刚刚毫无准备，裸面啦FaceBook data scientist的视频面试， 挂的彻彻底底。
面试官是个很清秀的清华毕业的中国妹子，人特别特别好，说话清晰，逻辑思维明确，人有很好。
我因为上个月接了新的工作，所以这周正在搬家，上周末FB的recruiter突然找我，想着就是看看FB的面试什么样所以就接了面试，中间因为搬家还临时更改了一次时间。
也完全没准备，实在是对不起面试官妹子啦，看到了别怪我啊。
45 min SQL+ABtesting+product
SQL：
table1 ： date, receiver_id, caller_id, duration, country
Q1 for the user who firstly use the new feature on 2018-07-01, how many of them will use it again after X days?
(* no matter they are caller or receiver, they all count as active users)
expected output:
Day UserCount:.
0 99999
1 65789
2 35689
,
99 123
Q2 This new feature is only available to specific users for testing, and for each day the features exposed to the users as in the following table:
users wh have access to this feature might not use it at all.
(* receiver_id and caller_id in the upper table are all user_id)
table 2: date, user_id
question: if a user use this function for at least 3 seconds for one time a day, then we call this users as an active user of this day.
what’s the percentage of active users among all the users who have access to the feature for each day?
然后还问了我：
1 如何选择ABtesting的样本， 考虑哪些因素选择test group 和 control group， 做 random sampling吗
2 如果我们选择要不要推出一个产品， 做ABtesting的话会选择哪些metrics
3 如两组group有显著差异，那么我会考了哪些因素