Picking the right color palette for your visualization

At Bitmetric we continuously hone our skills, and we like to help you do the same. That’s why we regularly post a new Qlik certification practice question to our LinkedIn company page. Some time ago we asked the following Qlik Data Architect certification practice question about picking the right color palette for a data visualization.

Which option do you think is best?

The correct answer is D: diverging colors

To learn why, let’s look a the different types of color palettes we have available for our visualizations:

• Single color

• Qualitative / Categorical colors

• Sequential colors

• Diverging colors

ℹ This article is a primer on selecting the right color palette for your visualization. There’s a lot more to learn beyond the basics and we’ll be returning to this topic in future blog posts. Sign up for our newsletter if you want to be notified of new posts.

Let’s take a closer look at each of the palettes!

Single color

Very often, the simplest solution is the best. If you only want to visualize a single data series then all you need is a single color. Consider for example the chart below from Google Trends, which shows the search interest over time for the term “AI”.

This chart very clearly shows how the interest in AI exploded after ChatGPT was released in November 2022. There is absolutely no need to add any further color to make it easier to understand.

Of course, using only a single color, even if it’s your favorite one, can become boring after a while. Fortunately, there are more than enough reasons for multi-color palettes. Let’s look at those next.

Qualitative or categorical colors

A qualitative, or categorical, color palette consists of different colors, or “hues”. A categorical color scheme should be used for charts in which the dimensions/categories have no inherent order. An example of this type of palette can be found in the “Classic QlikView” theme on SenseTheme.

In those cases where categories have an inherent order, for example from strongly agreeing with a statement to strongly disagreeing, this palette doesn’t work as well. You can sort of make out what percentage of respondents agree or disagree with a particular statement, but you’ll frequently have to look at the legend to see which color corresponds to which intensity of (dis)agreement.

A better way to visualize this data is to use a sequential color palette, which we’ll see in the next section.

Sequential colors

A sequential color scheme doesn’t use different colors, it uses different shades, or “intensities” of a single color. The example below shows the same chart as before, but we’ve changed the colors to a sequential color palette. This is much easier to read. The darker the color, the more strongly respondents agree, and vice versa.

You may wonder “If this color palette works well, why wasn’t answer C the best answer then?”. That’s because there’s an even better color scheme that we could be using, the diverging color palette. Let’s take a look at that next.

Diverging colors

A diverging color palette is very useful for data series where there’s a natural midpoint. This enables the reader to see clearly see values that are above and below that value. For example:

• A zero value, for example temperature in degrees Celsius, where we can distinguish between above and below freezing point.

• Devation against an expected value. For example a budget, target or previous period, where we can distinguish between values that are exceeding expectations and values that are falling short.

• A halfway point, for example how many respondents agree or disagree with a particular statement.

The chart below shows how a diverging color palette can be applied to our earlier example. This makes the chart even more easy to read, the more blue the value is, the stronger respondents agree, while more red signifies more disagreement. Those who don’t really hold a strong opinion are shown in the middle range in white/gray.

Looking back to our original question, we want to visualize a data series that ranges from -100% to +100%. Without any additional information, we can assume that we want to show deviation from the midpoint of 0%. That makes a diverging color palette the best choice in this scenario.

That’s it for now, but of course there is lots more to explore so we’ll certainly be revisiting this topic in future blog post.

More from the Bitmetric team