Communicating Data for Impact

Peter Speyer
Chief Data and Technology Officer
Institute for Health Metrics and
Evaluation (IHME)
Brian Pagels
Chief Impact Officer
Forum One
Nam-ho Park
Managing Director, West Coast
Forum One
COMMUNICATING
DATAFOR
IMPACT

Communicating Data for Impact 2
Table of Contents
Introduction .................................................................................................. 3
Challenges in Communicating Data ...................................................... 6
Identifying the Right Audience ...................................................................... 7
Understanding Audience Needs .............................................................. 9
Case Study: Making Global Health Data Actionable ......................... 10
Making Data Impactful ............................................................................. 20
About the Authors ..................................................................................... 22

Introduction
The year was 1854. On the Crimean peninsula, British and French
soldiers were engaged with Russia in a battle over religion and
territory. Soldiers were dying in droves.
Across the Black Sea, a young yet accomplished British nurse
volunteered to serve in the military hospitals in Turkey, where
injured soldiers were transported for care. When she arrived, the
sanitary conditions in the hospitals were horrendous—as was the
manner in which the hospitals were collecting data about their
patients.
The young nurse took notice, and improved
both during her tenure. Then she did
something truly remarkable.
In the mid-19th century, data storytelling and
public policy converged. Florence Nightingale
– Britain’s beloved “Lady with the Lamp”
and the young nurse from the story – saved
many more lives with her data analysis and
storytelling than she could have saved alone
as a nurse. Nightingale believed it was her
religious imperative to study statistics. “To
understand God’s thoughts,” she said, “we
must study statistics for these are the measure
of His purpose.”
Following her service in the Crimean War,
Nightingale partnered with the accomplished
statistician William Farr to analyze the hospital
mortality data. The numbers were clear: more
soldiers died from diseases spread through
poor sanitary conditions than from combat wounds. Nightingale
and Farr published these findings in the behemoth 830-page
Notes on Matters Affecting Health, Efficiency, and Hospital
Administration of the British Army. Farr would have been willing
to stop there, but Nightingale was concerned that the dense
report and data tables alone wouldn’t move Queen Victoria or the
British Army to improve sanitary conditions in military hospitals.
Florence Nightingale
Source: Wikipedia

“We do not want impressions, we want facts…You complain that
your report would be dry. The dryer [sic] the better. Statistics
should be the driest [sic] of all reading,” Farr warned her.
Nightingale was aware of emerging data visualization techniques.
Scottish economist and Renaissance man William Playfair
invented the line graph and bar chart in the late 18th century,
and then the pie chart and circle graph in the early 19th century.
Nightingale knew she could leverage these methods to illustrate
her findings.
So she ignored Farr’s advice. Instead, Nightingale created polar
area diagrams – often referred to as Nightingale’s “Rose” or
“Coxcombs” – which used colored wedges to represent causes
of death in the army during each month of the war.
Florence Nightingale, “Diagram of the causes of mortality in the army in the East” was published in Notes on Matters Affecting
the Health, Efficiency, and Hospital Administration of the British Army and sent to Queen Victoria in 1858. Source: Wikipedia

Nightingale said her intention was “to affect thro’ the Eyes what
we fail to convey to the public through their word-proof ears.”
Her report and graphics had a significant impact on military
hospital practices and conditions. By considering her
audiences and selecting the best data visualizations to affect
change, Nightingale became one of the earliest effective data
communicators.
Just like Florence Nightingale, data owners who want to make an
impact must consider their intended audiences and tailor the data
communication strategy accordingly. Today, more data are being
collected, scrubbed, analyzed, and shared than ever before. But
the sheer amount of data creates its share of problems for those
who wish to improve conditions in the world.
What data are available to highlight or help solve societal issues?
Which data and sources should we trust? And once we’ve
identified the right data, how do we effectively communicate
them to those in a position to influence public policy?

