Case study: How would you investigate a Drop in User Engagement?

Data Science Interview QuestionsCategory: Data ScienceCase study: How would you investigate a Drop in User Engagement?
MockInterview Staff asked 7 years ago

Before starting, be sure to read the overview to learn a bit about Yammer as a company.
Yammer’s Analysts are responsible for triaging product and business problems as they come up. In many cases, these problems surface through key metric dashboards that execs and managers check daily.
The problem
You show up to work Tuesday morning, September 2, 2014. The head of the Product team walks over to your desk and asks you what you think about the latest activity on the user engagement dashboards. You fire them up, and something immediately jumps out:

The above chart shows the number of engaged users each week. Yammer defines engagement as having made some type of server call by interacting with the product (shown in the data as events of type “engagement”). Any point in this chart can be interpreted as “the number of users who logged at least one engagement event during the week starting on that date.”
You are responsible for determining what caused the dip at the end of the chart shown above and, if appropriate, recommending solutions for the problem.
Getting oriented
Before you even touch the data, come up with a list of possible causes for the dip in retention shown in the chart above. Make a list and determine the order in which you will check them. Make sure to note how you will test each hypothesis. Think carefully about the criteria you use to order them and write down the criteria as well.
Also, make sure you understand what the above chart shows and does not show.
If you want to check your list of possible causes against ours, read the first part of the answer key.
Digging in
Once you have an ordered list of possible problems, it’s time to investigate.
For this problem, you will need to use four tables. The tables names and column definitions are listed below—click a table name to view information about that table. Note: this data is fake and was generated for the purpose of this case study. It is similar in structure to Yammer’s actual data, but for privacy and security reasons it is not real.

  • Table 1: Users
  • Table 2: Events
  • Table 3: Email Events
  • Table 4: Rollup Periods

Making a recommendation
Start to work your way through your list of hypotheses in order to determine the source of the drop in engagement. As you explore, make sure to save your work. It may be helpful to start with the code that produces the above query, which you can find by clicking the link in the footer of the chart and navigating to the “query” tab.
Answer the following questions:

  • Do the answers to any of your original hypotheses lead you to further questions?
  • If so, what are they and how will you test them?
  • If they are questions that you can’t answer using data alone, how would you go about answering them (hypothetically, assuming you actually worked at this company)?
  • What seems like the most likely cause of the engagement dip?
  • What, if anything, should the company do in response?

Source: Investigating a Drop in User Engagement

2 Answers
Jayavarshini Ilarajan answered 5 years ago

The basic criteria that can be evaluated during the analysis of the drop in the user engagement can be determining

  1. The total time a user spends using the product or a service, i.e., how long the user session lasts. To see if it has been reduced in the past time period. Various analysis of the session length can give an idea if the usage is declining.
  2. Satisfaction rate: By getting a feedback from the user, the product analysts can determine the net promoter score of the application. The users often will want to share their feedback.
  3. Concentrating on the active users than all the users. Some users may just have given a shot in using the product. By measuring the usage of the users who have been consistently using the product, gives us a clear picture.

How to improve the declining usage?
Create new features and experiment with different values of the same variable to see if it is a better value. Have a control and a variant group to monitor the efficiency. Ensure these tests can practically run and if we can derive meaningful results.
Interpret p value for Analyzing the hypotheses. Reject the null hypothesis for lesser p value (May be if p<0.05) or else consider.    

Your Answer

14 + 18 =