求一起总结case problem题目既已经毕业了的渣渣DataScience找工学...

6月份刚从统计master毕业的非科班出身渣渣，最近两个月连个面试都没有，小公司直接投了没消息，大公司连内推也不敢。但是再不努力就不行了，赶紧来打卡督促自己学习，每天都至少学习面试的一部分内容。
我的一个很悲惨的经历先分享给大家(两次被同一家公司放鸽子，记录一下我的凄惨经历
http://www.1point3acres.com/bbs/ ... &fromuid=148112
(出处: 一亩三分地论坛))
，不知道我还能不能找到这么好的职位了。但是志存高远，我相信我可以找得到，不单是sql处理数据的工作，一定要用R/python，做modelling，还有机会接触一些大数据工具，比如spark，tensorflow等等。

再就是跟随小k大神的脚步，虽然自己底子烂，学校烂，但是能学到一点是一点。上次听完讲座，我马上去买了cracking the PM interview 这本书，直接从case problem看起。
今天先来总结一题， what is your favorite product?
1. What problems does the product solve for the user?
2. How does the product accomplish these goals? What makes it "neat"? What makes users fall in love with the product?
3. How does it compare to the alternatives?
4. How would you improve it?

我之前面试时，就被面试官问过what is your favorite online shopping website? What do you like it most? How would you improve it/what is its shortage?
整理答案如下，希望大家一定不吝赐教，一起修改答案，做到最好

Q: What is your favorite online shopping website? Why? Howwould you improve it?

A: Nordstrom.com.

1. Nordstrom.com is an online shopping website for the chaindepartment store. It provides online catalog and ordering features to make upthe in-store purchase for customers.

2 &3. I love shopping on Nordstrom.com, especially itsmobile app. I usually browse all kinds of products from different departmentsand also get inspired to buy products either online or get to the store to purchase.The features I like the most are: 1. Free shipping and returns anytime for anyamount of purchase. 2. In-app notification of order status. 3. Pick-up atnearest stores. These three features are outstanding compared to otherdepartment store online shopping platforms.

4. There is one feature that I think can be improved toincrease the revenue. In the main page, there is a part, ‘get inspired @Nordstrom on Instagram’. It shows posts from customers and I feel like if thereare direct links to products shown in those posts, it would increase customer’sdesire to purchase.

Q: How to improve?

A: Now we need to launch this new feature, tag or direct linkto the particular product. This new feature is designed to attract customerswho have habits to check this part on app.

The ultimate goal of this new feature is to finally increasethe revenue, that is to say, the amount of sales increase. Therefore, weconduct an A/B test to investigate the hypothetical increase in the sales ofthe particular products in post.

Now we design the experiment.

1.    Primary metric: the average spent per userwithin one week/month(keep posts same in this certain time period)

a)      May also look at: conversion rates, clicks/addto cart, clicks/purchase, etc.

b)     This metric is a mean, therefore it followsnormal distribution and we know the standard error once we have data.

2.    Choose target population: app users/previouscustomers who took a look at/had habit checking ‘get inspired @ Nordstrom onInstagram’ before launching new feature.

3.    Randomized sampling with a sample size:

a)      Use power analysis to set an appropriate samplesize N:

                     i.           Need: confidence level(95%), smallestpower(80%), effect size[(mu1-mu2)/sigma](difference between hypotheticalincreased revenue and existing revenue that we wanted to test out)

b)     Don’t forget duration vs. exposure tradeoff.