“Communicating data and
maximizing impact is about
supplying the right audience with
the right amount of data in the
right format.”
Challenges in Communicating Data
Communicating data and maximizing impact is about supplying
the right audience with the right amount of data in the right
format. And in order to increase the likelihood that a target
audience will pay attention, one must consider the appropriate
timing and channel(s) for delivery of the data.
Nightingale’s and Farr’s 830-page
tome was a massive achievement and
found attentive readers. However, to
drive change, Nightingale targeted very
specific audiences: the Queen and
the leadership of the Royal Army. And
she needed to provide them with the
data in a way that would capture their
attention. Nightingale’s “coxcomb”
visualizations combined those data
in easy-to-understand graphs, and
delivered a clear message.
Any individual or organization providing data to audiences has
the opportunity to decide how to share their data. The larger the
dataset, the more options there are to create impact (and the more
decisions must be made to provide audiences with the right data.).
For some audiences, access to all available data is crucial.
However, this also means users must navigate vast amounts of
information, which puts most of the burden on them. For other
audiences, like news consumers, fewer, punchier numbers
are best. Vast amounts of data can be condensed to one key
illustrative number or visual via aggregation and filtering, and the
onus of picking and creating this one number is on the datapoint
provider.
Those are two extremes. The best ways to provide the data to a
given audience will probably be somewhere in between, ranging
from API access to the full database, sophisticated or basic query
tools, interactive visualizations, reports, infographics, and many
more.

Identifying the Right Audience
To maximize impact, data providers must identify audiences they
need to address, and then provide them with the right amount
and granularity of data.
They must then choose the appropriate data products and tools
for those audiences, which range from complex interactive
visualizations or query tools to simple graphs or illustrations.
These data products can then be packaged and delivered as part
of scientific books, policy reports, press releases, blog posts,
email newsletters, social media outlets, and more.
The table on the following page provides an overview of
key audiences, along with relevant data and tools for these
audiences. For each of the audience “types” below, we selected a
role that best represents that audience’s data-related needs.

Type Description
Audience
examples
Data Products & tools
The
Casual
User
The Data
Actor
The
Analyst
The
Researcher
Interested individuals take
in the data. If the data
trigger action, these
individuals may move into
a more active role. They
typically have little data
and domain knowledge.
Data actors act on and
leverage the data to drive
change. They have
significant domain
knowledge, but often
only limited time.
Other data actors may
also be looking for
interesting stories or data
to back up a story. They
need to review
information quickly, and
often need a specific
figure.
Analysts use data to
create deeper
understanding, while
informing data actors and
consumers. They have a
deep domain knowledge,
extensive data knowledge,
and will review and
condense large amounts
of data for a given topic
into reports,
presentations, apps, etc. .
Researchers work in the
trenches to collect,
analyze, and synthesize
data for the groups
above; they could have
done this data collection
/ analysis themselves.
Anyone with a
casual interest in
the topic.
Policy and decision
makers in
governments,
nonprofits, and
corporations
Journalists,
bloggers, advocates
Domain experts at
int’l, national, local
levels; often staffers
for decisions
makers
Web/software
developers, “data
geeks”
Researchers,
academics, analysts,
modelers
Specific data
points, trends,
developments
Curated
datasets, e.g. by
topic,
country/region
See above
Comprehensive
datasets
showing global
trends, data by
topic, country
etc.
Full database
Full database;
source data &
methods (input
for dataset)
Infographics,
declarative/narrative
visualizations,
illustrative diagrams
Press releases,
reports/briefs,
limited interactive
visualizations, search
data tools
See above
Query tools,
exploratory
visualizations
Application
programming
interface (API)
Query tools,
exploratory
visualizations,
data catalogue,
data repository,
visuals to explain the
methods

Understanding Audience Needs
When releasing data for public consumption, data providers
should be prepared to make a few upfront decisions. One
fundamental question must be asked: “How much data should I
present to the user?”
The above table provides a high-level illustration of the range of
data needs, related products, and desired levels of interactivity,
all based on audience characteristics.
To see this in action, consider a policy
maker interested in health. Perhaps
she’s looking to improve health
outcomes and lower the cost of the
delivery of health care services for
individuals with complex needs. Her
data format and delivery needs likely
differ greatly from those of a web
developer looking to create a map
that highlights differences in health
care delivery costs across the country
(which this policy maker may in fact
find useful).
With the big picture already considered, let’s take a deeper dive
into the audiences highlighted above. We will look at each of
the audience groups in turn, and some overlap between these
audiences will emerge. We’ll also provide concrete examples
from a recently published study of health patterns around the
world, the Global Burden of Disease study.
“One fundamental question must
be asked: “How much data should I
present to the user?”

