Test Questions Flashcards
From third-party study guides and Facebook's MarSci practice exam
See the image below:
A: Marketing Mix Modeling
A
D
C
D
B
C
A & C
See the image below:
Location
An advertiser is running a Conversion Lift test on a new media platform. In discussing how the campaign will be measured, the media platform’s team says that it will compare the conversion rates for the test group exposed to ads vs. a pre-selected control group. The platform does have the ability to run a public service announcement instead, but feels that its methodology is valid.
What concern should the analyst have with the test design?
Randomization is required for valid experimental design
Pre and post measurements will provide a more accurate assessment
There may be contamination between the test and control groups
An A/B test will provide a more accurate assessment
Randomization is required for valid experimental design
An advertiser’s primary product offering is a series of subscription boxes. The advertiser wants to increase user retention. Approximately 25% of customers who buy a three-month subscription do NOT buy another subscription the following year.
A key objective is to reduce the churn rate by 5%. A data scientist develops a model to identify users who have a high probability of churning and to create exclusive offers designed to entice these users to buy another subscription.
Which type of model should the data scientist use?
Logistic regression
Linear regression
Multinomial regression
Support vector regression
Logistic regression
An online shoe brand ran two Conversion Lift studies. Test 1 resulted in the control group showing a 20% conversion rate while the test group showed a 25% conversion rate. The results were significant at 60%. Test 2 resulted in a 20% conversion rate in the control group while the test group showed a 30% conversion rate. These results were 95% significant.
What can the analyst conclude?
Test 2 results are more reliable than Test 1
Test 1 results are more reliable than Test 2
Both tests are equally reliable
A comparison cannot be made between Test 1 and Test 2
Test 2 results are more reliable than Test 1
A brand runs a multi-cell experiment to confirm whether its campaign is generating sales lift. The results were as follows.
- Cell A
- sales lift = 4.0%
- 90% confidence interval = (0.010,0.070)
- Cell B
- sales lift = 4.5%
- 90% confidence interval = (-0.010, 0.084)
- Difference between Cell A and Cell B
- sales lift = -0.5%
- 90% confidence interval = (-0.023, 0.001)
What is the correct interpretation of the results?
Cell A performed positively, Cell B did not perform positively, Cell A performed better than Cell B
Cell A performed positively, Cell B performed positively, Cell B performed better than Cell A
Cell A performed positively, Cell B performed positively, and it is not possible to tell which was better
Cell A performed positively, Cell B did not perform positively, and it is not possible to tell which was better
Cell A performed positively, Cell B did not perform positively, and it is not possible to tell which was better
The correct answer was “Cell A performed positively, Cell B did not perform positively, and it is not possible to tell which was better”.
A confidence interval containing zero means we’re uncertain whether there is an effect. Having zero in the confidence interval implies that the effect could have a positive or negative effect on the outcome of interest. This means we can’t conclude that cell B resulted in a positive outcome, nor can we conclude there was any difference between cells A and B.
Refer to the chart.
An analyst in a technology company aggregates 42 Brand Lift studies that contained favorability questions. The analyst plots these studies against the engagement rate that the ads received. The engagement rate is a summation of likes, comments, and shares over the total reach in each campaign.
Each dot in the chart is a separate campaign.
What is the correct interpretation from this chart regarding the correlation between the engagement rates and favorability lift?
No correlation
Weak positive correlation
Weak negative correlation
Strong positive correlation
No correlation
An automotive manufacturer launches a campaign to sell its newest model. The manufacturer uses a last-click attribution model with a 90-day conversion window to evaluate its media investment.
The attribution results show lower-than-expected online conversions on Facebook. The manufacturer needs to validate these results.
In order to validate the results, what should they be compared to?
The results from last-click attribution model with a conversion window of 30 days
The number of offline conversions
The results from Facebook Data-Driven Attribution
The results from Facebook Brand Lift tests that were run during the campaign
The results from Facebook Data-Driven Attribution