Journal of Management Analytics

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tjma20

Enhanced data narratives

Judd D. Bradbury & Rosanna E. Guadagno

To cite this article: Judd D. Bradbury & Rosanna E. Guadagno (2021) Enhanced data narratives,
Journal of Management Analytics, 8:2, 171-194, DOI: 10.1080/23270012.2021.1886883
To link to this article: https://doi.org/10.1080/23270012.2021.1886883

Published online: 12 Mar 2021.

Submit your article to this journal

Article views: 220

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tjma20
Journal of Management Analytics, 2021
Vol. 8, No. 2, 171–194, https://doi.org/10.1080/23270012.2021.1886883

Enhanced data narratives

a*
Judd D. Bradbury and Rosanna E. Guadagnob
a
Business Analytics Program, Jindal School of Management, University of Texas at Dallas,
Richardson, TX, USA; bThe Center for International Security and Cooperation, Stanford
University, Stanford, CA, USA
(Received 17 December 2020; accepted 3 February 2021)

Data narratives are an emerging form of communication that employs enhanced

media for effective knowledge transfer of complex information. Researchers in the
fields of data visualization and artificial intelligence have begun to pioneer new
structures of communication to improve the efficiency of construction and the
retention of information provided by the knowledge transfer experience. In this
paper, we report the results of an empirical study conducted to compare the
performance of various narrative communication techniques including frame
based narrative visualization, documentary narrative visualization, computer
generated text narratives and human generated text narratives. We assess the
knowledge transfer performance for each of these data driven narrative
structures. Across all conditions, an identical set of knowledge retention
questions assessed participants’ recall of details from their assigned narrative
communication. Statistical analysis on group performance answering the
knowledge retention questions revealed that some narrative communication
techniques perform better with general audiences.
Keywords: analytic communication; data visualization; algorithmic storytelling;
video analysis; enhanced data narratives

Introduction
Data narrative is a category of communication methods that utilize an underlying
quantitative data set as content for a knowledge transfer experience using narrative
structure. As technological advances increasingly facilitate access to “big data”
(Manyika et al., 2011; McAfee, Brynjolfsson, Davenport, Patil, & Barton, 2010),
research teams are developing new methods for the construction of narrative com-
munication using quantitative data sets. Researchers in the fields of data visualization
and artificial intelligence are working to enhance historic traditions of human com-
munication. The need for enhanced communication techniques has evolved in paral-
lel to a rapidly expanding scope of data. Modern analytical methods have the capacity
to process data sets with thousands of variables and millions of records (Elmqvist,
Stasko, & Tsigas, 2008; Eriksson, Byrne, Johansson, Trygg, & Vikström, 2013;
Keim, 2002; Shneiderman, 2008). How do you properly construct communication
about four thousand different facets of a phenomenon that is one million records
deep?

*Corresponding author. Email: [email protected]

© 2021 Antai College of Economics and Management, Shanghai Jiao Tong University
172 J. D. Bradbury and R. E. Guadagno

To effectively infer and communicate meaningful information from datasets of

this size requires that millions of records be summarized and presented to accurately
represent the insights gained from the data without significant loss of the meaning.
Data narratives, in their emerging forms, are the early attempts at forming advanced
methods of human communication with the capacity to summarize and encode these
deep and complex data sets. Here we focus on evaluating enhanced data narratives
with summarizations of large record data sets.
Natural language written by a human author is a basic building block of human
communication. Humans have also created images and used them for visual com-
munication for many millennia, using the unique ability to encode enormous
amounts of information in a single picture. In the field of human- computer inter-
action (HCI), several methods of data narrative have emerged. This study focuses
on three particular methods including Visual Data Narratives, Documentary Data
Narratives, and Computer Generated Text Data Narratives. In this paper, we focus
on these methods as part of a model for measuring the performance of enhanced
data driven communications.
Visual Data Narratives summarize a quantitative data set through a process of
graphically encoding data elements and presenting them in an overall visual frame
(Hullman et al., 2013; Hullman & Diakopoulos, 2011; Segel & Heer, 2010; Ziemkie-
wicz & Kosara, 2008). Among others, visually encoded objects known as glyphs are
presented in a visual view as representations of summarized variables and values from
a quantitative data set (Borgo et al., 2013; Munzner, 2014; Ward, 2008). Information
is communicated through a visual frame or a series of frames that engage the audience
in a review of glyphs as representations of many data values. Documentary Data Nar-
ratives utilize a similar style of visually encoded data viewing with the enhanced struc-
ture of full motion video (Bradbury & Guadagno, 2020). Often the presentation of
evidence in Documentary Data Narrative is delivered using verbal “voice-of-god”
narration as a second channel of communication further improving the knowledge
transfer experience. The enhanced capabilities provided by a moving video experience
in Documentary Data Narrative opens up the possibility for exploration of animated
data visualizations in a short film (Amini, Henry Riche, Lee, Hurter, & Irani, 2015;
Kwon, Stoffel, Jäckle, Lee, & Keim, 2014; Robertson, Fernandez, Fisher, Lee, &
Stasko, 2008; Yee, Fisher, Dhamija, & Hearst, 2001). Computer Generated Text
Data Narratives summarize quantitative data sets with totals, ratios and values
woven into a journalistic text-based story (CITO Research & Narrative Science,
2015). The story is composed using algorithmic text in place of the human author,
constructing stories with typical phrases found in human communication.
Storytelling has been effective in human communication as expressed in the long
and diverse history of teaching, education, and socialization of information using a
narrative format (Green, 2004; Green & Brock, 2002; Oatley et. al., 2002). Utilization
of narrative in business strategy allows for the exploration of characters, relation-
ships, and cause and effect outcomes that simple bullet points cannot (Gershon &
Page, 2001; Shaw, Brown, & Bromiley, 1998). At early age, we are taught to
process information using the narrative arc of children’s stories. We emulate these
stories through the explanation of events that occur in our daily life. In essence we
are programming each other to understand our lives using our characters, our
relationships, and the retelling of outcomes from our decisions. Narrative, as a tech-
nique, helps us to place information in an appropriate context. Storytelling helps to
 Journal of Management Analytics 173