Health data are collected in many places, including censuses,
surveys, vital registration systems and registries run by governments,
health records, claims data, administrative data, and scientific
literature. To understand the state of health around the world,
we must identify, access, compile, analyze, and evaluate all these
data comprehensively—a seemingly
insurmountable challenge.
In the early 1990s, Professors
Christopher Murray and Alan Lopez
accepted that challenge. Called Global
Burden of Disease (GBD), their approach
measures the impact of diseases,
injuries, and risk factors that shorten
lives or create health loss through
short- or long-term disabilities.
In 2012, the Institute for Health Metrics
and Evaluation (IHME), under the
leadership of Chris Murray, published the results of the GBD 2010
Study —a complete revision done in collaboration with 488 co-
authors from 50 countries around the world.
Based on all health data available to the researchers, GBD 2010
covers 187 countries and provides data for 291 diseases and
injuries, and 67 risk factors. 3 different metrics (YLLs, YLDs and
DALYs) are available as numbers, rates, or percentage, by age (20
age groups) and sex, for 1990, 2005 and 2010, with uncertainty
intervals.
The results dataset has more than 1 billion data points. There
were a number of significant challenges in creating this dataset,
from identifying and accessing data, to cleaning and preparing
data for analysis, to the actual analysis and evaluation. IHME built
out a 3,000-node computer cluster to facilitate the analysis,
and even then, it took about 1 week to run the analytics from
beginning to end.
Three metrics are used to measure health loss:
• Years of Life Lost (YLL) due to premature
mortality
• Years Lived with Disability (YLDs)
• Disability-Adjusted Life Year (DALY)
Global Burden of Disease (GBD)
Case Study

The Casual User
The general—or perhaps interested and informed—public sits
at the top of the table. They don’t necessarily have domain
expertise with an issue, but might be passionate about one or
more topics. If they’re not initially interested, data providers must
pique their interest, or even motivate them to act.
If, say, a member of this audience
wants to donate to the international
relief organization with the best
track record, he’ll be looking for key
data points and trends that give him
answers and better insight into the
problem at hand. These data points
are best delivered as numbers, or as
infographics telling a simple data story.
Selecting one number to focus on
could require tremendous effort from the data provider. Despite
the challenge, and although the user experience isn’t exactly
interactive, boiling a complex issue down to one number can
create a powerful and elegant statement.
If the data story requires more than one number, data providers
and analysts have many options for visualizing their data. These
techniques include: tables, bar charts, tree maps, dashboards,
cartograms, etc.
Infographics may include one or more of these graphical
approaches to represent data, and can be either static or
interactive.
If they’re not initially interested,
data providers must pique their
interest, or even motivate them
to act.

As we’ve discussed, data geared toward general public should
be limited to key data points, trends, or stories. While some
visualizations aimed at data actors may interest the general
public, IHME created a higher-level visualization with key
summaries per
country, called GBD
Insight. And a printed
infographic, the GBD
2010 Poster, appeared
in Lancet, providing
some key metrics, a
country ranking by
life expectancy, and
more details on a few
example countries.
GBD 2010 Poster.
Source: IHME
Case Study
GBD Results for Data Consumers

The Data Actor
This is a crucial audience that includes legislators, ministers,
nonprofit leaders, business executives, etc. Since they’re in the
position of influencing and often creating/refining policy, their
data needs are more complex than those of the general public.
However, policy makers are not likely to spend hours wading
through reams of data to find what they need. Instead, they’ll
look for packaged briefs with corresponding figures they can use
in a speech, or maybe a curated dataset or visualization that only
requires selecting a few filters or settings to get the information
they’re looking for.
In recent years, innovators - inspired
by the visual storytelling of leading
online news media - have made the
traditional paper or PDF briefs more
interactive and accessible. Modern
briefs will include text, pictures, videos,
statistics or interactive graphs, and
other contacts. The objective, however,
remains the same: provide the most
relevant data in a way that makes best
use of a policy or decision makers’ time.
Most journalists, advocates, and bloggers often need data at
about the same level as policy makers, but are more likely to use
an interactive tool to find a unique data story to highlight.
Providers might approach interactivity in a few different ways. For
instance, interactive visualizations allow users to review a specific
story and further drill into the dataset. Alternatively, data query
tools can allow users to find individual data points or create small
tables. They can provide access to the data with curated filters
or guides to point users in the right direction and allow them to
easily retrieve relevant data.
Policy makers are not likely to
spend hours wading through
reams of data to find what they
need.

