How to Optimize Marks and Channels for Maximum Impact in Data Visualizations

Data visualization is an art, but implementing it in the most effective way relies on science. Any given visualization is very specific to the particular nuances of that dataset, that company, that question, etc., but there are some guiding principles that can help you to craft the most effective set of visuals that you can.

This will not be an exhaustive guide, literal dissertations have been and will be written on this subject. This is intended to be a starting point to get you thinking about the best ways to present various types of data, and make you think about whether there might be a better way when you are in the process of making a visual.


The geometric shapes that we use in a chart are referred to as marks.

Source: Visualization Analysis & Design: Chapter 5 by Tamara Munzner

The type of mark that we use is dictated by the data that we want to display and the type of chart that we choose.


The properties of marks that we modify in a chart are called channels.

Source: Visualization Analysis & Design: Chapter 5 by Tamara Munzner

Channels are not equally effective in visualizations. The fact is that human visual processing interprets these different channels with different levels of accuracy.

Source: Visualization Analysis & Design: Chapter 5 by Tamara Munzner

As you can see from this chart, we can perceive the length of a mark with a high level of accuracy, but other visual channels are perceived less accurately. An area mark that is twice the the area of another will be perceived as only ~1.6 times as large, while a colored mark with double the color saturation of another will be perceived as being ~3.2 times as saturated.

Guiding Principles

In designing a chart, there are two principles that should always be present in your mind: the Expressiveness Principle and the Effectiveness Principle.

Expressiveness Principle

“The expressiveness principle dictates that the visual encoding should express all of, and only, the information in the dataset attributes. The most fundamental expression of this principle is that ordered data should be shown in a way that our perceptual system intrinsically senses as ordered. Conversely, unordered data should not be shown in a way that perceptually implies an ordering that does not exist.” – Tamara Munzner

Effectiveness Principle

“The effectiveness principle dictates that the importance of the attribute should match the salience of the channel; that is, its noticeability. In other words, the most important attributes should be encoded with the most effective channels in order to be most noticeable, and then decreasingly important attributes can be matched with less effective channels.” – Tamara Munzner

But which channels are most effective for what? Fear not, we have some insight into this as well.

Source: Visualization Analysis & Design: Chapter 5 by Tamara Munzner

Data that are ordered should be presented with magnitude channels, while unordered data should be presented with identity channels. Most of the tools that we use assign these kinds of channels by default and so we don’t even think about this. For example, if you make a Scatter Chart to display many point marks with two ordered dimensions, Tableau (or GDS, or Looker, or virtually any other tool) will display these points along the x and y axes, setting the x and y positions of each point to show the values of each of these dimensions. A Bar Chart, on the other hand, may have one type of ordered data, and one type of categorical data. In this case, x position is used to distinguish between different categorical groups, and then the height of the rectangular area marks uses y position to measure the magnitude of the ordered data.

Extending the example, say you took that scatter plot, but in addition to the two axes of ordered data that you display, you also want to show categorical group membership. The best way to show group membership is with space, and so you could take each of those groups and break them out into separate scatter plots. That would allow the viewer to very easily tell the groups apart from one another and see the scatters for each group. However, by doing this it is less easy to compare the groups to one another, because they are no longer all on the same chart. The next best way to show group membership is with color hue. Note that color hue is inherently unordered (ROYGBV, the order of the rainbow, is learned, it is not inherent to our visual processing, and our brains do not interpret them as ordered without active effort). On one scatter chart, you can color the points for all of the different groups with different color hues. It is a little bit more difficult to tell the groups apart than if you made entirely separate scatter plots, but it is easier to compare them. Neither version is inherently better than the other, it is entirely dependent on what aspect of the data you are trying to emphasize.

An Example

An example scatter plot, using Tableau’s Superstore sample data

Observe this example scatter plot. It uses x and y position for ordered data, as well as x position separation for the first major category. It also uses color hue for the second major category, and then point shape for a third category. Finally, it also uses Area (2D size) of the points to add an extra level of ordered data. Note which information channels are the easiest to see and which require more work to pick out. Next time that you’re making a chart, consider whether you are using the most salient channels to display the most important pieces of data, and whether there is anything you can change to better emphasize your point, or to add some necessary nuance.


Submit a Comment

Your email address will not be published. Required fields are marked *