mentally construct a world where information content lives, connecting that world to
personal experience. In many ways, storytelling is the human operating system for
exchanging complex information within a context. This begs the question: which of
the extant enhanced approaches is the most effective technique for creating data nar-
ratives that produce a high degree of knowledge transfer? Below, we review the
benefits of each data-driven approach.
Visual Data Narratives benefit from a significant ability to summarize and com-
putationally encode complex data sets (Larkin & Simon, 1987). Quantitative data sets
with a large scope can be engaged visually with navigation, layering, so-called small
multiples, and multi-dimensional representation (Card, 1999; Munzner, 2009; Shnei-
derman, 1996; Yi, Kang, & Stasko, 2007). The challenge for Visual Data Narratives is
in the scope of audience familiar with frame based presentation of data visualization.
Frame based communication techniques have been used previously in comic books,
storyboards, slide show presentations, new journalistic approaches, and business
intelligence dashboards. While there is a significant amount of experience with
frame based presentation of information, the experience of a general audience with
data visualization presented in frame-based presentations with stepper navigation is
significantly less than the experience of the same audience with communication in
text narrative.
Documentary Data Narratives benefit from a large cross section of people experi-
enced in watching films and from engaging two channels of information processing
versus one (Mayer & Anderson, 1991; Mayer & Moreno, 1998; Mayer & Sims,
1994; Tversky, Morrison, & Betrancourt, 2002). Similar to Visual Data Narratives,
Documentary Data Narratives are constructed from data visualization where audi-
ences have significantly less experience interpreting the information. However,
general audiences do have a great deal of experience watching documentary videos.
When verbal narration (i.e. the “voice-of-god”) is engaged, Documentary Data Nar-
ratives benefit even further from the long history of verbal storytelling containing a
more universal reach of literacy.
Computer Generated Text Data Narratives benefit from looking and feeling
similar to traditional Human Generated Text Data Narratives. Without clearly indi-
cating the source of computer generated data narratives, research subjects may not
be able to determine the difference between a data narrative generated by a human
author versus one generated by a computer. Using the common communication tech-
nique of the written word, Computer Generated Text Data Narratives and Human
Generated Text Data Narratives share a familiar knowledge transfer experience.
Over a number of repetitive reading experiences, if the same construction template
is used, audience members might develop commentary about boredom with the con-
sistent pattern of the stories. This commentary is balanced by the positive feature of
consistency of information communication in a series of stories. Based on literacy in
general, Computer Generated Text Data Narratives have the ability to reach a very
wide audience. Computer Generated Text Data Narratives have challenges when
encoding a large scope of information in a concise way without the use of visual
imagery. Text describes data well, but its nature may be more limited in its ability
to summarize or computationally encode a larger scope of information.
Empirical research regarding the knowledge transfer performance of these
methods for enhancing data narratives has yet to address which of these methods
is superior. The purpose of this paper is to evaluate the knowledge transfer capability
174 J. D. Bradbury and R. E. Guadagno

for each of these emerging types of enhanced data narratives. Specifically, we examine
the knowledge transfer for each of these data narrative methods, alongside the classic
example of text stories written by humans. We assess the recall performance for par-
ticipants who viewed one of the four data narrative methods, communicating infor-
mation about the common data set.
Specifically, we explore three primary research questions related to a general audi-
ence in this paper:

. Whether enhanced forms of data narrative transfer knowledge more effectively

than traditional human generated text narratives?
. Are data narratives constructed with data visualizations more effective at
knowledge transfer than text narratives?
. Are Documentary Data Narratives composed with video of data visualizations
more effective at knowledge transfer than frame-based visualization data
narratives?

Enhanced forms of data narrative

With rising expectations to create meaning from an expanding ocean of data, tech-
niques for encoding and communicating information have begun to emerge. These
techniques are designed to meet the challenge of synthesizing and summarizing
large data sets into meaningful information for intellectual consumption. This
study examines three taxonomic groups of data narrative construction defined here
as Visual Data Narrative, Documentary Data Narrative, and Computer Generated
Text Data Narrative. Each of three classes of data narrative uses communication
techniques to represent a data set. Raw data sets are generated from observations
about a phenomenon. The first two classes are defined as Visual Data Narrative
and Documentary Data Narrative utilize data visualization to encode communi-
cation views representing the data set. The third class defined as computer-generated
text data narrative summarizes data elements into quantitative numbers. The quanti-
tative numbers are then mapped into common linguistic phrases using variables. Lin-
guistic phrases are then composed into narrative paragraphs using standard subject
matter templates (Narrative Science, 2015). The equivalent construction for a
Human Generated Text Data Narrative has the author summarizing a data set
into quantitative measures. The measures are then integrated into sentences and para-
graphs necessary to create a narrative story.

Visual data narrative

Visual Data Narrative refers to a style of information presentation that constructs
stories from frames of data visualization organized using a traditional narrative story-
telling arc. Visual Data Narratives use avisual frame or view constructed from a data set
as the building block for the overall construction of the data narrative. Data Visualiza-
tions in the Visual Data Narrative are typically constructed from an underlying data set
using shapes, maps and other visual encoding to reveal a particular insight about a
phenomenon represented by the data measurements. Views can also be constructed
using data objects like text, pictures, and isotypes (small iconic pictures like gingerbread
people) to help structure the overall narrative story (Haroz, Kosara, & Franconeri,
 Journal of Management Analytics 175

2015; Howson, 2014; SAP, 2015) Each frame of data visualization takes the shape of an
event in a traditional story, linking cause and effect within time and space (Bordwell,
Thompson, & Smith, 1997). Like narrative storytelling, Visual Data Narratives con-
struct a series of data visualizations positioned as events, linked by cause and effect
when appropriate, sequenced in time and located in space or geography.
One of the simplest methods for constructing a series of visualization events is
defined in the Aristotelian tradition of a three-act play defined as a beginning,
middle, and an ending (Aristotle, 1984). The beginning view creates a frame of refer-
ence for the audience. From the standpoint of spatial and temporal location, this first
frame should passively help the audience locate their proximity to the phenomena
represented in the Visual Data Narrative. The beginning might locate the audience
within an industry, company, geography, virtual concept, and/or point in time. Rheto-
ric, for example using the Aristotelian modes of ethics, logic, or emotion can be
engaged as a theme, bringing human meaning to the presentation. The introduction
of rhetoric helps the audience identify with the narrative. How does this information
relate to me? The middle view then uses a visualization view to highlight an issue or
insight with a significant amount of tension. The more tension created by this middle
view, the more effective the Visual Data Narrative. The ending presents a view of data
that reconciles or summarizes the tension created during the middle view. Risk miti-
gation, alternate courses of action, and new opportunities are fertile approaches for
reconciling the tension created in the middle view. Titles, headings, and annotations in
each of the views help to provide narration of the insights defined in the data
visualizations.
In cases where the content requires a more elaborate treatment of events, a struc-
ture similar to the classic Freytag Pyramid of Setup, Rising Action, Climax, Falling
Action, and Resolution may provide a more interesting composition (Freytag,
1896). The rising action, climax, and falling action provide more intricate treatments
of the presentation of tension in comparison to the simpler middle view of the Aris-
totelian model.
A more recent approach created by Neil Cohn is called Visual Narrative Structure,
emulating the presentation approach in comic books (Cohn, 2013; McCloud, 1993).
Visual narrative structure can be utilized to organize and sequence a series of cause
and effect-linked data visualizations designed around unique styles of visual views.
The visual narrative structure model sequences five styles of visual frames known
as establishers, initials, prolongations, peaks, and releases (Cohn, 2013). The Estab-
lisher frame sets up the action in a scene. Characters are often introduced in the estab-
lisher frame. The Initial frame begins tension as a path of narrative arc. The Initial
frame begins anticipation of the goal/peak of a path. The Prolongation view marks
the middle of the narrative arc and is designed to expand anticipation in the story.
This frame may be a somewhat passive or cyclical engagement of an on-going
process. The Peak frame marks the high point of narrative tension as an event. The
Peak often represents a critical event, or the goal of a path being achieved. The
Release frame then relieves the tension created in the overall narrative arc.
For the present study, we constructed a Visual Data Narrative in the form of an
interactive slide show with multiple frames of data visualization. The interactive slides
are presented in a predefined sequence commonly referred to as a stepper in Tableau
Storypoints. The categorization is derived from a navigation feature representing an
arrow of time that allows the user to advance through data visualization frames, one
176 J. D. Bradbury and R. E. Guadagno

