Key points

  • Big Data is straining our ability to interpret information
  • This makes visual renderings of data more important than ever...
  • ...But organizations don’t think enough about how to visualize data
  • Part of the solution is to develop a common visual language between functions and units.
  • Consider both the user and designer’s biases.

Over the past two decades, the average business has seen the amount of data it collects grow astronomically. The ability of computers to analyze all that data has also multiplied, but not enough that they can make sense of the facts without the guidance of human experts.

Unfortunately, even context experts are limited in their ability to process data, guide computer-supported analysis and interpret findings. It’s not always easy to find the insights hidden within that mountain of data and then to communicate them clearly. Common situations can be extremely challenging to visualize. For example, how can you show the impact of even a single point of disruption across a complex supply chain – especially when, as is often the case, the underlying numbers that describe the disruption can be truly understood only by people with different kinds of expertise?

A surprising number of answers can be found by looking at a theory developed over a hundred years ago by an American scientist and philosopher named Charles Peirce.

Often referred to as the father of the Pragmatist school of philosophy, Peirce was concerned with how people communicated – in particular, with how they conveyed meaning through signs. In his Theory of Signs, Peirce concluded that developing an effective sign or symbol depended on answering three questions, which might be paraphrased as:

  1. What is the meaning this message is intended to convey?
  2. What are the key connections between the factors that make up that message? and finally,
  3. What is the best way to represent that message visually for a particular audience?

For a basic example of how to use Peirce’s three questions, imagine you are a Wall Street analyst assigned to develop an infographic that compares the performance of two companies. The answer to the first question (what is the meaning?) may be simple, such as “We need to show the difference in earnings for these two firms over the last 10 years.” The answer to the second (what are the connections between the factors in this data?), in this case, the equations used to derive those earnings. And the answer to the third, (what’s the most accessible way to explain this material to this audience?), might be a scatterplot of date-labeled points that compares the two companies’ performance (assuming that your intended audience is familiar with the format).

This might sound simple enough, but it can easily become much more complex. What if the problem is not so easily defined? What if the audience is not familiar with the graphical form chosen? Or if you have a view of the context that differs from the audience’s point of view, and focuses on elements of the context that are outside the audience’s experience? Such decisions can be crucial: the wrong choice may not only lead to an incorrect interpretation, it may encourage bad or at least ineffective behavior.

Unfortunately, many analysts and even entire organizations don’t worry much about these kinds of concerns. Many regularly put the cart before the horse in data visualization, first choosing their software and methodology and only afterward thinking about how that representation fits the understanding of the audience and the information they need to make an informed decision.

Fortunately, it’s possible to minimize the risk that your own bias can lead you to choose a less-than-useful visual interpretation. One way is to give managers access to multiple varieties of visual renderings, to familiarize them with the idea that data can be visualized in many different ways. There might be a single truth, but the slice that is essential for someone in the warehouse may be beside the point for someone managing a supply chain. Another is to actively question the assumptions behind a software system’s design, such as asking why a particular ratio has been given pride of place.

One path forward is for a company to develop a consistent system of visualization—a shared visual language that helps ensure that everybody in the enterprise interprets a set of visual forms the same way, including more exotic forms used by functional specialists. The language must be based on contextual needs and incorporate not only the present visual biases of the organization but the ways in which they are probably evolving. Explicit, company-wide adoption of a well-known group of signs (such as the change-dynamic notation of the Standard Convention : is a good place to start, but care must be taken to make sure it fits the company’s needs.

It’s also important to try to align this language as much as possible with external users of the data. Preferences for certain forms and interpretative tendencies allow design processes to not only sidestep mismatches in form and interpretation, but also to capitalize on those forms that resonate most with a particular audience.

Today’s smartest companies don’t take the design of their data for granted. They understand that the way you represent the message of your data becomes part of the data’s message. It affects not only how the data is read but also the manager’s future actions. Only by taking great care with data visualization can you make sure it stays consistent with what your numbers actually say and ensure you have communicated it in a visual language your audience will comprehend.

1This article is taken from “Fit, Bias, and Enacted Sensemaking in Data Visualization: Frameworks for Continuous Development in Operations and Supply Chain Management Analytics,” Elliot Bendoly, Journal of Business Logistics, 2016, pp. 1-12.

Elliot Bendoly is a full professor in the Department of Management Sciences, and was the 2015 OM Distinguished Scholar (Academy of Management). He is the co-editor of Visual Analytics for Management: Translational Science and Applications in Practice, scheduled to be published in 2017 (Taylor & Francis/Routledge, New York).

Elliot Bendoly Distinguished Professor of Operations and Business Analytics, Editor-in-Chief, Journal of Operations Management (2024+), SMB-Analytics Co-Director, Affiliate Intl. Institute for Analytics, Former Assoc. Dean of Undergraduate Students/Programs
Faculty Profile