10 Tips For Great Charts Part 1

1) Have a Clear Purpose

At the highest level, there are two types of charts. One type is designed to answer a question and the other is designed to communicate the answer to the question. This is an important distinction, because a chart that is effective for analysis isn't necessarily the most effective chart for communicating the insight we derive from that analysis.

For a controversial example, consider Minard's graph of Napoleon's March. Edward Tufte, the leading expert on information graphics, refers to it as "the best statistical graphic ever drawn." (http://tinyurl.com/6k885) In contrast, Seth Godin, a best-selling author of business books, adamantly disagrees and has said that Tufte was "completely out of his gourd. It is the worst graph ever made." (http://tinyurl.com/2ob4aa)

The difference in whether it is the best or the worst depends on its purpose. As a graph to answer a question, it exemplifies the way in which an effective visualization can facilitate analysis. However, that insight is lost on anyone who doesn't want to spend an hour appreciating its nuances.

If the purpose of your graph is to communicate insight derived from an analysis to decision makers, what is the key message? Generally speaking, it's a one to one relationship. When there are multiple key messages, multiple graphs are often required to communicate them effectively.

2) Consider Your Target Audience(s)

Intuitively we know that people at different levels in the organization require different levels of detail. Thus, a chart for an executive might have much less detail than a line manager, or more likely, we might provide the executive an entirely different chart.

Unfortunately, in the struggle to balance the many best practices for charts, the best practice of adapting your reports (and charts) to your target audience(s) often gets lost. However, it isn't possible to take a strategic approach to reporting without giving it due consideration.

3) Select a Chart Type

When it comes to selecting the right chart, there is also confusion around when to use a line graph vs a column graph and when to use a pie graph vs a bar graph. The answer depends on what you're trying to achieve with the graph, so it is essential to have a clear purpose.

Line graphs are good for showing changes over time, but column graphs are better it it's important to know the exact value of a particular time period. Let's take visits from search engines over the past year as an example. If your key message for the graph was that visits from search engines have increased steadily over the past year, a line graph would be best. If your key message was that visits from search engines peaked in September when the new site was released, a column graph would be better.

Pie graphs are best to show composition, while bar graphs are better if you want to do a comparison. For example, if your key message was that the majority of your visits were referred by Google, a pie graph would work. But, if your key message was that there were slightly more referrals from Yahoo than MSN, a bar graph would be better. Practically speaking, bar graphs are almost always a better choice than pie graphs, but neither can effectively show more than one time period at a time which limits their value.

When selecting a chart, there are two relatively new chart options to consider. Both are designed to show a lot of information in a very small space. The first option is Edward Tufte's sparklines (http://tinyurl.com/zz39j). Sparkline (and sparkbar) graphs are great to add meaning to tabular data which provides the strengths of both tables and graphs. The second option is Stephen Few's bullet graphs (http://tinyurl.com/3cjfwc) which are essentially hybrids of progress bars and gauges. They can show the metric compared to the target and indicate whether the value should be considered good, bad, etc.

4) Add Context

• Trend: How does the metric change over time?
• Rank: What is its relative position compared to other metrics?
• Composition: What percentage is it of the total?
• Correlation: What is the relationship between this metric and another metric?
• Distribution: What is the frequency of the items within a range? For example, how many of your pages are low demand, medium demand, high demand?
• Target: How does this metric deviate from the goal?

The comparison type will influence both the choice of chart as well as elements within the chart, so one could argue that this step occurs both before and after chart selection. In any case, after the chart has been selected, the biggest opportunity to add context is with tabular data.

• Adding sparklines or sparkbars provides a quick view of historical performance for that metric (see example below).
• Adding change indicators, such as up/down arrows, provides a quick view of whether the metric went up or down (see example below).
• Color coding the change indicators provides a quick view of whether the change is good or bad. For example, an increase in visits would result in a green up arrow while an increase in bounce rate would result in a red up arrow (see example below).
• Adding a column that shows the percentage of change from the previous period provides a quick way to show how much the metric changed.

5) Eliminate the Nonessential

• Turning off the gridlines option
• Eliminating the third dimension when it isn't meaningful
• Removing dark background in the plot area
• Deleting gridlines (when appropriate)
• Removing the box around the legend
• Reducing excessive axis values by adjusting the scale (ie if the maximum axis value is 100%, it would be better to have major unit values display every 25% rather than every 10%).

Note from the author: Many people use the words "chart" and "graph" interchangeably. Technically, a graph is a type of chart, but the distinction is meaningless in most situations. For the purposes of this article, we'll make the distinction anyway.

Stay tuned for the next issue of WebSight as our guest writer, Jennifer Veesenmeyer, concludes her final 5 tips for great charts! View 10 Tips for Great Charts Part 2.