Which Chart to Use

Data Types and Scales

• Most 2D charts contain 2 types of data:
Quantitative - the numbers
Categorical - the categories those numbers fall into
• Different types of categories are:
Regions - north, south, east, west or Asia, Americas, Europe etc.
Departments - sales, marketing, operations etc.
Time - days, wekks, months, quarters, years etc.
Bands / Bins - 0-10, 11-20, 21-30 etc.
• Categories fall into one of 3 types:
Nominal - think ‘name’
Ordinal - think ‘order’
Interval or Ratio - think ‘groups’



What's Your Message

• There are 7 types of quantitative messages
Nominal comparison - comparing departments, regions etc.
Time-series - trend over time
Ranking - largest to smallest, smallest to largest
Parts to the whole - actual vs. budget/target, regional share of overall revenues
Deviation - data compared to the norm or expected result expressed as the difference
Frequency distribution - counts of data per interval or range
Correlation - comparison of two paired sets of measures to understand the relationship between the measures
• If you have multiple messages, consider building multiple charts or making the charts dynamic so the user can choose the view


Which Chart to Use

Nominal ComparisonColumn or Bar charts
Time SeriesLine or Bar or Dot or Step chartsLines focus on the overall pattern Bars focus on individual values Dots - best of both worlds Step - Useful for displaying how the levels in your data increase, remain constant or decrease over time
RankingBar or in-cell or Sparklines Charts
Parts to a WholeBar or Column or Bullet charts Stacked Bars (occasionally)Only use Pies if there are <= 3 items / segments
Deviation or VarianceColumn or Bar or Line chartsWhen your target is a range, you can use a line or column chart compared to a range or tolerance
DistributionHistogramBox Plots (box and whisker chart) are another alternative but they're not as easily interpreted to the untrained user
CorrelationScatter chart with a trend line


A Word on Zero Axis

• A column or bar chart must always start at zero


• A common accepted alternatives when starting the axis above zero is to use markers instead of columns