step at a time. We selected this style of Visual Data Narrative as it is representative of

a large cross-section of previous narrative visualization work used by the New York
Times, The Guardian, and a number of other mass media publishers (New York
Times, 2015; The guardian, 2015). The stepper approach allows the media publisher
to check user attention through recorded clicks at particular steps. The interactive
slide show with stepper navigation is also the designated style of storytelling rep-
resented in the Storypoints functionality in Tableau Software, as well as the Stories
functionality in SAP Lumira Software (SAP, 2015; Tableau, 2015). The interactive
slide show with stepper navigation is a leading archetype that is representative of
Visual Data Narratives (Segel & Heer, 2010).
The Visual Data Narrative story presented in Figures 1–6 engage the more elab-
orate narrative structure models utilizing elements from the Freytag and Cohn
approaches to visual narrative structure. The first frame in the story is presented as
a setup or establisher to orient the audience to Intel as a one of the leading producers
of microprocessors in the world using a simple line of text (Intel Corporation, 2015).
The second frame is presented as an initial introducing the tension of profit decreas-
ing in the previous quarterly financial results. A simple bar chart of four quarters was
utilized to represent the underlying data measurements for quarterly profit. The third
frame was presented as rising action increasing tension as revenue is also falling short
of the previous year comparison. A bar chart referencing the quantitative measures of
annual revenue are provided for comparison. The fourth frame is presented as the
peak or climax of the story with earnings estimates increasing over a ninety-day
period. A trend line chart is presented as the visualization containing three plotted
points representing quantitative earnings measure estimates from investment analysts
at three separate points in time. The fifth frame is presented as falling action with the
quarterly and annual earnings measurements reported in alignment with analyst esti-
mates. A bar chart of reported earnings measures for the current quarter and the year
are presented beside earnings measures for the previous three quarters. The last frame
is a release using simple text statements suggesting that analysts were optimistic about
Intel with reported earnings reported consistent with expectations.

Figure 1. Visual data narrative of Intel earnings: setup-establisher.

 Journal of Management Analytics 177

Figure 2. Visual data narrative of Intel earnings: initial.

Documentary data narrative

Documentary Narrative Visualization is the presentation of a data visualization
experience using the techniques of documentary film (Bradbury & Guadagno,
2020). Engaging some of the techniques from Documentary film, the data visualiza-
tion views are set in motion using video and voice-over narration to enhance knowl-
edge transfer. Documentary Data Narratives can be defined as a video that presents
indexical visual evidence using verbal rhetoric (Bradbury & Guadagno, 2020). The
video is often structured in one of three modes of evidence presentation traditionally
known as expository, participatory, and observational (Nichols, 2010). Expository
videos structure evidence in a rhetorical frame, where the audience is transported
to a world of evidence that closely resembles personal experience. Expository narra-
tive visualization typically includes the features of indexical evidence, voice-of-god
commentary, narrative storytelling, and rhetorical proofs. Participatory data narra-
tive videos engage the visualization creator as a character in the narrative presen-
tation of evidence. The visualization creator may also provide voice-of-god
narration for the video. Observational videos present evidence in the most objective
manner possible without voice-of-god narration from the visualization creator or any
other source.
In our study, we constructed the Documentary Data Narrative example using the
expository mode of documentary containing presentation of indexical evidence,
178 J. D. Bradbury and R. E. Guadagno

Figure 3. Visual data narrative of Intel earnings: rising action.

voice-of-god commentary, and narrative storytelling. Our indexical evidence was rep-
resented using the identical six visualization frames utilized in our construction of
Visual Data Narrative with stepper navigation. The navigation of the visualization
frames was set to motion using video instead of user guided stepping. The video
was enhanced with voice-of-god narration engaging the text based insight annota-
tions at the top of each visualization frame. Rhetorical enhancements were intention-
ally minimized to avoid injecting bias into our measurements.
Statements using the themes of ethics and emotion were avoided altogether.
Careful attention was paid to include logical statements about insights that were con-
sistent with the constructions of Visual Data Narrative and Computer Generated
Text Data Narrative. See Figure 7 for the example of Documentary Data Narrative
used in our study.

Computer generated text data narrative

Computer Generated Text Data Narrative is a form of communication that utilizes a
computer algorithm to compose a text-based story. The composition process uses
Natural Language Generation (NLG), to compose sentences and paragraphs from
a quantitative data set by referring to a template of typical speech patterns that are
common for an industry (Reiter & Dale, 1997). Natural Language Generation is a
sub-field of artificial intelligence and computational linguistics that is concerned
with the construction of computer systems that can produce understandable texts
 Journal of Management Analytics 179

Figure 4. Visual data narrative of Intel earnings: peak-climax.

in English or other natural languages from some underlying and non-linguistic rep-
resentation of information (Reiter & Dale, 1997). The development of this type of
automated construction of news stories is being pioneered by organizations like Nar-
rative Science, Automated Insights, Yseop, and the Los Angeles Times (Automated
Insights, 2015; Narrative Science, 2015; Slate, 2015; Yseop, 2015). As a leader in
this sub-field, Narrative Science began with a business strategy of constructing jour-
nalistic stories for audiences of one that were cost prohibitive for human authors to
write (Borgo et al., 2013). The Gamechanger website application constructs a sports
news story for a little league team using an uploaded box score from the game (Allen,
Templon, McNally, Birnbaum, & Hammond, 2010; Business Insider, 2014). This
novel approach allows feature length stories to be constructed about little league
baseball games without incurring the high labor cost of a professional journalist.
In its second major initiative, Narrative Science created a partnership with Forbes
magazine where the Narrative Science algorithm creates stories about earnings pro-
jections for publicly traded corporations that are published online a few days
before the formal earnings announcements (The Atlantic, 2015). The earnings projec-
tions use a quantitative data set provided by Zacks Investment Research that is
applied to a narrative template of financial sentence phrases using conditional if/
then logical rules to construct a narrative text story about earnings that is then pub-
lished in an online column in Forbes Magazine (Forbes, 2015). The algorithm driven
investing column provides a low cost posting of a news story without the lower return
on investment and tedious effort of a human author.
180 J. D. Bradbury and R. E. Guadagno

Figure 5. Visual data narrative of Intel earnings: falling action.

Figure 6. Visual data narrative of Intel earnings: release.

 Journal of Management Analytics 181

Figure 7. Intel earnings documentary data narrative with voice-of-god narration.

