Uber数据科学面试真题

职位分类
全部
数据科学
数据分析
计算机科学
人工智能
产品经理
BQ
面试题
全部(67)
A/B testing(0)
Statistics(2)
Product Case(55)
Coding(8)
Modeling(1)
高频题(1)
全部(67)
A/B testing(0)
Statistics(2)
Product Case(55)
Coding(8)
Modeling(1)
高频题(1)
1.如何让你定价
2.A/B test
3.Hypothesis testing
4.Bigram
5.Square root计算器
6.Order zeros to the front
7.Merge two sorted lists
8.A/B testing
9.设计推荐司机去哪的algorithm
10.Predict the cost of a claim
11.机场algorithm
12.两个coupon A/B test
13.How to price incentives
14.ETA algo
15.How to optimize dispatch?
16.Prediction
17.Overcharged rides
18.Ride feature during off hours
19.How to improve the home page of Ride App
20.Arrangement of category tab
21.Pros and cons of arrangement of category tab
22.Ads display
23.Measure movement of browse tab
24.Impact of changing ads word
25.Measure giving incentive to existing riders
26.Design push notification for driver app
27.Investigate reasons of decreasing adoption in some cities
28.Analyze bad experience of sending request to pick up
29.Restaurant ranking
30.Pros of adding carousel
31.Measure the ETA change
32.Measure the new feature to donate restaurant
33.Investigate the reasons why ETA doubled last month
34.Measure the supply and demand of Uber eats, and the reasons why number of drivers decreased
35.Measure the success of instagram
36.Investigate reasons of increasing delivery time
37.Evaluate Uber rides marketplace health
38.Investigate undersupply
39.Measuring features of adding more time to match drivers for far distance
40.Measure the success of Uber eats search bar
41.Quantify the impact of project
42.Which store has higher delivery fees
43.Network effect of two-sided market
44.Diagnose steps of metrics fluctuation
45.Tradeoff evaluation
46.Factors of location recommendations
47.Pros and cons of Uber Pool
48.Square root function
49.Hypothesis testing
50.Measure pre-order on Uber eats
51.Improve the performance of pick up and drop off experience
52.Investigate the sudden decrease of total trips
53.Measure category on uber eats
54.Possible problems of driver picking up
55.Improve pick up experiences
56.Investigate the increase of average delivery time
57.Measure the new feature of adding green vehicle
58.monthly subscription pricing
59.Measure adding category of new business
60.Transforming date type from int to float
61.Measure success of batch delivery
62.Measure success of batch delivery
63.What to do before launch uber eats in a new city
64.How to get more driver to deliver on off peak hours
65.Random number generator
66.Build a ML model to predict how long a pick-up order will take
67.Log and p-value
1. 如何让你定价
黑车外卖这个软件有两个收费项目。一个是运费,可能$1.99, 或者$2.99,还有一个服务费,比如是你总单的20%。 这样子的收费模式叫什么?为什么要用这样的收费模式,好坏是什么?为什么不增加运费从而不收‍‌‍‍‍‍‌‍‍‍‍‌‌‍‍‌‍‍‍服务费,或者增加服务费而不收运费?如果让你定价,你怎么定? 
2. A/B test
背景大概是driver取餐可能会提前到餐厅或晚到餐厅,有什么影响,怎么验证这个影响,如何optimize dispatch使得客户体验最好,

如何设计metric,如何A/B test
3. Hypothesis testing
先生成一些随机数,然后再去做hypothesis testing.
4. Bigram
Given a string find all the bigram (two adjacent words) and their counts
5. Square root计算器
写一个square root计算器
6. Order zeros to the front
Python (order zeros to the front)
7. Merge two sorted lists
merge two sorted lists
8. A/B testing
Case study是一道关于A/B testing的题 ,包括collect什么data,sample size多大,需要collect data多久,假设检验H null和alternative是什么,test statistic长什么样,什么是p value. 应该都是A/B testing里的基础问题.
9. 设计推荐司机去哪的algorithm
Uber要launch一个可以推荐司机去哪的algorithm,然后关于这个algorithm如何设计,以及如何validate之类.
10. Predict the cost of a claim
1) Given the description of claims and their corresponding costs [0, xxx), how to predict the cost of a claim (assuming data feature has been finished)

a) How to deal with the skewed numerical response

b) How to deal with 0 counts assuming a majority of observations will have 0 costs

c) In terms of modeling, could use a mixture model
11. 机场algorithm
黑车在机场的algo和平时的algo会不一样吗?为什么不一样?
12. 两个coupon A/B test
从两个coupon出发,问了A/B test的design,问了把A/B test放到一个regression里的分析,问了t stat, F stat, R square,

