SHRM-SIOP Science of HR White Paper Series

Data Visualization: Get Visual to Drive

HR’s Impact and Influence

Evan F. Sinar
DDI

Copyright 2018
Society for Human Resource Management and Society for Industrial and Organizational Psychology
 Evan F. Sinar
DDI
[email protected]

Evan Sinar is Chief Scientist and a Vice President at DDI. He designs

and executes analytics linking leadership programs to business
outcomes and produces actionable insights about talent trends and
people practices, including leading DDI’s Global Leadership
Forecast and High-Resolution Leadership research programs. Evan authored data
visualization chapters in the book Big Data at Work and the SAGE Encyclopedia of Industrial
and Organizational Psychology. He is an active social media author and curator and has
been recognized as a top worldwide influencer on data visualization as well as other
leading-edge topics such as HR analytics, big data and the Internet of things. He has
authored over 70 professional presentations and articles for major publications and
professional conferences and serves on the editorial board of several journals. Evan holds a
Ph.D. in industrial and organizational psychology and is a Fellow of the Society for
Industrial and Organizational Psychology.
ABSTRACT
The goal of this paper is to establish a rationale and roadmap for adopting visualization techniques to
represent HR data. By using these approaches, HR professionals can unlock a level of data
engagement and influence difficult to achieve using tables and numbers alone. Many HR datasets are
well-suited to visualization as a component of storytelling and messaging for nontechnical,
professional and executive audiences. Data visualization is also a more approachable skill than ever,
with many easy-to-use and quick-to-learn tools available. This paper provides an overview of
foundational concepts for effective visualization design and a recommended set of practical resources
and principles for rapidly incorporating data visualization into your own work.

Data Visualization: Get Visual to Drive HR’s Impact and Influence

Data visualization—defined as representing quantitative variables through visual

properties such as position, length, area and color—is an immensely powerful

technique for exploring and communicating information. To a degree far beyond

traditional text or number-based formats, well-designed visualizations make data

accessible, clear, enticing and engaging to business audiences. Yet, despite the recent

surge in visualization’s use, many HR professionals aren’t proficient in these compelling

approaches. Thus, they miss golden opportunities to expand their influence and

audience reach.

The purpose of this paper is to:

• Provide an overview of key concepts for effective data visualization.

• Review the evidence base supporting its use.

• Recommend a set of practical resources for adopting these techniques in

your own practices.

1
Why Visualization Now?
The historical roots of data visualization extend back centuries: from the early 1700s-

era work of pioneers such as William Playfair (1759-1823, considered a pioneer in

graphically displaying statistical information) and Charles Minard (1781-1870, best

known for creating data-rich geographic representations, most notably his stunning

map of Napoleon’s 1812 Russian Campaign) to the foundations built by (still active) late

20th-century practitioners William Cleveland and Edward Tufte. However, the past

several years have seen an explosion of prominence for data visualization within the

business community.

I see three main trends driving this upswell: big data availability, data-driven

decision-making and information transparency.

1. Big Data Availability. The ever-increasing proliferation of information—in scale,

speed and variety—has created business environments where analysis expertise to

sort through and extract insights from these large-scale data is increasingly

valuable. Data visualization taps into our formidable visual processing skills to

counteract and inoculate against big data overload and confusion.

Visualization techniques also pair well with other facets of data science. These

techniques serve as a “face” for the advanced machine learning and artificial

intelligence methods in expanded use in many organizations with the data gathered

from and about their employees.

2. Data-Driven Decision-Making. Many organizations have shifted away from “gut

instinct”-based decision making, to instead adopt an expectation—and often a

2
 mandate—for data to drive strategic and tactical choices. This has hugely increased

the proportion and number of business professionals being asked to make decisions

based on the ever-growing amount of data. Through data visualization techniques,

organizations can involve a broader range of experts to explore and extract value

from their datasets.

3. Information Transparency and Ethics. The initial wave of unbridled optimism

about big data’s impact has faded, replaced—particularly for employee data—by an

orientation toward data-gathering that guards against bias and unethical data use.