In our study, we utilized the Computer Generated Text Data Narrative composed
by the Narrative Science algorithm that ran on July 13, 2015 in the electronic version
of Forbes Magazine (Forbes, 2015). The text of the narrative was provided to our
research participants in a single view as illustrated in Figure 8.

Figure 8. Intel earnings computer generated text data narrative.

182 J. D. Bradbury and R. E. Guadagno

Human generated text data narrative

A human authored text data narrative was constructed for our study to provide a
baseline of comparison for the enhanced data narrative communication conditions.
Journalistic stories about corporate earnings are standard fare for financial news
media services. The earnings announcement column is a typical assignment for new
journalists in the financial industry. Our objective in engaging a Human Generated
Text Data Narrative was the creation of an expert human authored story on corpor-
ate earnings using phrases typically used by human authors, edited to mirror the same
quantitative measures presented in the enhanced data narrative presentations. The
Human Generated Text Data Narrative for our study was constructed using sen-
tences from two earnings articles published by the financial media websites of
Zacks Investment Service and Seeking Alpha (Zacks, 2015).

Human communication proxy

The Mathematical Theory of Communication, presents a model of communication
where transmitters send messages across communication channels to receivers
(Shannon & Weaver, 1951). In the case of in person human communication, the trans-
mitter is the speaker, the verbal voice over the air is the communication channel, and
human listening is the receiver. Communication channels use presentational media
including voice, spoken words, facial expressions, gestures; and representational
media in the forms of books, text, paintings, and images; transmitted through tech-
nological media of telecommunications, internet, radio, and television (Fiske,
2010). The combinations of transmitters, receivers, and communication channels
provide unique approaches to human communication.
The most common form of communication is found with one person interacting
with another person during a face-to-face verbal exchange. Alternative approaches
utilize technological media substitutes for various communication elements of the
sender or receiver . Our findings in this paper speak to the concept of a Human Com-
munication Proxy that works to emulate aspects of a one on one conversation
between two people. Cognitively, the face-to-face human communication approach
appears to function as a normative model of audience preference (Barkhi & Pirkul,
1999; Daft & Lengel, 1986; Daft, Lengel, & Trevino, 1987; Dennis, Fuller, & Valacich,
2008; Denstadli, Julsrud, & Hjorthol, 2012; Kock, 2004; Kydd & Ferry, 1991; Lee,
2010).
Our results suggest that general audience preferences for enhancements to data
driven storytelling are being measured against the normative face-to-face human
communication model. The HCP substitutes elements of presence, verbal narration,
story structure, and representational media transmitted to the receiver. When the
HCP is engaged, audience preference follows how rich or realistic the representation
is evocative of the normative face-to-face sender/receiver experience. The closer the
emulation is to a conversation with another person, the more attractive the communi-
cation approach is to a general audience. Our findings support media naturalness
theory and the psychobiological model predicting that lower naturalness in the nar-
rative experience translates to higher cognitive effort in consuming a narrative (Kock,
2004). Thus, the empirical results presented in this paper suggest that general audi-
ences express a preference for communication models where they have more experi-
ence using them (Figure 9).
 Journal of Management Analytics 183

Figure 9. Human communication proxy.

Enhanced Data Narratives using Computer Generated Text utilize a communi-

cation model with an algorithmic author that crafts a text based representation of
a story. The text based story is published or provided in some manner by an algorith-
mic enhanced sender. The published text media functions as the communication
channel. The human receiver then consumes the text through a process of a human
reading it, using any of several media including newsprint, computer screen, or
smart device. The HCP in this model is found in the algorithmic story structure
applied in the creation of the computer generated text. The written text also functions
as a substitute for the in person narrative experience provided by a sender. As a sur-
rogate for the in person narrative experience, the text narrative transports the reader
to a narrative world (Gerrig, 1993; Green & Brock, 2002).
184 J. D. Bradbury and R. E. Guadagno

The enhancement of computer automated authoring is often not visible to the

reader. With an optimized natural language generation algorithm, the experience
of the receiver reading text may be indistinguishable from a similar experience
reading a human authored equivalent. The only distinguishable difference for the
reader might surface in the form of a predictable sentence pattern after reading
several narratives written by the same algorithm. This phenomenon is also known
to exist with formulaic human authors.
Visual Data Narratives utilize a communication model where a human author con-
structs a sequence of data visualizations in the form of a navigated stepper slide show.
Once published, the author becomes the sender. The step-based views, visualizations,
and rhetoric become the communication channel. The receiver interacts with the
Visual Data Narrative to decode and consume the intended messages. The stepper
slide show becomes the HCP standing in for the in person narrative experience. The
standalone nature of the Visual Data Narrative makes it absent any form of verbal nar-
ration. This leaves all the communication to be performed through avisual only experi-
ence. The audience may be transported to a narrative world, but that world is silent, or
limited to the narration provided from titles and annotations in the visual frames.
Documentary Data Narratives utilize a communication model where a visual data
narrative is recorded and transformed into a data narrative video. Voice over narra-
tion is combined with a predetermined step through sequence of data visualization
views. In most constructions, the creator becomes a character or narrator in the
video. The creator publishes the recorded video that functions as the sender. The
audio visual experience of the video playing is functioning as the communication
channel. The receiver plays the video, consuming the communication by watching.
The combination of voice over narration, visualizations, and in some cases visible
human actors, altogether represent a Human Communication Proxy. The video
engages visualization, titles and annotations in the visual channel of communication.
Voice over narrative acts as a substitute for the in-person narrative experience, enga-
ging the audio channel of communication. In concert, the video provides a higher
fidelity communication experience that emulates the pacing and structure of a narra-
tive conversation between two people.

Materials and methods

Our primary motivation in this study was to determine how effective each of the three
emerging forms of data narrative is at transferring knowledge to a general audience.
To accomplish this objective, we measured the knowledge transfer capability of the
three major forms of data narrative using the most typical examples of each construc-
tion technique.

Design of the study

Using an identical data set, four narrative stories were constructed using each of the
distinct communication techniques for narrative construction. The common data set
utilized for each story construction was sourced from the financial data set for the
Intel Corporation published by Zacks Investment Service (Zacks, 2015). Careful
attention was made to ensure that the specific data elements used in each story
were represented identically regardless of the type of narrative construction. The
 Journal of Management Analytics 185

