谢分享~~