“Black box” analytical models resulting from complex data science procedures are

also facing increased scrutiny because they are largely impenetrable to lay

audiences. These forces place a premium on data transparency. Visualization is

inherently well-suited to information transparency, providing a platform for data

openness, engagement and interactivity. It also provides a more direct view of the

raw data, counteracting the potential for biased interpretations on the part of the

researchers and giving the audience an opportunity to see the underlying data for

themselves.

The HR Skills Gap—and Opportunity

For HR professionals—serving as the conduit between employee data and talent

strategy—visualization is a vital area of proficiency to fuel success. Yet, it’s not typically

a strength for those in the HR function.

In data gathered in 2017 (DDI, The Conference Board & EY, 2018), we found that

only 16% of organizations were successfully implementing data visualization and

3
storytelling techniques. This is despite the utility these methods have in linking an

organization’s bench strength (supply of leaders ready to immediately step in to fill

critical roles) and its financial success.

These skills gaps within HR have further consequences at the organizational

level—for example, where HR professionals are responsible for growing others in these

areas despite often lacking these skills themselves.

Business Applications and Research

Data visualization has three main objectives: exploring data, explaining it and building

others’ engagement in resulting actions.

1. Exploring Data: Data visualization allows more efficient detection of

relationships, patterns and differences within people datasets than information

presented in tabular or textual form.

2. Explaining Data: Data visualizations convey complex concepts in a format that

surfaces and clarifies data trends. Visualized forms of data also guide audiences

to contribute their observations and ideas about why and how patterns exist.

3. Building Engagement: Data visualizations elicit a powerful and unique type of

audience connection to quantitative information that is unlikely using more

common business presentation formats of endless bulleted text lists or

inscrutable tables of numbers.

Data visualization research conducted in a business environment is in early

states but rapidly gaining momentum. Three recent research studies provide a

4
promising foundation for visualization’s advantages over more traditional forms of

business communication.

A 2014 study by Pandey and colleagues found the visually-presented

information was more persuasive and resulted in greater attitude change to identical

information presented using tables. Al-Kassab and his research team (2014) found that

visualization improved managers’ information synthesis and processing abilities while

making decisions. And a 2015 study by Kernbach, Eppler and Bresciani demonstrated

higher attention to, agreement with and recall of business strategies when presented

visually.

Foundational Issues
Objectives and Business Questions. Data visualization is most effective when well-

aligned with a business question and sought-after action. That is, what is the goal of the

analysis and resulting visualization? Five common visualization objectives and example

associated business questions for HR/people data are to:

1. Compare values across groups—for example, employee engagement scores by

region, generation or function to target wellness investment to remedy

deficiencies.

2. Display connections or relationships between variables—for example, between

employee level and turnover risk score category to guide retention efforts.

3. Show hierarchical or part-to-whole structures – for example, number of yearly

hires by role within a department within a division to allocate onboarding budgets.

5
4. Illustrate change over time—for example, training completions by course and by

quarter to design learning paths and phase out underused courses.

5. Display data patterns within maps or geographies—for example, average

assessment score by country to inform recruitment strategies.

What Data are Best to Visualize? While there is no one “right” kind of data to

visualize, certain data characteristics are better for taking advantage of visualization’s

strengths.

Generally, higher-volume data (for example, from large-scale employee

populations) are more suited to visualization than low-volume data. Over-time data—

for example, multi-year survey or test administrations—are particularly well-matched

to visualization techniques, as there are entire classes of visualizations designed to be

used with data spanning time periods.

Visualization is also especially powerful for multivariate data—that is, described

by several distinct dimensions such as competency scores, as well as for data

segmented into hierarchical, nested categories such as region, department and job

title. Geographical, map-based data can also be represented visually using techniques

designed for that purpose.

Of course, baseline standards for data quality and veracity must be met for

visualization as with other analytics. For a thorough discussion of this topic, I

recommend the SHRM-SIOP Science of HR White Paper “Trends and Practices in

Talent Analytics” (Kaur & Fink, 2017).

6
 Audience and Format. Before creating any data visualization, consider your

audience and format(s) for presentation. What’s the audience’s familiarity with

visualized data? What’s worked well (and what hasn’t) for this audience in the past?

Ultimately, what insight, decision and actions are you looking to drive with your

audience?