factual content and statements were the same in each of the data narrative construc-
tions, providing focus on the construction techniques as dependent variable con-
ditions for measurement.
In the first version, an example of Visual Data Narrative using Storypoints func-
tionality in Tableau was created. The story building process constructs narratives
using a stepper type function to advance from one data visualization frame to the
next visualization in a sequence. The Visual Data Narrative was crafted with four
data visualization frames of bar charts and a line chart. A blue and orange color
scheme was utilized encoding blue objects as historical data and orange objects as
current data. An introduction frame with a defining text sentence about Intel was
sequenced in front of the four visualization frames, followed by a summarization
frame with two text statements about the optimism and buy rating of financial ana-
lysts. A summarizing statement of insight was defined above each visualization frame
in a light grey box, functioning as a sequential navigation button. A title was provided
for each frame summarizing the content with the assistance of standard axis labelling
and details where appropriate.
In the second version of our story, the Documentary Data Narrative used was
constructed as a full motion video walk through of the same data visualization
frames presented in the example of Visual Data Narrative. The example includes
human voice over narrative describing the dynamic changes to the data represented
in the data visualization.
In the third version, our source for the Computer Generated Text Data Narrative
is an investing column projecting earnings for the Intel Corporation (Forbes, 2015).
The earnings projection story was published a few days before the actual earnings
were released. The article was written by an algorithm at Narrative Science and pub-
lished online in Forbes Magazine (Forbes, 2015). The headline and text of this data
narrative was presented to research subjects in the same manner that it is presented
in the investing column at Forbes Magazine.
In the fourth version, the Human Text Data Narrative uses natural language
written by a human author sourced from Zacks Investment Service from an article
published for the equivalent Intel second quarter earnings announcement (Zacks,
2015). In cases where text sentences equivalent to the Computer Generated Text
Data Narrative were unavailable in the Zacks article, sentences were sourced from
an article published by Seeking Alpha for the Intel second quarter earnings
announcement (Seeking Alpha, 2015). Our objective was the construction of a text
narrative example containing identical content written by a human that was
aligned as much as possible with the Algorithm Text Data Narrative. Sentences for
the Human Text Data Narrative were incorporated from their article sources in
their entirety. Sentences in the Human Text Data Narrative were edited to ensure
the exact same quantitative data elements were presented.
The four types of data narrative communications along with descriptions of their
unique features appear in Table 1. Visual Data Narrative creates the test condition
listed first in the table with sequential frame based visualization construction. Docu-
mentary Data Narrative makes up the next test condition listed second in the table
with full motion video construction. Computer Generated Text Data Narrative
makes up the test condition listed third in the table with Natural Language Generated
text construction. Human Generated Text Data Narrative is the fourth test condition
listed in the table with human authored sentence construction.
186 J. D. Bradbury and R. E. Guadagno

Table 1. Recall performance of data narrative tests.

Each of the four examples of data narrative were presented to research subjects in
an equivalent manner with the narrative treatment residing within an equal sized 1024
by 768-pixel frame for viewing content. Instructions for reviewing each type of data
narrative were provided at the top left of the viewing frame as a headline in bold.
Study participants were randomly assigned to view one of the four data narrative con-
structions described in Table 1 and then asked to answer the same set of ten questions.
Participants were allowed to engage the content at their own pace, with all partici-
pants spending a roughly equivalent amount of time on their narrative treatment.
Participants were prevented from re- reviewing the data narrative content while
answering the questions. Participants assigned to review the same data narrative con-
struction were defined as a group of subjects representing that test condition.
Using the four groups of subjects, we applied a quantitative approach to examine
the significance of differences in the average quiz score performance for each of the
research subject groups.

Recall of results
A common data set was utilized to support each of the four types of data narrative
content. The exact same quantitative measurements were defined in each type of
data narrative. Where possible, sentences and headlines used the same size and font
across all of the data narrative types, paying respect to the original intent of the
type of data narrative construction. The factual statements used to construct the
stimulus materials for each data narrative is described below:

. Intel is a large global manufacturer of microprocessors.

. Three quarters of increasing profits followed by the most recent quarter of
profit declining by 29%.
. Revenue decreased this quarter by 6% to 13.07 billion dollars compared to
13.83 billion dollars in the same quarter of 2014.
. Consensus earnings estimates were .48 per share ninety days before the report,
.52 per share one day before the report, and .51 per share the day of the report.
. Earnings per share of .51 were expected for the quarter, with analysts expecting
$2.11 in earnings for the fiscal year.
 Journal of Management Analytics 187

. Analysts are becoming more optimistic as Intel reported earnings consistent

with analyst expectations today.
. Most of the analyst’s rate Intel a buy, equivalent to eight similar companies.

To assess recall, participants were asked to respond to 3 multiple choice questions

about the trend of a measure, 6 multiple choice questions requesting recall of a
specific quantitative measure, and 1 multiple choice question regarding the trend of
analysts sentiment. The questions related to a trend were defined as:

. What happened to profit last quarter?

. What happened to revenue last quarter?
. What was the trend of earnings estimates?

All trend questions provided three randomly sequenced answer choices for the
possible trend directions of increased, decreased, or stayed the same. The questions
requiring quantitative measure recall were specified as:

. What was the amount of change to profit?

. What was the amount of reported revenue?
. What was the earnings estimate 90 days before the quarterly report?
. What was the earnings estimate 1 day before the quarterly report?
. What were the earnings reported in the second quarter of 2015?
. What were the earnings estimated to be for the entire year of 2015?

Four answer choices were presented randomly in a vertical sequence for each
quantitative measure question. An example of the randomly sequenced quantitative
measure choices for the question regarding earnings estimates in the year 2015 is
found below:

. 3.16
. .89
. 1.84
. 2.11

The final question regarding the trend of analysts’ sentiment was stated as?

. Are analysts becoming more optimistic or less optimistic about Intel?

Three randomly sequenced answer choices were provided for analysts’ sentiment
as more optimistic, less optimistic, and about the same. The ten questions were pre-
sented in the same sequence of profit questions, revenue questions, earnings ques-
tions, and finally the sentiment question.

Method
Our study was conducted to assess the knowledge transfer capabilities of emerging
forms of data narrative. Data narrative constructions involving Visual Data Narra-
tive, Documentary Data Narrative, Computer Generated Text Data Narrative, and
188 J. D. Bradbury and R. E. Guadagno

Human Text Data Narrative were presented as data driven communication for par-
ticipants. Each participant was presented with one of the four data narrative con-
structions followed by a ten-item survey of questions designed to assess their recall
of the information presented in the data narrative. Prior to reviewing the data narra-
tives, participants were asked a series of preliminary questions to determine their
interest in financial investments, their knowledge in financial matters, and their
opinion of the products and reputation of the Intel Corporation. The survey con-
cluded with a short set of demographic questions regarding standard indicators of
age, gender, income, industry, level of education, and occupation.

