Facebook 200-101 問題集

質問 1：
An ecommerce brand plans on running a Conversion Lift test on a new media platform. The digital marketing team wants to reach existing customers who have purchased one of its products within the past year. To enable this targeting decision, the digital marketing team asks an analyst to provide a list of email addresses of customers who made a purchase in the past 365 days. The total count of these customers is 10,000,000.
The platform does not have a lift tool in production, so it recommends that the advertiser hold out 10% of the customer list to compare against the other 90%.
Which sampling methodology should the analyst use to most efficiently select the test/control groups?
A. Stratified random sampling
B. Cluster sampling
C. Simple random sampling
D. Multistage sampling
正解：D

質問 2：
A taxi company is working on building an understanding of household customer lifetime value. Some of their customers order via digital platforms, some via phone, and some alternate between the two. They currently calculate lifetime value (LTV) by looking at all hashed order data including email addresses for online customers and all phone numbers for phone orders. The results showed that email customers had a yearly LTV of $100, and the phone customers had a yearly LTV of $80. However, the company is aware that a group of people are introducing some noise into the results by ordering via both phone and online.
What solution should the analyst recommend to enhance the ability to create a functional LTV model?
A. Select the average of the two models
B. Ask online customers to provide their phone number in a post-order survey
C. Gather hashed email and phone number for each order on any channel
D. Remove all customers who use both from the LTV model
正解：D

質問 3：
An ecommerce brand needs to determine if additional placements will drive more results than focusing the budget primarily on Facebook Feed.
A/B test results:
* Campaign 1: Facebook Feed
* Cost per purchase: S5
* Campaign 2: Four placements
* Cost per purchase: S3
* Facebook News Feed cost per purchase: $1.50
* Instagram feed cost per purchase: $2.50
* Messenger cost per purchase: $6
* Instagram Stories cost per purchase: $7
Which action should the brand take?
A. Remove Messenger and Instagram Stories from the test campaign and retest with two new placements: Facebook News Feed and Instagram feed
B. Deliver budget against Facebook News Feed, Instagram feed, Messenger, and Instagram Stories within each ad set
C. Deliver budget to Facebook News Feed within each Ad Set
D. Deliver budget to only Facebook News Feed and Instagram feed within each Ad Set
正解：D

質問 4：
An analyst has calculated an average lift from a random sample of 300 similarly sized campaigns. The analysis showed the following:
* Minimum lift is 98
* Maximum lift is 177
* 95% confidence interval for the average lift per campaign is (133, 145)
* Average p-value for each individual campaign's lift is p = 0.03
What should the analyst conclude about the distribution of lifts?
A. There is 99% certainty that next campaign will have lift covered by the interval of (98, 177)
B. There is 95% certainty that the true average lift of all campaigns is covered by the interval of (133, 145)
C. 95% of the 300 campaigns had observed lift covered by the interval of (133, 145)
D. There is 95% certainty that next campaign will have lift covered by the interval of (133, 145)
正解：C

質問 5：
A quick-serve restaurant brand is launching a Conversion Lift test to understand the effect of media in surrounding neighborhoods. It offers the option of paying in cash, credit card or through its app. One week after the Facebook campaign launches, these are the results:
* The daily budget for each ad set was spent in full
* The majority of people reached live within two miles of any given location
* 90% of transactions were made in cash
* Offline event sets were only uploaded at the end of the campaign resulting in a total of 10,000 conversions
What should the analyst consider when the test is complete to determine if the results are reliable?
A. The match rates across the test group only
B. The percentage of transactions that are not traceable
C. The offline event sets were not uploaded frequently enough
D. Geography level tests should target only users within one mile of a given location
正解：D