Guide

Choosing the Right Chart: A Decision Guide for Analysts and PMs

The right chart makes a point obvious in seconds; the wrong one hides it or, worse, suggests something untrue. Choosing well is less about design taste and more about one question: what do you want the reader to compare?

Start with the question, not the chart type

Write the message first, in one sentence. “Sales grew every quarter.” “Email brings in more leads than paid search.” “Most orders are under $50.” The message tells you which comparison matters, and the comparison tells you the chart.

The decision guide

You want to show… Use Avoid
Change over time Line chart (bars for a few periods) Pie charts
Comparison between categories Bar chart, sorted by value 3D bars, unsorted bars
Parts of a whole Stacked bar, or a pie with 2–4 slices Pies with many slices
Distribution of one variable Histogram or box plot Bar chart of raw values
Relationship between two numbers Scatter plot Two separate line charts
Many relationships at once Correlation matrix / heatmap A dozen scatter plots
One headline number A large figure with context (“+12% vs last year”) A chart with one bar

Line charts: trends over time

Lines show direction and rate of change better than anything else. Keep time on the horizontal axis, label the last value directly, and limit yourself to a few lines — more than four or five becomes spaghetti. For seasonal businesses, plotting this year against last year on the same axis answers “are we really growing?” instantly.

Bar charts: comparing categories

Bars are the workhorse. Sort them from largest to smallest unless the categories have a natural order (months, age groups). Use horizontal bars when labels are long. Always start the value axis at zero — bar length is what readers compare, and a truncated axis exaggerates small differences. For a quick bar chart in a message or email, the text bar chart generator works anywhere plain text does.

Pie charts: use sparingly

People judge angles poorly, so pies only work when there are two to four slices with clearly different sizes and the point is “share of a whole.” If you need to compare slices precisely, or there are more than a few categories, use a sorted bar chart instead.

Histograms: how values are spread

When the question is “what does a typical order look like?” or “how long do deliveries usually take?”, a histogram shows the shape — clustered, skewed, or with two peaks — that an average hides. Try different bin counts until the shape is clear; the histogram generator makes this quick.

Scatter plots: relationships

To show whether two measures move together — ad spend and sales, price and units sold — plot one on each axis. Add a trend line if helpful, and report the correlation coefficient. Remember that correlation is not causation; our guide to correlation vs regression explains the difference. For many variables at once, a correlation matrix is a good first look.

Five rules that improve any chart

  1. Title it with the takeaway (“Email drives 40% of leads”), not just the topic (“Leads by channel”).
  2. Remove clutter — heavy gridlines, borders, 3D effects and unnecessary legends.
  3. Label directly on lines and bars where possible, instead of making readers match colors to a legend.
  4. Use color with purpose — grey for context, one accent color for the point you are making.
  5. Show the source and time period so the chart can be trusted and repeated.

Charts that mislead (even unintentionally)

Key takeaways