Participants and procedure

Participants for the study were recruited using the Amazon Mechanical Turk plat-
form with worker requirements defining US Citizens with at least a 98% approval
rate for their hits (work items), that have completed more than 100 hits (Heer &
Bostock, 2010). Each participant received $0.50 for engaging the study. Upon com-
pletion, participants were prevented from performing the study more than once
using their Mechanical Turk ID. Participants began the survey by clicking on a
link and providing their informed consent. A tracking feature was used to ensure par-
ticipants were reviewing the narrative content they were assigned. We identified 5 par-
ticipants that diligently reviewed the content of their assigned data narrative and
completed all survey questions, except those regarding demographics. We respect
the decision of these participants to avoid providing their demographic information.
We included their responses for the questions not related to demographics in the
results of the study. Participants that did not review the content of their assigned
data narrative or did not answer the survey questions for knowledge recall were
excluded from the study.
A group of 512 adult participants (261 men, 246 women, 5 unreported) from the
United States were included in the study. They ranged in age from 18 to 74 (M =
35.84, SD = 11.35). Participants reported the following education levels: 43.8%
attained a 4-year College Degree, 18.8% completed some college, 11.9% attained a
2-year College Degree, 10.9% attained a master’s degree, 8.8% completed High
School, 2.1% attained a Professional Degree, and 2.1% attained a Doctoral Degree.
The three occupational categories of Legal, Computer/Mathematical, and Sales
accounted for 38.1% of the participants in aggregate. More than 61% of the partici-
pants were drawn from the following industries: Legal (13.9%), Computer/Math-
ematical (13.1%), Sales (11.1%), Office or Administration (7.8%), Arts Design &
Entertainment (7.8%), and Management (7.4%). Reported personal income was
under $25,000 for 20.3% of the participants, with 68.5% classified in the $25,000 to
$100,000 range, and 11.2% classified greater than $100,000.
We assessed participant responses regarding their interest in financial infor-
mation, their knowledge of financial issues, as well as their attitudes about the Intel
Corporation. For the participants surveyed, 92.6% exhibited at least some interest
in financial information, with 81.8% having some knowledge of financial issues.
Less than 3% of the participants had a negative bias about the Intel Corporation
and its products.
Participants were randomly assigned to view one of the four data narrative test
conditions (Group 1: Frame Based Visualization Story; Group 2: Video Visualization
 Journal of Management Analytics 189

Story; Group 3: Computer Authored Text Story; Group 4: Human Authored Text
Story). Answers to the 10 questions assessing recall/understanding formed the
scoring baseline for comparison of the groups.

Results
A one-way between-subjects analysis of variance (ANOVA) was conducted to explore
the impact of data narrative construction on knowledge transfer, as measured by the
survey questions on recall for the factual content of the data narratives. Participants
were divided into four groups according to the type of data narrative they reviewed
(Group 1: Visual Data Narrative; Group 2: Documentary Data Narrative; Group
3: Computer Generated Text Data Narrative; Group 4: Human Text Data Narrative).
There was a significant main effect of data narrative type on recall scores for the four
groups: F(3, 508) = 9.84, p < .001, ŋ2p = .05. The results of the post-hoc analysis is pre-
sented in Table 1.
Post-hocs using the Tukey Honest Significant Difference (HSD), highlighted by
an asterisk in Table 1, indicated that participants’ mean recall scores for the Visual
Data Narrative type of construction performed significantly lower with our test audi-
ence than the other three types of data narrative construction. Mean recall scores in
Groups 2, Group 3 and Group 4 did not differ significantly from each other.
Bonferroni corrections were applied to the p-values for the relevant comparison
groups. The null hypothesis of a Computer Generated Text Data Narrative being
as effective at knowledge transfer as a Human Generated Text Data Narrative was
affirmed with a value of p = .9992971, with a correction of (1.000000). Visual Data
Narratives underperformed Human Text Data Narratives with a value of p
= .0000246, with a correction of (.000025). Visual Data Narratives also underper-
formed Computer Generated Text Data Narratives with a value of p = .0000181,
with a correction of (.000018). Documentary Data Narratives significantly outper-
formed Visual Data Narratives with a value of p = .0001470, with a correction of
(.00015).

Discussion
In this paper, we began with the purpose of assessing various forms of data narrative
that have emerged over the last several years. Data Visualization, Narrative Visualiza-
tion, Data Stories, and Natural Language Generated Text Stories have become very
popular in the field of management. A great deal of time and energy has been devoted
to these increasingly used methods. Our analysis was designed to apply an empirical
approach to determine if enhanced data narrative communication techniques worked
as well as traditional text narratives, and if they did, which ones worked better, and
why?
Our first research question was an inquiry about the ability of enhanced forms of
data narrative transferring knowledge better than traditional human generated text.
In the case of Computer Generated Text Data Narratives, our results were consistent
with our expectation. We expected participants to understand and recall the infor-
mation presented in the Computer Generated Text Data Narrative at least as well
as the Human Generated Text Data Narrative. After reviewing the Computer Gener-
ated Text Data Narrative, participants scored as well on the ten survey questions as
190 J. D. Bradbury and R. E. Guadagno

the participants that had reviewed a Human Generated Text Data Narrative. Our
results suggest that our example of Computer Generated Text Data Narrative, as
an enhanced form of data narrative, did not perform better or worse than our
example of a data narrative written by a human. In terms of utility, a Computer Gen-
erated Text Data Narrative is not necessarily designed with the objective of improving
knowledge transfer compared to a human authored text data narratives. The objec-
tive of Computer Generated Text Data Narrative is providing a communication
medium that is as effective at knowledge transfer as a human authored text narrative.
The benefit of using Computer Generated Text Data Narratives is located in the cost
savings of not hiring a human to write the data narrative. Given this specific objective,
Computer Generated Text Data Narratives performed as well as Human Generated
Text Data Narratives in our study.
In the case of Visual Data Narratives, we expected Visual Data Narratives to
exhibit a superior capability of knowledge transfer. The results of our study suggest
that the Visual Data Narrative we created underperformed their text-based counter-
parts to a significant degree. Participants that reviewed the example Visual Data Nar-
rative content scored much lower on the ten survey questions requesting recall of
information presented in the data narrative. Participants that reviewed the example
Visual Data Narrative scored significantly lower on average than the participants
that reviewed the example Human Generated Text Data Narrative and Computer
Generated Text Data Narrative. The results of our study suggest that for the
chosen examples we created, the model of Visual Data Narrative did not transfer
knowledge as effectively as the examples of text based data narratives, when reviewed
by a general audience that may lack advanced visual data literacy.
Our second research question asks if data narratives constructed with data visu-
alizations are more effective at knowledge transfer than text narratives. Our study
had mixed results on this question. When our example of frame based Visual Data
Narratives was used without the enhancements of documentary film, the results of
our study suggested that the Visual Data Narrative did not perform better at knowl-
edge transfer than the text data narrative counterparts. On the other hand, our results
suggest that when our example of Visual Data Narrative is constructed using video
presentation, accompanied with verbal voice-of-god narration, the recall perform-
ance of participants improves to the same level as our examples of text data narratives
generated by a human or computer.
Our third research question explored if Documentary Data Narratives are more
effective at knowledge transfer than Visual Data Narratives. Our results suggest
that our examples of Documentary Data Narrative communication techniques out-
performed our examples of Visual Data Narrative communication techniques to a
significant degree. Participants that were asked to review the Documentary Data Nar-
rative example scored significantly higher on the ten survey questions requesting
recall from the information presented, when compared to the group of participants
that reviewed the same content presented in a stepper click based Visual Data Nar-
rative approach.
Our original aspiration in developing this study was to provide some empirical
evidence in favor of enhanced techniques for information presentation. The results
of our study suggest that enhanced techniques for information presentation do
play a productive role in the presentation of complex information for a broad audi-
ence. There is a large body of expert knowledge in the academic community that
 Journal of Management Analytics 191