Type I / II error,sample size,有的是类似给定义,有的是结合case分析. 
13. How to price incentives
How to price incentives (A/B testing, optimization, machine learning model都有问到)
14. ETA algo
How to evaluate new ETA algo
15. How to optimize dispatch?
Driver取餐可能会提前到餐厅或晚到餐厅,有什么影响,怎么验证这个影响,如何optimize dispatch使得客户体验最好,

如何设计metric,如何A/B test
16. Prediction
如何predict 送餐ETA
17. Overcharged rides
 | Case:Uber health(1) What' s the difference between normal uber and potential problem
 (2) How to track those overcharged rides
 (3) How to improve that
18. Ride feature during off hours
 | Schedule a ride feature during the off hour
 (1) Metrics for three side and guardrail metrics?
 (2) How to design the experiment
 (3) How variance impact on sample size and how to control that
 (4) What' s the p value?
 (5) How to decide MDE
 (6) if cancellation is high, how will you recommend?
 (7)lf one city did not pass stats significance, What to do?
 (8) How to apply pvalue for ABN test?
19. How to improve the home page of Ride App
 | How to improve the home page of Ride App
20. Arrangement of category tab
 | How to decide the arrangement of category tab, Where to, destination shortcuts?
21. Pros and cons of arrangement of category tab
 | Should we customize the arrangement of the Home tab for users? Pros and cons?
22. Ads display
车顶放了滚动广告装置,问如何measure impact,选哪些metric,如何设计实验
23. Measure movement of browse tab
 | Uber Eats把browse tab移动到home page的一个search bar的pros/cons,怎么measure,用什么metrics,怎么算sample size。如果看到click through rate增加但是revenue/order下降你有什么hypothesis,会怎么调查。
24. Impact of changing ads word
 | for surge area, we currently show drivers "you can earn X times for your next trip". Product team wants to change the message to ""you can earn +X$ for your next trip". Why do we want to do this? What's the goal metrics and how to measure the impact?
25. Measure giving incentive to existing riders
 | we want to give existing riders incentive for their future trip. What's the pros and cons? What's the goal metrics and how to measure the impact?
26. Design push notification for driver app
 | driver app push notification test 怎么设计
27. Investigate reasons of decreasing adoption in some cities
 | Uber One (uber 的membership program)突然发现从四个月开始,有些city的adoption下降,问为什么
28. Analyze bad experience of sending request to pick up
 | 分析从用户send request到司机到达pick up location流程中的bad experience
29. Restaurant ranking
 | 可以用哪些信息来rank餐厅
30. Pros of adding carousel
 | 现在是scroll down看餐厅,如果中间加一个carousel样式的滑动,会有什么pros
31. Measure the ETA change
 | Uber eat range ETA changed from 10 minutes to a wider range.分析ETA范围变大的Pros and cons,用什么metrics测,以及做什么实验来检验?
32. Measure the new feature to donate restaurant
 | Launch a new feature that can donate restaurants tips to survive through covid.主要是说这个feature的pros and cons, metrics,实验
33. Investigate the reasons why ETA doubled last month
 | ETA double compared to last month,why?
34. Measure the supply and demand of Uber eats, and the reasons why number of drivers decreased
 | How to measure the supply and demand on Uber Eat? Why did the number of drivers suddenly decrease on Uber eat?
35. Measure the success of instagram
 | How to measure the success of Instagram? If engagement decrease, what should do(long term and short term methods)
36. Investigate reasons of increasing delivery time
 | Average delivery time increase,是什么原因。然后一步一步deep dive到root cause,是undersupply。是什么原因造成了undersupply.设计new feature去solve 这个问题。最后考了A/BTesting的问题(具体是dynamic pricing这个feature),如何设计sample size, running time,还有treatment和 control group的ratio distribution.
37. Evaluate Uber rides marketplace health
 | How to evaluate Uber rides marketplace health?
38. Investigate undersupply
 | Target on Supply-Demand. What's Undersupply? Why will happen? What kind of metrics to evaluate? Some features to Optimize?
39. Measuring features of adding more time to match drivers for far distance