Answers to these questions will guide what types of visualizations you use, as

well as the level of detail you share. A group of senior executives may expect a more

straightforward, recommendation-focused visualization whereas a technical audience

may seek additional detail causing you to include more such information in the

visualization. If possible, create separate presentations for audiences varying widely in

their orientation toward data visualization to avoid creating a single view of the data

that doesn’t fully meet the needs of either group.

In terms of presentation format, visualizations presented on slides will typically

need to be higher-resolution, with larger text and visual elements, and less detailed

than visualizations shared in a printed report. Slide presentations can have their own

challenges—factors such as room lighting, screen size and projector resolution can have

a major impact on which visualizations you’ll use.

If presenting using a projector, I strongly recommend a trial run of your

visualizations in the same room and with the same A/V setup as the final presentation.

Be sure to factor in enough time to make adjustments (e.g., enhancing contrast,

enlarging graphics, adding call-out boxes to highlight key data trends) to ensure that

visualizations show up in high legibility and interpretability.

7
Visualization Techniques
I recommend three overarching techniques to increase the accuracy, interpretability

and persuasiveness for the data visualizations you create: contrast, annotate and

sequence.

Contrast. Use contrast within the visualization to draw attention to the most

notable segments of or patterns within the data. This can be done by using focal colors

(see color recommendations below) to accentuate critical visualization features. It’s

important to note that creating contrast also involves deemphasizing less important

data elements. This may be because they aren’t statistically significant, are too small to

be meaningful or simply aren’t the focus of a specific part of the presentation (but may

be given focus at a later point). This is best done by using gray to fade nonfocus

elements out of the foreground.

Annotate. Data visualizations are rarely complete immediately after emerging

from the program used to create them. Post-creation annotations are often essential

for clarifying your message for the audience. Annotations include labeling the

visualization not merely with a descriptive title (e.g., “Applicant volume over time by

region”). Create a title that drives action (e.g., “Recruitment staff needed in the

Southeast to support a surge in candidates”). Also, a thorough analyst will already

know the background for notable visualization features such as a spike in the trend line

in a particular quarter. “Callout” annotations to layer in this context will help the

audience understand the reasons for key patterns.

8
 Annotations can also be used to note data provenance such as the source and

date for the data. This pre-empts basic clarification questions from the audience and

helps build confidence in data quality. When used properly, annotations can improve

memorability of data visualizations, reduce working memory demands the audience

faces, and help establish a common frame of reference and understanding for the

visualization. Another valuable outcome is that informative visualizations allow

discussion time to focus less on basic questions and more on sophisticated discussions

about causes, actions and implications.

Sequence. Think of visualizations not as stand-alone data views but as

components in a larger narrative that you’re building and sharing with the audience.

Well-sequenced visualizations can match and take advantage of established

storytelling principles. For example, consider Freytag's Pyramid (see Figure 1) —from

early “exposition” use within a presentation to introduce and set up the topic, to next

highlight “rising action” as supporting facts are revealed, to the “climax” revealing the

primary insight, to “falling action” showing how the conflict is addressed, and finally to

the “conclusion” as a visual summary of the entire story.

9
Figure 1: Freytag’s Pyramid

Source: https://www.clearvoice.com/blog/freytags-pyramid-using-classic-storytelling-

techniques-successful-marketing/

Sequencing visualizations can also involve adopting concepts of a “data comic”

(Bach, Riche, Carpendale, & Pfister, 2017): progressive panel-by-panel (or slide-by-

slide) views shifting from one to an alternate view of the same data, changing the time

scale (one year to the next) or moving from a higher to a lower level of detail. See

Figure 2 for an example of a data comic-style sequence of visualizations.

Figure 2: Data Comic Panel-by-Panel Example

Source: http://aviz.fr/~bbach/datacomics/Bach2017datacomics.pdf

Sequencing that gradually steps up from a basic to a complex data visualization

(orienting the audience to each step before moving on) is also a counterweight to

10
pressures based on the overstated assumption that “the audience needs to be able to

understand a visualization in five seconds!” Taken literally, this approach will lead to

simplistic data views (and weaker impressions about the presenter). Rather than

deploying visualization to simplify the complex (often futile or misleading given the

true sophistication of the modern business environment), instead use well-designed

(high-contrast, annotated and sequenced) visualizations to convey and clarify the

complex.