GBD Results for Data Actors
IHME has created several tools that allow exploration of data
at a higher, global level to give decision makers access to the
specific data-supported conclusions they need to affect change.
Most notably, GBD Cause Patterns shows a condensed view of
21 broad cause groups, and allows review of patterns by age,
geography, sex, and chronology. GBD Arrow Diagram allows
a quick comparison of disease and risk factor rankings in 1990
and 2010. It helps users look at the big picture and identify key
priorities, based on size or growth of burden. Users can then drill
down by country, age, and sex to see what populations are most
affected. More examples of visualizations can be found on the
GBD Visualizations page.
In addition, a query tool
makes it easy for users
to find individual data
points or small tables
with the information
they choose. The
information is then
displayed as a simple
visual and made
available for download.
But people still
appreciate paper. In
addition to the online
tools, IHME created
a GBD Policy Report
that was provided
in printed form and
for download online.
Given the tremendous
demand for this kind
of information, the report was followed up
with regional editions in collaboration with
the World Bank, as well as a report for EU
and EFTA in collaboration with the European
Union.
GDP Arrow Diagram.
Source: IHME
Case Study

The Analyst
Staff who work for data actors at NGOs, in government, or for
corporations are often domain experts in a particular field. Their
work requires data analysis in order to plan or evaluate programs
and policies.
Analysts need access to filtered
datasets for the topic or geographic
region in which they operate. Some
of this group may wish to export a
portion of a dataset to run their own
analyses, while others are content to
use interactive tools provided through
the web interface.
Web or software developers and
“data geeks” also need full access to
the database, but they often seek Application Programming
Interfaces (APIs) to more easily integrate data into existing
applications, or to create entirely new applications with the data.
APIs allow developers to tap into a database and use it to create
their own applications, or combine datasets via multiple APIs,
while ensuring the applications are refreshed dynamically when
the source is updated. APIs also free them from the burden and
cost of storing large datasets.
Developers sometimes serve as data promoters of sorts—their
products or services may reach any of the above audiences.
Analysts need access to filtered
datasets for the topic or
geographic region in which they
operate.

Experts and analysts understand global health, specific diseases,
injuries, and risk factors. They also understand the metrics used
to measure them.
Their real interest lies in reviewing the data to find patterns and
trends, and to answer questions. Then, they can use the data to
plan interventions and programs. Just like researchers, they want
lots of detail, but also more intuitive ways to interact with the data.
To satisfy this level of need, IHME built a exploratory data
visualization for data experts and data analysts. Called GBD
Compare, it’s IHME’s flagship tool for interaction with GBD data.
The visualization allows users to review the data in 5 different
chart types: treemaps, maps, age and time plots, and stacked bar
charts.
There’s a 2-panel view with linkage between panels, e.g.
clicking on country in the map will set the other chart to that
country. Some charts
have many different
view options; for
instance, a treemap
illustrating diseases
and injuries can
filter by trend data,
uncertainty bounds, or
risk factor attribution.
The visuals are
complex and require
the user to have some
prior knowledge of
burden of disease
measurement. They’re
powerful for exploring
trends, from high-level
to great detail. Users
can export screenshots as JPGs; they can download the
underlying data as a CSV; or they can share a permalink
with the specific settings they’re using via email or social
media.
GBD Compare.
Source: IHME
GBD Results for Data Analysts
Case Study

The Researcher
Researchers and academics almost always want full access to a
database in order to analyze the data and create models for their
research. They’ll often look for the “download” option before
spending time exploring an interactive
tool. Providing downloadable files
in machine readable format for easy
importing into statistical analysis tools
is a useful mechcanism here.
Researchers may also require
background information on underlying
data and methodology, including
survey instruments, published papers,
references to underlying data used in
analysis, and similar documentation.
The researcher will often look for
the “download” option before
spending time exploring an
interactive tool.

GBD Results for Researchers
Researchers and other data experts want to review and use the
results at the most comprehensive level. They’re comfortable
working with databases, and would rather manipulate the data
themselves. In addition, they need to understand the methods
and data used to generate the results. Fully detailed accounts of
the GBD methods are available across 8 papers and appendices
(together far more than 1000 pages), accessible on the Lancet
website.
To provide researchers with information about the more than
20,000 datasets used for GBD, the IHME team catalogued the
data in the Global Health Data Exchange (GHDx) with a link to
the data provider, as well as extensive metadata like geography,
years, summary,
keywords, publisher,
and more.
Full results of GBD
2010 are available
for download in CSV
files in the GHDx.
Given the size of
the results database,
the data have to
be downloaded by
disease, injury, risk
factor, or country.
IHME created
the Mortality
Visualization, which
illustrates the starting
point for these results.
Here, users can review trends and find citations
and other metadata for all data points. See also
the COD Visualization for all data points used for
estimating causes of death.
Mortality Visualization.
Source: IHME
Case Study