4.    Specify control group and experiment group:

a)      Control group: at least N number of customerswho took a look at ‘get inspired @ Nordstrom on Instagram’ before and now stillhave this feature with no links

b)     at least N number of customers who took a lookat ‘get inspired @ Nordstrom on Instagram’ before and now have this featurewith direct links to particular products

Once we run the experiment for a while, we can analyze theresult:

1.    We can conduct a two-sample t-test comparingcontrolled group and experiment group

a)      Null hypothesis: the difference in average spentper user between experiment group and controlled group is 0.

b)     Set type 1 error: 5%

c)      Calculate t-score = (xbar1-xbar0)/Sp*sqrt(1/n1 +2/n2), where Sp is the pooled sample variance, xbar1 and xbar0 are the averagespent per user for experiment group and controlled group, respectively.

d)     Decision criteria:

                     i.           Calculate 95% CI, if includes 0 then not statsignificant, if no 0 then stat significant

                   ii.           Get p-value, if p-value > 0.05 then not statsignificant, vice versa

                  iii.           Compare t-score with critical value according to5% significance level. If the absolute value of t-score larger than criticalvalue, then stat significant.

Once we have the result, we should make recommendation.

If significant, we may launch the new feature, but we still need toconsider:

1.    is the impact large enough to cover the cost?

2.    Are there seasonality effects?

If not significant, probably notlaunch, but need to check more cross effect, e.g. confounding variable effect.

又是全新美好的一天！昨晚上又写了一道题，但是感觉自己写的很不好，没啥内容，希望路过的筒子不只是留下脚步，还能看过我的总结后一起讨论一下！

There’s a 5% drop in the number of visits to our website homepage today. What happened?
What? How? Why?
1.        Clarify: What is the drop look like? Is it really a drop?
a)        Might because of seasonality, holidays etc. We need to analyze historical data to decide. Year-on-year analysis, etc..
b)        What is the overall variance of it?
c)        Do our daily active users (DAUs) drop as well? (since DAUs is more like a KPI than # of visits)
2.        Why?
a)        Global/industrial recession?
b)        Competitors?
c)        Marketing events/promotions?
d)        Segmentation:
i.        By sources: organic search, direct, paid, social, referral
ii.        By countries/region/location etc.

今天的任务，把已经做过的一个project再好好做做，复习一下R的一些包和function
发一个data.table的cheat sheet，今天好好复习一下
https://www.google.com/url?sa=t& ... v6z6e7qToOpS9WdXOSQ

唉呀妈呀，注意力太涣散了，现在又开始在DataCamp上面学data table的课了，这些新的code 哦，真的学一遍忘一遍，因为都是为了准备面试，而不是真的有工作要一直用来做project， 头疼，除了一遍一遍的练习没有办法

楼主加油呀，肯定前面有更好的offer等着你!
我也是非stat出身的废柴研究生，马上就研二开始准备找工作了，很慌啊，我这一年只用SAS和SQL，因为我们专业是那种很传统的统计
因为想跟男朋友去加州不想异地，所以也开始学起R和Python来了，虽然我知道大概率是拿不到加州的offer的
但还是要有希望！！！！

答案的要点是逻辑清晰，结构化，不能想起一点是一点。我有个面试题，但是一点结构化的思路都没有，就只能brainstorming的想起什么说什么，面试官直接说，就给我最重要的一点，不要穷举

赞楼主，我也是刚毕业在找ba方面的工作，每天都好绝望，连学习都没有动力了，目前也在补business方面的知识，向楼主学习。

csphoto 发表于 2017-8-22 08:05
赞楼主，我也是刚毕业在找ba方面的工作，每天都好绝望，连学习都没有动力了，目前也在补business方面的知识 ...

一起加油，我来发帖子也是因为我已经毕业了周围没有人一起学习了，觉得自己支撑不下去，想来和大家一起寻找那个学习的氛围

大家走过路过给我留个评论吧，心里实在太难过了，今天白天发了好几封邮件一个人都没有理我的，强忍眼泪不哭出来

赞楼主！new grad同在找da的工作，深知其中的不易。楼主一定要加油！

XiaonanD 发表于 2017-8-22 11:08
赞楼主！new grad同在找da的工作，深知其中的不易。楼主一定要加油！

谢谢支持！一起加油！

同是数据科学。。。真的是负面情绪太多了。。。

楼主加油！一起总结