Avoiding Common Design Missteps. With visualization, minimalist design

often produces maximum clarity—extraneous elements such as backgrounds, 3D

effects, gridlines, borders and shadows draw focus from the data, can reduce accuracy

of interpretation and should be stripped back as much as possible.

Responsible data visualization also includes awareness and avoidance of

misleading (and, if done intentionally, unethical) design choices—for example,

truncating the Y-axis for column charts (which can make even small changes appear

massive); using dual Y-axes (which can show a false correlation between two variables).

Make visualizations easy for the audience by direct-labeling data points

whenever possible. This is preferable to relying on axes and captions focusing them to

oscillate back and forth across the visualization. Also, provide common baselines—for

example, aligning bar segments for different groups by their left edge—to aid

comparisons when presenting multiple groups or categories in the same visualization.

11
 Data Visualization Versus Infographics. Infographics are related to data

visualization. Yet, the two forms of information presentation are distinct in several key

aspects.

Infographics place a stronger emphasis on design and aesthetic features,

sometimes at the expense of data accuracy. Data visualizations are typically defined as

being automatically generated (e.g., using software), in contrast to infographics, which

are often illustrated with the involvement of a graphic designer.

Infographics tend toward subjectivity whereas data visualizations are more

objective in their message. Infographics are often made up of several individual data

visualizations, arranged in a story sequence or around a common theme.

Though this white paper is about data visualization, well-designed infographics

share many of the same effective design principles (e.g., use of color; graphical

properties; annotations). I recommend the work of the Accurat Studio

(https://www.accurat.it/) and David McCandless (https://informationisbeautiful.net/).

Visualization Principles and Techniques

Accuracy of Visual Features. With an eye toward accurate audience interpretation of

the patterns illustrated by a data visualization, the various graphical features used to

denote data are far from equal and should be prioritized accordingly.

Research summarized by Lunbald (2015) rank-ordered visual features by

accuracy (Figure 3). Features higher on the list will be better-suited for visual design

features, leading to higher interpretative accuracy than those lower on the list.

12
 For example, for quantitative interval/ratio data, varying data element positions

will be more accurately interpreted than varying their length, which, in turn, will be

more accurate than varying their volume.

It’s important to understand these distinctions when making design choices for

a visualization. Use features higher in these lists to illustrate more important

distinctions within the data. Avoid those lower in the list unless necessary (Figure 3).

Variation in relative accuracy also explains why certain visualizations are

recommended over others. For example, because volume differences are interpreted

less accurately than those based on area or length, 3D graphics are rarely appropriate.

Figure 3: Visual Features by Data Type, Ranked by Interpretative Accuracy

13
 Using Color. Color is both a powerful and problematic graphical property for

representing data. Many visualizations used in business settings use color excessively or

inappropriately as a graphical property.

One particular risk area is the use of “rainbow” color scales. These fail to

recognize that visual distinctions are not consistent across the color spectrum, leading

to inaccuracy and misinterpretation about the magnitude of score differences (Kosara,

2013).

A more accurate approach is to consider color variation using saturation rather

than hue for continuous scales. A second color-related risk area is use of palettes that

don’t account for colorblindness. Online resources exist for checking graphics for

colorblindness suitability and for identifying color schemes that accurately represent

quantitative distinctions (see, for example, http://colorbrewer2.org).

Types and Techniques

Data visualization types are guided by the research question being asked—or the data

story to be told. I use Kirk’s (2012) five-category taxonomy to classify visualization

types, which also build on the five main objectives for visualization as listed above.

For each category, I provide a brief description of the key question addressed by

the visualization, identify several representative types and show simplified versions

(without labels or annotations for this purpose). For further “in the wild” examples of

each type—very useful for getting ideas and gauging feasibility for your own data—I

highly recommend searching Google Images for the name of each type.

14
 • Comparing Values Across Groups—data visualizations that aid comparisons

among multiple groups based on their quantitative properties. These

visualizations are designed to illustrate data patterns (e.g., which groups are

larger than others; which groups differ most between two time periods; which

groups are over-represented in how their members are distributed across

categories). Examples are slopegraphs, Sankey/Alluvial diagrams, word clouds,

bar charts (Figure 4).