Communicating Data: Cheat Sheet
The table below maps our audience types from above to their
attributes as they pertain to data, along with the extent to which
these attributes are present in the audiences. As an example,
data consumers represent a large audiences size with limited
attention span and rudimentary data manipulation skills and
domain expertise. Data experts on the other hand represent
a small audience, but they are willing to engage with the data
(attention span) and have in-depth data skills and topic expertise.
Audience size Attention
Data
manipulation
skills
Topic
expertise
The
interested
individual
The data
actor
The
analyst
The
researcher
Large
Small
Medium
Small
Low
Low
High
High
Low
Medium
High
High
Low
Medium
High
High
Example
Exploratory visualization with all data available
Infographic
Narrative
visualization /
brieﬁng
Detailed
topical
visualization
Exploratory
visualization
with all data
available
Ability to
drive change
Low
High
Medium
Medium

Conclusion: Making Data Impactful
Today, data are being collected at an unprecedented scale,
and they exist in many different shapes and formats. The tools
for data analysis are becoming ever more sophisticated, and
there are more and more powerful tools to present data and
engage users. Data are being looked to as an important aspect
of communication, as input for decision making, and as a means
to evaluate the effectiveness and performance
of organizations, programs, projects, policies,
etc. But many people and organizations
struggle with unlocking the power of data and
communicating it in the most effective way.
Ultimately, data providers want to inspire action
among their audiences. They understand we
need to use these data to solve our current set
of problems, and increasing the distribution and
understanding of data can accelerate the pace
of positive change.
Depending on their domain expertise and
data skills, individuals in organizations and
governments require data in relevant formats
to inform these decisions. Despite more
data, better tools and more data skills, the
fundamental challenge for communicating
data has remained the same since Florence
Nightingale’s time: supplying the right audience
with the right amount of data in the right format.
Data curation and decisions about the data delivery mechanism
are as important as the data themselves.
Careful curation of the data tailored to the audience it serves
determines how that data is purposed for positive social change,
whether it be saving lives in hospitals, making informed choices
for the health of your family, or crafting public policy that will
affect large populations. How we shed light on the data does
influence its impact.
Florence Nightingale.
Source: Wikipedia

GBD Study Conclusion
The GBD Study provides the most comprehensive dataset ever
compiled on global health outcomes. Experts have likened the
study to the Human Genome project in its importance, and
the response to the publication of the GBD Study has been
overwhelming. Analysts have worked with or requested custom
datasets. Domain experts have engaged with and commented
on data reviewed in the visualization tools. Policy makers have
requested conversations and workshops to interpret the data.
Journalists have used the data for health coverage and for
sparking Twitter conversations about specific data points; Wired
Magazine turned one of the visuals into a beautifully rendered
infographic.
Part of the reason for this response was that IHME took a
hard look at the GBD results, evaluated the needs of different
audiences, and created a suite of tools to share the results of
the study. Analyzing the world’s health data to generate the GBD
results was a monumental effort, but the impact of the study
wouldn’t have been the same without tools that made the data
easy to access and explore for different audiences.
Case Study

About the Authors
Peter Speyer
Chief Data and Technology Officer
Institute for Health Metrics and Evaluation (IHME)
Peter and his team fuel IHME’s research with data from
organizations around the world, provide state-of-the-
art data management and computational infrastructure,
and share IHME’s research results with all the different
audiences mentioned in this paper.
speyer@uw.edu
Brian Pagels
Chief Impact Officer,
Forum One
Brian helps foundations, government, and nonprofits
explore opportunities to manage, visualize, and share
data more effectively. He is particularly passionate about
domestic health and environmental issues, and has
managed data-focused projects for the Robert Wood
Johnson Foundation, the AARP, and the EPA, among
others.
bpagels@forumone.com
Nam-ho Park
Managing Director, West Coast
Forum One
Nam-ho works in Seattle with organizations focused on
global health and development, furthering their missions
and influence through digital communications strategy
and web technologies.
npark@forumone.com

Special Thanks
Reviewers
Kevin Merritt, President & CEO, Socrata
Noah Illinsky, Visualization & UX Expert, ComplexDiagrams.com
William Heisel, Director of Communications, Institute for Health
Metrics and Evaluation
Kurt Voelker, Chief Technology Officer, Forum One
Jim Cashel, Chairman, Forum One
Editor
Kristina Bjoran, Forum One
Designer
Risa Willmeth, Forum One
This white paper may be used and distributed under the following
Creative Commons license.
Attribution-NonCommercial-NoDerivatives 4.0 International

Communicating Data for Impact

More Related Content

What's hot (20)

Similar to Communicating Data for Impact (20)

More from Peter Speyer (9)

Communicating Data for Impact