 | Tagret on一个Lyft已经有的feature: wait for more time to match available drivers from relatively far distance,去定义evaluation metrics,解释network effects,如何分析实验结果(it' s not statistical significant, but we find the positive result on our key metrics, how to explain to our non-technical stakeholders?)
40. Measure the success of Uber eats search bar
 | Ubereats search bar这个feature,有什么metrics来衡量success,如果把search bar转移到recommendation page,这样的pros 和cons
41. Quantify the impact of project
 | Project的impact如何量化,这个impact为什么benefit from your analysis,如何确定是因为相关性还是因果性。
42. Which store has higher delivery fees
 | 麦当劳和寿司店,哪家店才用higher delivery fees,如果实际用的delivery fees和assumption截然相反如何解释
43. Network effect of two-sided market
 | 详细解释two-sided market的network effects,是underestimate还是overestimate,替代方案是什么,Switchback test的原理,back and forth 时间的如何set up,30min是一个合理的时间吗
44. Diagnose steps of metrics fluctuation
 | metrics fluctuation 的diagnose steps
45. Tradeoff evaluation
 | 一个公司想从一个地方运货到另一个地方,有两个channel去place order,第一个是phone,email,account manager,第二个是web channel,现在data显示第一个channe的order cancellation rate是20%,第二个channel的order cancellation rate是10%,分析原因。如果web channel的下单率更高,但cancellation rate也更高,如何evaluate这个trade off
46. Factors of location recommendations
 | Driver team launch一个新的feature:在Uber司机的app上面推荐locations,推荐的地点需要考虑什么factors
47. Pros and cons of Uber Pool
 | Uber Pool这个feature想要re-launch,需要考虑的Pros and Cons
48. Square root function
 | 定义开方函数
49. Hypothesis testing
 | 给一个场景从头开始设计hypothesis testing,问了p-Value怎么解释给nontechnical people,t-test和z-test,type error,sample size等各种细节
50. Measure pre-order on Uber eats
 | Ubereats用户目前不能在restaurant close的时候下单,如果现在增加一个feature让用户可以在restaurant在App上还是close状态时提前order,这样做的prosand cons,怎么测试?
51. Improve the performance of pick up and drop off experience
Drive的pick up and drop off experience。只考虑driver接单到最终pick up rider的过程,需要track哪些主要的metrics,总体怎么提升这个过程的performance?(过程中每一步可能出现什么问题,每个问题用什么数据identify,有什么解决方法,怎么测试)
52. Investigate the sudden decrease of total trips
 | total trips类然下降,怎么diagnose?
53. Measure category on uber eats
 | 测试ubereats主页上category部分的设计(比如之前是展示的不同的cuisine type现在改成了展示不同的category:food,grocery,alcoho等等;还有改变size),怎么test。
54. Possible problems of driver picking up
driver在pick up rider的时候可能会出现什么问题(比如driver已经到达地点但是联系不上rider,互相找不到彼此之类),你有什么解决方法?
55. Improve pick up experiences
 | pick up experience中有什么可以提升的,主要关注driver已经到达地点之后,可能会出现什么问题,你有什么solution,你可以看什么metircs来看performance,怎么测试你的idea?好像过程中还有问到有了idea,
56. Investigate the increase of average delivery time
 | 围绕average delivery time上升,问怎么分析
57. Measure the new feature of adding green vehicle
 | 给打车的时候多一个选项,green vehicle(hybrid,ev),围绕这个问一些metrics和分析
58. monthly subscription pricing
 | uber eats的monthly subscription的定价问题
59. Measure adding category of new business
 | uber eats想promote一些新的business打算把他们主页上加一些category(pharmacy,alcholol),问应该想哪些metrics,怎么计划experiment
60. Transforming date type from int to float
 | 如何将data type 从int 变成float 例如1/3(int/int)会成为0
61. Measure success of batch delivery
 | Uber eat 想推出batch delivery,即一个骑手可以同时接同个餐馆多个单,有什么pros and cons, success metrics, 使用什么实验来验证效果?
62. Measure success of batch delivery
 | Uber eat 想推出batch delivery,即一个骑手可以同时接同个餐馆多个单,有什么pros and cons, success metrics, 使用什么实验来验证效果?
63. What to do before launch uber eats in a new city
 | Uber eats want to launch in a new city. what should you consider before the launch? What' s your rollout plan for the first 3 months? How do you evaluate if the launch is successful?
64. How to get more driver to deliver on off peak hours
 | How to get more driver to deliver on off peak hours
65. Random number generator
写一个random number generator,要按照一定的weights,比如weights是[1,2,3,4.5],需要先normalize一下。
66. Build a ML model to predict how long a pick-up order will take
 | build一个ML model,来预测一个pick-up order多久能够完成
67. Log and p-value
(1)怎么算 log;(2)怎么算 p-value, 假如比较的是连续变量