Figure 4: Representative Visualizations for Comparing Values

• Displaying Connections or Relationships Between Variables—data

visualizations designed to show how two or more variables relate to one

another. These visualizations display patterns such as strength of relationship

between variables, what values most often co-occur with others, and the

multivariate profiles of particular cases. Four examples are bubble plots,

scatterplots, parallel coordinates and word trees (Figure 5).

15
Figure 5: Representative Visualizations for Displaying Connections or Relationships
Between Variables

• Showing Hierarchical or Part-to-Whole Structures—data visualizations

designed to depict patterns such how a single data category relates within a

broader grouping (e.g., the scope and depth of an ordered data structure; how a

group breaks down into its proportionate members). Four examples are circle

packing diagrams, tree maps, sunburst diagrams and pie/donut charts (Figure

6).

Figure 6: Representative Visualizations for Showing Hierarchical or Part-to-Whole

Structures

• Illustrating Change over Time—data visualizations designed to show variation

across a time span. These visualizations show data patterns such as how certain

groups become larger in an absolute or relative sense compared with others

across years and the trajectory and magnitude of a category’s growth or decline

over time. Four examples are horizon charts, stream graphs, bump charts and

line charts (Figure 7).

16
Figure 7: Representative Visualizations for Illustrating Change over Time

• Displaying Data Patterns Within Maps or Geographies—data visualizations

designed to display data patterns overlaid on maps, with geographical entities

the foundation onto which other visual properties (for example, value-sized

shapes or color saturation) are placed. Two example types are Choropleth maps

and dot maps (Figure 8).

Figure 8: Representative Visualizations for Displaying Data Patterns Within Maps or

Geographies

For an expanded, searchable view of over 150 data visualization types, including

descriptions, definitions, data structure guidance, and examples, peruse The Data Viz

Project (http://datavizproject.com/).

17
Tools and Continuous Learning
Tools. Approachability and ability to create high-quality data visualizations may be

easier than you think. Although it’s certainly possible to invest time and expense to

either learn a visualization-focused programming language or to invest heavily in

commercial visualization software, it’s no longer necessary to do so for the vast

majority of HR professionals seeking to adopt and apply these techniques.

With a wide range of data visualization tools available—and growing every day—

I focus in this paper on three tools that are open-source or widely available, involve a

shallow learning curve and are capable of producing presentation-quality graphics.

1. RawGraphs (http://rawgraphs.io/) is an open-source online tool that

allows users to enter their own data and to generate many of the visualization types

listed above, and that can be paired with a vector graphics editor to rapidly produce

free, high-quality visualizations.

2. Voyant (http://voyant-tools.org/) is an open-source online tool for processing text-

based data to produce word trees, bubble lines and a dozen other visualizations for

graphically displaying patterns in a large-scale text corpus.

3. Microsoft Excel. Excel is surprisingly capable for creating visualizations, particularly

after overruling its default settings, which violate many of the foundational

principles above. Several prominent data visualization practitioners also release

templates designed specifically for Excel, and the 2016 version of the program

further extends its visualization abilities.

18
 A Small-Scale Example. The data and steps below walk through a small-scale

example of data visualization using the RawGraphs tool described above; this example

can be scaled up to match your own data specifications.

Table 1. Sample Data for Bump Chart

Region 2013 2014 2015 2016 2017

Region 1 800 529 672 664 578

Region 2 1122 622 728 531 866

Region 3 268 647 196 92 109

Region 4 353 524 591 253 425

Region 5 322 273 143 622 737

To create a visualization using this dataset, start by opening the website

http://rawgraphs.io/.

• Click on the “Use It Now!” button to progress to the “Load your data” screen.

• Copy and paste the data above into the entry box.

• Click on the “Click here to stack it” button in the lower right, then select

“Region” as the dimension to stack on (this reorients the data into a vertical

format).

• Scroll down to view the available visualization types and select “Bump Chart.”

(Note: Select “Try our samples” in the “Load your data” section to view and

experiment with additional visualization types.)

• Scroll further to configure the visualization: Drag and drop “Region” into the

Group box, “column” into the Date box and “value” into the Size box.

19
 • The visualization below will appear. The “Download” section can be exported

directly into an image (.png) format (picture file with transparent background),

or to a vector graphics (.svg) file for further editing in a graphics program to

adjust font sizes, labels and so on. (I recommend Inkscape,

https://inkscape.org/en/, for this as a free and richly featured image processing

tool.)

