The ultimate guide to product analytics interviews. Strategies developed by FAANG engineers to craft structured, impactful answers and increase your chances of landing a data science job.

Product analytics interviews evaluate your skills and creativity as an aspiring data analyst. Interview questions evaluate if you can use your technical knowledge to guide the further development of a product. Learning how to answer these questions is crucial to ace most job interviews for data scientists and analysts.

In this course, you’ll first gain an overview of what product analytics interviews look like and what kinds of questions usually come up. Next, you’ll learn to use a framework for answering product analytics questions, showcasing your analytical skills. Finally, you’ll learn how to apply said framework to different kinds of product analytics interview questions by looking at several examples.

With this course, you’ll increase your chances of landing a job in data science by learning how to effectively answer product analytics interview questions. You’ll be able to provide well-structured and well-communicated answers, and show you can creatively tackle any business problem you may face.

Ace the Product Analytics Job Interview

# Problem

[Let's say that Facebook would like to change the user interface of the composer feature (the posting box) to be more like Instagram. Instead of a box, Facebook would add a "+" button at the bottom of the page. How would you test if this is a good idea?](https://www.interviewquery.com/questions/changing-composer)

Try answering this on your own before continuing. Use the hints below if you are stuck.



We can begin thinking about this question in terms of the metrics that Facebook wants to improve. Which metrics relate to posting, and which specific posting metrics would correspond to Instagram’s style rather than Facebook’s? It helps to bring it back to the goal of each platform. 

What is Facebook’s mission, and what is Instagram’s?


Consider the trade-offs between Instagram posts and Facebook posts. Which app utilizes which type of post more often? 

Hypothesize the difference if Facebook were to shift to an Instagram model and propose data to validate it.

Think about the drop-off rates for both Instagram and Facebook on the post composer tool. What steps in the funnel would users start falling off for both apps? Let’s say we propose to measure the trade-offs between the existing UI of the Facebook composer versus the Instagram UI. 

While the current composer feature on Facebook may make it easier to share status updates and geo-location or sell items, the Instagram composer may make the user more inclined to share photo posts. Now, how can we decide if switching to more photo inclined posts is a good idea?


# Solution

Let's make some initial assumptions. We can guess that we want to try a new user interface to improve certain key metrics that Instagram does better than Facebook in. Noticeably, given that Instagram is a photo-sharing app, we can assume that Facebook wants to improve:

* Posts per active user
* Photo posts per active user

Additionally, we have to measure the trade-offs between the existing UI of the Facebook composer versus the Instagram UI. While the current composer feature on Facebook may make it easier to share status updates and geo-location or sell items, the Instagram composer may make the user more inclined to share photo posts.

Therefore, given this hypothesis, one way to initially understand if this test is a good idea is to measure the effects of an increase in the proportion of photo posts to non-photo posts on Facebook and how that affects general engagement metrics.

For example, if we compare the population of users that have a percentage of photo posts from 10% of the total versus 20% of the total posts, does this increase our active user percentage at all? Would it increase monthly retention rates?

Another thing we have to be aware of is the drop-off rate for the Facebook composer versus the Instagram composer. The drop-off rate would directly affect the general amount of posts that each user makes. We can look at the drop-off rate between the two composers by different segments as well, such as geographic location, device type, and demographic markets.

If we want to run an AB test to test the differences instead of just analyzing our existing segments, we would have to evaluate these same metrics but make sure not to compare by specific segments unless they are a large sample size of the population. Doing it by market/segment may leave it so that we get a Simpson’s paradox scenario, where for most markets, we get a certain result, but in aggregate, the result is different.

In running the AB test in addition, it’s important to add in the specific rigidity that the test must be run. For example, sample size and distribution are important to make sure that we have a sufficiently large sample size in both control and test to get a statistically significant result. We should also randomly assign people to either test/control as well as reach significance and single variable change on the composer element.



 Try to solve this question to test your understanding of the material.


Sample Question: Changing Composer

Try to solve this question to test your understanding of the material.

Introduction

Investigating Metrics

Fractional Metrics

Measuring Success

Feature Change

Metric Trade-Offs

Growth

Conclusion

Sample Question: Changing Composer

Problem

Solution