Visualize Correlation Between Load Time and Session Duration

What's really being asked

This question assesses your practical data visualization knowledge. The interviewer wants to see if you can select the most effective chart to communicate a relationship between two variables (correlation) and then strategically add a third dimension to explore potential root causes. It's a test of data storytelling and analytical depth, not just memorizing chart types.

The full answer

First, for the initial finding, you should propose a scatter plot. This is the standard and most effective way to visualize the relationship between two continuous metrics. Second, assign the independent variable (app load time) to the X-axis and the dependent variable (session duration) to the Y-axis. Third, to explore the influence of network type, a categorical variable, you should use color. Assign one color to data points from WiFi sessions and another to points from Cellular sessions. This creates two distinct, but overlaid, clouds of points. Finally, explain what you'd look for: do the two colored groups show different trends? For example, is the negative correlation much stronger for Cellular users? This would suggest network type is a significant factor.

The mistakes people make

Suggesting an incorrect chart type like a bar chart or line chart is a major red flag. These are not suitable for showing correlation between two individual, continuous metrics. A line chart incorrectly implies a time-series relationship. Proposing two separate scatter plots, one for WiFi and one for Cellular, is also a weaker answer. While not entirely wrong, it's less effective than overlaying them with color on a single chart, which makes direct comparison of the trends much easier for the viewer. Finally, suggesting a bubble chart where bubble size represents network type is incorrect; bubble size is for a continuous third metric, not a binary categorical one.

What usually comes next

Expect questions like, "What if the third variable was continuous, like 'data transferred during session'?" The correct answer is to then use a bubble chart, where bubble size represents the amount of data transferred. Another follow-up could be, "How would you handle outliers, like a few sessions taking 30+ seconds to load?" A good response is to consider using a log scale on the X-axis to compress the range, making the bulk of the data more visible, but always noting that you must clearly label the axis to avoid misinterpretation.

A concrete example

I would create a scatter plot where the X-axis is "App Load Time (ms)" from 0 to 5000ms and the Y-axis is "Session Duration (seconds)" from 0 to 600s. Each dot represents a single user session. We'd see a downward trend from top-left to bottom-right. Then, I'd color all dots from WiFi sessions blue and all from Cellular sessions orange. If the orange dots are clustered more towards the high-load-time, low-session-duration area, it confirms network type is a key factor influencing the relationship.