Insights at a Glance: The Power (and Pitfalls) of Data Visualization

The process of gathering and analyzing customer experience data involves several translations.

  • Desired outcomes are translated into measurable attributes.
  • Attributes are translated into feedback tools (such as survey questions).
  • Customers translate their sentiments into quantifiable scores – data points.
  • Data points get translated into ratios, averages, and frequencies.
  • The collected data can then be translated into knowledge.

Customer experience researchers and data analysts are charged with the task of following all these translations step-by-step, but in the end, most non-analysts are only interested in that ultimate goal – the knowledge.

That’s where data visualization comes in. As humans, our understanding of data relies heavily on how that data is presented to us. Visualizations are among the best tools for making that final translation from information to insight.

Visualizations make data memorable.

Have you ever struggled to remember the name of a particular actor even though you can see his or her face clearly in your mind? For most people, it’s easier to remember something they have seen than something they have heard or read. Once translated into an image or graphic, data becomes less abstract and makes a distinct impression – one with staying power.

Visualizations make clear connections between parts of the whole.

Insight comes from connecting A to B to C and so on. A number by itself doesn’t say much until it’s put into a context of other numbers. Sometimes, even a simple table can help, but the more complex your data, the more difficult it is to glean insights just from reviewing the figures. Visualizations are handy shortcuts that make the relationships between data points immediately clear, getting you straight to that “light bulb moment.”

Visualizations influence how data gets interpreted.

Data is objective, but visualizations are subjective. There are a number of factors that influence the message a person receives from looking at a graphic: size, scale, color, even font choice. What’s more, the most basic types of visualizations – pie charts, bar charts, line graphs, and scatter plots – are each best suited to different purposes. Using the wrong kind of graphic for the type of data can be misleading or obscure the possible conclusions.

The pie chart can show what percentage of respondents in 2015 chose each answer. It cannot show a comparison to other years, nor can it show any other metrics.

The bar chart can show all three years’ worth of data at once. It’s decent for showing the change in each measurement over time, but a line chart would be a better fit.

The line chart shows how each measurement changed over time. Note that instead of representing the total for each year between 2013 and 2015 (which would produce very short lines with little variation), this chart shows month-over-month trends for each response. Line charts work best when comparing a more thorough set of dates.

A scatter plot, would be a good choice for displaying each response to a particular survey question. The above charts each show the total number of responses for each category, out of 1,285 responses to the same question, and each of those totals represents one point on the chart (or one slice of the pie). In a scatter plot, each of those 1,285 responses would generate its own dot, and the way those dots group together would reveal the trend.

Beware: Things aren’t always what they seem.

Visualizations are useful for drawing conclusions at a glance, but sometimes looks can be deceiving. Like statistics, visualizations can be manipulated to produce a particular effect – for better or for worse. For example, bar and line graphs depend on the scale of their vertical and horizontal axes. By increasing or decreasing the scale of either axis, bars can be made to look smaller or larger, or trends to look more or less dramatic.

Because the maximum value on this bar chart is 560, the top of the range (vertical axis) is 600. The red bars for “mostly satisfied” are far longer than any of the light blue bars for “completely dissatisfied,” making it look like hardly any respondents chose the latter.

The second chart uses the same data as the previous one, but omitting some of the categories. Without the higher-scoring “mostly satisfied” values, now the maximum value is 338, making the top of the range 400 instead of 600. Even though none of the actual values changed, now the blue “completely satisfied” bars look much more significant. Likewise, some of the light blue “completely dissatisfied” bars that barely appeared on the first graph are now visible here.

All it takes is a closer look to see whether the graph’s scale is skewing the effect, but ideally, you should be able to get an accurate sense of the information at just a glance. Otherwise, the visualization isn’t doing its job effectively.

These aren’t the only kinds of visualizations, of course, especially in this age of customized metrics and creative infographics. They do, however, represent the basis of data visualization, and knowing how to read them prevents those valuable insights from getting lost in translation.

More articles on using data to your advantage:

Share the Post:

Related Posts