Course contentsShow
← Data Science
Lesson 1216 of 2,14531. Data Visualization PrinciplesFree lesson

Choosing the Right Purpose

Determine whether your analysis stage requires exploration or explanation to select appropriate visualizations.

Choosing the Right Purpose

You'll learn how to decide whether a visualization should help you explore data or help others understand your findings.


Every visualization you create should answer one fundamental question: Am I making this for myself or for someone else? This choice determines everything—from how polished it needs to be to which chart type you'll use.

The Two Purposes

When you're knee-deep in analysis, trying to spot patterns, test hunches, or understand anomalies, you're in exploration mode. Here, speed matters more than beauty. Rough scatter plots, quick histograms, and messy correlation matrices are perfect. You're the only audience, and the goal is discovery, not communication.

Once you've found something meaningful—a surprising trend, a key relationship, or an answer to your business question—you switch to explanation mode. Now you're building for stakeholders, colleagues, or report readers. These visualizations must be clear, focused, and tell a specific story. Every element should support your message.

Think Like a Chef

Exploratory visualization is like a chef tasting and adjusting seasonings—messy, iterative, private. Explanatory visualization is the plated dish you serve to diners—refined, intentional, and designed for their experience.

When to Choose Which

  • Early in your workflow (profiling, hypothesis generation, outlier detection): exploration
  • During deep dives (testing relationships, checking distributions): exploration
  • In final deliverables (reports, presentations, dashboards): explanation
  • When answering "what does the data say?": exploration
  • When answering "what should we do?": explanation

Key Takeaway: Match your visualization purpose to your analysis stage—explore messily when discovering, explain clearly when communicating. The same data may need both types at different points in your workflow.