• The resulting example graphic (Figure 9) shows year-over-year trends by region

(X-axis is year; wider streams indicate larger values; higher placement on the Y-

axis indicates rank-ordering of regions by year).

Figure 9: Example Bump Chart Created Using RawGraphs (Pre-editing in Inkscape)

Sources for Continuous Learning and Inspiration. Befitting a rapidly emerging

and energetic field, many of the leading data visualization practitioners are prolific

curators and authors on the topic. I recommend the work of the four experts below for

staying current on new data visualization tips, templates and examples to emulate:

20
 • Andy Kirk (http://www.visualisingdata.com/blog/).

• Stephanie Evergreen (http://stephanieevergreen.com/category/blog/).

• Cole Nussbaumer Knaflic (http://www.storytellingwithdata.com/).

• Robert Kosara (https://eagereyes.org/).

Beyond the work of these experts, I advocate tracking the Twitter hashtag

#dataviz to efficiently monitor a steady stream of visualization examples and guidance.

Seek further visualization inspiration through compilations such as the Information is

Beautiful awards (https://www.informationisbeautifulawards.com/showcase) and

Harvard Business Review’s Visual Library (https://hbr.org/offers/visual-library) and year-

end summaries (e.g., https://hbr.org/2016/12/what-we-learned-about-management-in-

2016-in-19-charts).

Conclusion
HR professionals armed with skills in information visualization will be more effective in

exploring the data they draw on to make workplace decisions, in explaining

quantitative trends to nontechnical audiences, and in engaging and persuading others

about their messages.

Visual storytelling skills are highly developable, and they are also exceedingly

valuable. As HR professionals become proficient in data visualization (currently not a

skill strength for most), they can make their messages more compelling, memorable

and engaging to the full range of employees and business stakeholders they must

influence.

21
References
Al-Kassab, J., Ouertani, Z. M., Schiuma, G., & Neely, A. (2014). Information visualization to
support management decisions. International Journal of Information Technology &
Decision Making, 13, 407.

Bach, B., Riche, N. H., Carpendale, S., & Pfister, H. (2017). The Emerging Genre of Data Comics.
IEEE Computer Graphics and Applications, 38(3), 6-13.

DDI, The Conference Board, & EY. (2018). Global Leadership Forecast 2018: 25 research insights
to fuel your people strategy. Pittsburgh, PA: DDI.

Kaur, J., & Fink, A., (2017). Trends and Practices in Talent Analytics. Society for Human Resource
Management (SHRM)-Society for Industrial-Organizational Psychology (SIOP) Science
of HR White Paper Series.

Kernbach, S., Eppler, M. J., & Bresciani, S. (2015). The Use of Visualization in the
Communication of Business Strategies: An Experimental Evaluation. International
Journal of Business Communication, 52(2), 164-187.

Kirk, A. (2012). Data visualization: A successful design process. Birmingham, UK: Packt Pub.

Kosara, R. (2013). How the rainbow color map misleads. Retrieved January 7, 2018, from
https://eagereyes.org/basics/rainbow-color-map.

Lunbald, P. (2015) Second Pillar of Mapping Data to Visualizations: Visual Encoding. Retrieved
December 14, 2017, from https://blog.qlik.com/visual-encoding.

Mackinlay, J. (1986). Automating the design of graphical presentations of relational

information. ACM Transactions on Graphics (TOG), 5(2), 110-141.

Pandey, A. V., Manivannan, A., Nov, O., Satterthwaite, M. L., & Bertini, E. (2014). The
persuasive power of data visualization. New York University Public Law and Legal Theory
Working Papers, Paper 474.

Sinar, E. F. (2015). Data visualization. In S. Tonidandel, E. King, & J. Cortina (Eds.), Big data at
work: The data science revolution and organizational psychology (pp. 115-157). New York,
NY: Taylor & Francis.

How to cite this paper:

Sinar, E. F. (2018). Data Visualization: Get Visual to Drive HR’s Impact and Influence.
Society for Human Resource Management (SHRM)-Society for Industrial-
Organizational Psychology (SIOP) Science of HR White Paper Series.

© 2018 Society for Human Resource Management and Society for Industrial and
Organizational Psychology

22