suggests advanced technologies and techniques like Natural Language Generation,

Narrative Visualization, and Documentary Narrative Visualization have enhanced
abilities to encode and represent large and/or complex data sets better than tra-
ditional communication techniques like human authored natural language text.
The challenge of these enhanced techniques may reside in the literacy, or lack of
knowledge building experiences, of general audiences regarding these new forms
enhanced data narrative communication (Carlson & Zmud, 1999; Lengler, 2006).
General audiences may struggle with some of the communication techniques that
are less familiar to them. Natural language text generated by a computer algorithm
shares the most fundamental and common approach for human communication.
The enhancement provided by Natural Language Generation is in the production
of information content. Once that content is produced, it is available in the least
common denominator, or most universal communication technique of text. If our
example is instructive, Visual Data Narratives may have more challenges with
broader audiences, that more often than not, have no formal training in data visual-
ization. This poses unique challenges when Visual Data Narratives are utilized for
general audiences. The challenge of visual literacy in general audiences can be miti-
gated when Visual Data Narratives are enhanced with communication techniques
that may be more familiar to the audience. Specifically, adding voice-of-god verbal
narration, and video presentation seems to benefit from audience’s experience with
film, the comfort of a human voice, and a second cognitive channel of
communication.

Disclosure statement
No potential conflict of interest was reported by the author(s).

ORCID
Judd D. Bradbury http://orcid.org/0000-0003-3664-2315

References
Allen, N. D., Templon, J. R., McNally, P. S., Birnbaum, L., & Hammond, K. (2010,
November). Statsmonkey: A data-driven sports narrative writer. In 2010 AAAI Fall
Symposium Series.
Amini, F., Henry Riche, N., Lee, B., Hurter, C., & Irani, P. (2015, April). Understanding data
videos: Looking at narrative visualization through the cinematography lens. In Proceedings
of the 33rd Annual ACM Conference on Human Factors in Computing Systems (pp. 1459–
1468). ACM.
Aristotle, R. W. (1984). Rhetoric. New York: Random House.
Automated Insights. (2015). Introducing wordsmith. Automated insights website. Retrieved
March 1, 2015. http://automatedinsights.com/wordsmith/
Barkhi, R., & Pirkul, H. (1999). An experimental analysis of face to face versus computer
mediated communication channels. Group Decision and Negotiation, 8(4), 325–347.
Bordwell, D., Thompson, K., & Smith, J. (1997). Film art: An introduction. New York:
McGraw-Hill.
Borgo, R., Kehrer, J., Chung, D. H., Maguire, E., Laramee, R. S., Hauser, H., & Chen, M.
(2013, May). Glyph-based visualization: Foundations, design guidelines, techniques and
applications. In Eurographics (STARs) (pp. 39–63).
192 J. D. Bradbury and R. E. Guadagno

Bradbury, J. D., & Guadagno, R. E. (2020). Documentary narrative visualization: Features and
modes of documentary film in narrative visualization. Information Visualization, 19(4), 339–
352.
Business Insider. (2014). If you don’t think robots can replace journalists, check out this article
written by a computer. Business Insider Website. http://www.businessinsider.com/narrative-
science-quill-gamechanger-2014-7
Card, M. (1999). Readings in information visualization: Using vision to think. San Francisco:
Morgan Kaufmann.
Carlson, J. R., & Zmud, R. W. (1999). Channel expansion theory and the experiential nature of
media richness perceptions. Academy of Management Journal, 42(2), 153–170.
CITO Research & Narrative Science. (2015). The automated analyst: transforming data into
stories with advanced natural language generation. Narrative Science Whitepaper.
Retrieved March 1, 2015. http://resources.narrativescience.com/h/i/114675300-white-paper-
the-automated-analyst
Cohn, N. (2013). Visual narrative structure. Cognitive Science, 37(3), 413–452.
Daft, R. L., & Lengel, R. H. (1986). Organizational information requirements, media richness
and structural design. Management Science, 32(5), 554–571.
Daft, R. L., Lengel, R. H., & Trevino, L. K. (1987). Message equivocality, media selection,
and manager performance: Implications for information systems. MIS Quarterly, 11(3),
355–366.
Dennis, A. R., Fuller, R. M., & Valacich, J. S. (2008). Media, tasks, and communication pro-
cesses: A theory of media synchronicity. MIS Quarterly, 32(3), 575–600.
Denstadli, J. M., Julsrud, T. E., & Hjorthol, R. J. (2012). Videoconferencing as a mode of com-
munication: A comparative study of the use of videoconferencing and face-to-face meetings.
Journal of Business and Technical Communication, 26(1), 65–91.
Elmqvist, N., Stasko, J., & Tsigas, P. (2008). Datameadow: A visual canvas for analysis of large-
scale multivariate data. Information Visualization, 7(1), 18–33.
Eriksson, L., Byrne, T., Johansson, E., Trygg, J., & Vikström, C. (2013). Multi-and megavariate
data analysis basic principles and applications. Umeå: Umetrics Academy.
Fiske, J. (2010). Introduction to communication studies. Oxfordshire: Routledge.
Forbes. (2015a). Narrative science partner. Forbes Magazine. Retrieved March 1, 2015. http://
www.forbes.com/sites/narrativescience/
Forbes. (2015b). Intel profit expected to slip. Forbes Magazine. http://www.forbes.com/sites/
narrativescience/2015/07/13/intel-profit-expected-to- slip/
Freytag, G. (1896). Freytag’s technique of the drama: An exposition of dramatic composition and
art. Chicago, IL: Scholarly Press.
Gerrig, R. J. (1993). Experiencing narrative worlds: On the psychological activities of reading.
New Haven: Yale University Press.
Gershon, N., & Page, W. (2001). What storytelling can do for information visualization.
Communications of the ACM, 44(8), 31–37.
Green, M. C. (2004). Storytelling in teaching. APS Observer, 17, 4.
Green, M. C., & Brock, T. C. (2002). In the mind’s eye: Transportation-imagery model of nar-
rative persuasion.
Haroz, S., Kosara, R., & Franconeri, S. L. (2015, April). Isotype visualization: Working
memory, performance, and engagement with pictographs. In Proceedings of the 33rd
annual ACM conference on human factors in computing systems (pp. 1191–1200). ACM.
Heer, J., & Bostock, M. (2010, April). Crowdsourcing graphical perception: using mechanical
turk to assess visualization design. In Proceedings of the SIGCHI conference on human
factors in computing systems (pp. 203–212). ACM.
Howson, C. (September 4, 2014). Visual data discovery: 4 Storytelling approaches compared.
Information Week.
Hullman, J., & Diakopoulos, N. (2011). Visualization rhetoric: Framing effects in narrative
visualization. IEEE Transactions on Visualization and Computer Graphics, 17(12), 2231–
2240.
Hullman, J., Drucker, S., Riche, N. H., Lee, B., Fisher, D., & Adar, E. (2013). A deeper under-
standing of sequence in narrative visualization. IEEE Transactions on Visualization and
Computer Graphics, 19(12), 2406–2415.
 Journal of Management Analytics 193

