Three variables is where I got tripped up.
Start by clarifying the goal of the visualization—whether it's for exploratory analysis, monitoring, or a specific business decision—and confirm the data available. Then propose a multi-faceted visualization that handles three variables, such as a scatterplot matrix or a 3D scatterplot with color encoding, while discussing trade-offs. Finally, tie the visualization back to actionable insights for the rideshare business.
Pro tip: Mention that you would first check for confounding factors like distance, time of day, or surge pricing, because raw correlations can be misleading. Also, suggest using interactive dashboards (e.g., Amazon QuickSight) to allow stakeholders to explore the data themselves.
Ask whether the visualization is for data scientists, executives, or product managers, and what decision it should inform. This determines the level of detail and interactivity.
Confirm that ride duration, fare, and ratings are available at the individual ride level, and consider additional variables like distance, time, and location that might affect relationships.
For three continuous variables, propose a scatterplot matrix or a 3D scatterplot with color/size encoding. Discuss alternatives like heatmaps for binned data or parallel coordinates.
Explain how to compute and display correlation coefficients (e.g., Pearson) and caution that correlation does not imply causation. Suggest controlling for confounders.
Describe how the visualization could reveal patterns, such as longer rides having higher fares but lower ratings, and how that could inform pricing or driver training.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.