Intel Corporation. (2015). Intel website. Retrieved 1, 2015, http://www.intel.com/content/www/

us/en/homepage.html
Keim, D. A. (2002). Information visualization and visual data mining. IEEE Transactions on
Visualization and Computer Graphics, 8(1), 1–8.
Kock, N. (2004). The psychobiological model: Towards a new theory of computer-mediated
communication based on Darwinian evolution. Organization Science, 15(3), 327–348.
Kwon, B. C., Stoffel, F., Jäckle, D., Lee, B., & Keim, D. (2014). Visjockey: Enriching data
stories through orchestrated interactive visualization. In Poster compendium of the compu-
tation + journalism symposium (Vol. 3).
Kydd, C. T., & Ferry, D. L. (1991, January). Computer supported cooperative work tools and
media richness: An integration of the literature. In Proceedings of the twenty-fourth annual
hawaii international conference on system sciences (pp. 324–332). IEEE.
Larkin, J. H., & Simon, H. A. (1987). Why a diagram is (sometimes) worth ten thousand words.
Cognitive Science, 11(1), 65–100.
Lee, C. E. (2010). Face-to-face versus computer-mediated communication: Exploring employ-
ees’ preference of effective employee communication channel. International Journal for the
Advancement of Science & Arts, 1(2), 38–48.
Lengler, R. (2006, July). Identifying the competencies of’visual literacy’-A prerequisite for
knowledge visualization. In Tenth International Conference on Information Visualisation
(IV’06) (pp. 232–236). IEEE.
Manyika, J., Chui, M., Brown, B., Bughin, J., Dobbs, R., Roxburgh, C., & Byers, A. H. (2011).
Big data: The next frontier for innovation, competition, and productivity.
Mayer, R. E., & Anderson, R. B. (1991). Animations need narrations: An experimental test of a
dual-coding hypothesis. Journal of Educational Psychology, 83(4), 484.
Mayer, R. E., & Moreno, R. (1998). A split-attention effect in multimedia learning: Evidence
for dual processing systems in working memory. Journal of Educational Psychology, 90(2),
312.
Mayer, R. E., & Sims, V. K. (1994). For whom is a picture worth a thousand words? Extensions
of a dual-coding theory of multimedia learning. Journal of Educational Psychology, 86(3),
389.
McAfee, A., Brynjolfsson, E., Davenport, T. H., Patil, D. J., & Barton, D. (2012). Big data: The
management revolution. Harvard Business Review, 90(10), 60–68.
McCloud, S. (1993). Understanding comics: The invisible art. Northampton.
Munzner, T. (2009). A nested model for visualization design and validation. IEEE Transactions
on Visualization and Computer Graphics, 15(6), 921–928.
Munzner, T. (2014). Visualization analysis and design. Natick: AK Peters/CRC Press.
Narrative Science. (2015). Storytelling is the last mile in big data and analytics. Narrative
science website. Retrieved March 1, 2015. http://resources.narrativescience.com/h/i/
124944227-what-is-natural-language-generation
New York Times. (2013). 2013: The year in interactive storytelling. New York Times
Newsgraphics. http://www.nytimes.com/newsgraphics/2013/12/30/year-in-interactive-
storytelling
Nichols, B. (2017). Introduction to documentary. Bloomington: Indiana University Press.
Oatley, K., Green, M. C., Strange, J. J., & Brock, T. C. (2002). Narrative impact: Social and
cognitive foundations. Narrative impact: Social and cognitive foundations.
Reiter, E., & Dale, R. (1997). Building applied natural language generation systems. Natural
Language Engineering, 3(1), 57–87.
Robertson, G., Fernandez, R., Fisher, D., Lee, B., & Stasko, J. (2008). Effectiveness of anima-
tion in trend visualization. IEEE Transactions on Visualization and Computer Graphics, 14
(6), 1325–1332.
SAP. (2015). SAP Lumira: Stories. SAP Help Website. Retrieved March 1, 2015. http://help.sap.
com/saphelp_lumira120/helpdata/en/1b/df5934709d410ea57d530ab7333fdc/content.htm
Seeking Alpha. (2015). Intel earnings preview: Strong quarter expected. seeking alpha website.
http://seekingalpha.com/article/2812135-intel-earnings-preview-strong-quarter- expected
Segel, E., & Heer, J. (2010). Narrative visualization: Telling stories with data. IEEE
Transactions on Visualization and Computer Graphics, 16(6), 1139–1148.
194 J. D. Bradbury and R. E. Guadagno

Shannon, C. E., Weaver, W., & Burks, A. W. (1951). The mathematical theory of communication.
Chicago: UI Press.
Shaw, G., Brown, R., & Bromiley, P. (1998). Strategic stories: How 3M is rewriting business
planning. Harvard Business Review, 76(3), 41–49.
Shneiderman, B. (2003). The eyes have it: A task by data type taxonomy for information visu-
alizations. In The craft of information visualization (pp. 364–371). San Francisco: Morgan
Kaufmann.
Shneiderman, B. (2008, June). Extreme visualization: squeezing a billion records into a million
pixels. In Proceedings of the 2008 ACM SIGMOD international conference on Management
of data (pp. 3–12). ACM.
Slate. (2014). The first news report on the L.A. earthquake was written by a robot. Slate
Magazine. http://www.slate.com/blogs/future_tense/2014/03/17/quakebot_los_angeles_
times_robot_journalist_writes_article_on_la_earthquake.html
Tableau. (2014). 8.2 Preview: storypoints. Tableau. http://www.tableau.com/about/blog/2014/5/
82-preview-tell-story-your-data-story-points- 30761
The Atlantic. (2012). Can the computer at narrative science replace paid writers. The Atlantic
Magazine. http://www.theatlantic.com/entertainment/archive/2012/04/can-the- computers-
at-narrative-science-replace-paid-writers/255631/
The guardian. (2015). datablog. Retrieved March 1, 2015. http://www.theguardian.com/data
Tversky, B., Morrison, J. B., & Betrancourt, M. (2002). Animation: Can it facilitate?
International Journal of Human-Computer Studies, 57(4), 247–262.
Ward, M. O. (2008). Multivariate data glyphs: Principles and practice. In Handbook of data
visualization (pp. 179–198). Berlin: Springer.
Yee, K. P., Fisher, D., Dhamija, R., & Hearst, M. (2001). Animated exploration of dynamic
graphs with radial layout. In IEEE Symposium on Information Visualization, 2001.
INFOVIS 2001. (pp. 43–50). IEEE.
Yi, J. S., Kang, Y., & Stasko, J. (2007). Toward a deeper understanding of the role of interaction
in information visualization. IEEE Transactions on Visualization and Computer Graphics, 13
(6), 1224–1231.
Yseop. (2015). What is smart BI. Yseop website. Retrieved March 1, 2015. http://yseop.com/
EN/smart-business-intelligence-why
Zacks. (2015). Earnings preview. Zacks Website. Retrieved March 01, 2015, http://www.zacks.
com/earnings.
Ziemkiewicz, C., & Kosara, R. (2008). The shaping of information by visual metaphors. IEEE
Transactions on Visualization and Computer Graphics, 14(6), 1269–1276.