Storytelling &

Visualization With Data

Casandra Campbell

IEOR E4572
April 4, 11, & 18, 2025
9-5pm
Course 1
Data Visualization
How we perceive visual

Overview
information and how to
chart data effectively.

2
Data Storytelling
How to tell a compelling
story with data that drive
action.

3
Dashboard Design
How to design effective
dashboards and use data
visualization technology.
Table of Contents
Days 1-2 Recap
How to Critically Assess Your Work
What Is a Good Dashboard
Collecting Dashboard Requirements
Dashboard Fundamentals
Dashboard Design Principles
💻Day 3 Activity 1
Common Mistakes in Dashboard Design
💻 Day 3 Activity 2
Introduction to Data Visualization Tools
Tableau Primer
Day 3 In-Class Assignment
Days 1–2 Recap
The data dictates which type of chart we should use.

Chart Type Best For

Simple Text When you only have a number or two to share.

Table When precise values are required.

Line Continuous data and visualizing trends.

Slope Categorical change in values and outliers.

Bar Comparing discrete groups on the same measure.

Area Sub categorical trends.

Scatterplot Showing correlation relationships.

Table Heatmap Making tables easier to interpret.

Food charts should usually be avoided.
Pie charts, doughnut charts, and spaghetti charts are more
likely to get you into trouble than help.
Gestalt principles of visual perception are used to design
more effective visualizations.
Charts should always decrease cognitive load as
much by removing all noise.
● Everything you add to a chart increases cognitive load.
● If it’s not helping to create signal, it’s noise and it shouldn’t
be there.
● When there’s too much to look at viewers tune out and
miss what’s important.
Remove noise by decluttering charts.
Remove Non-Critical Data
● Not all data is equally important. Remove anything
non-critical.
● Ask: Would removing this element remove meaning? If the
answer is no, remove it.

Reduce Importance of Necessary Non-Critical Data

● Push necessary but non-critical information to the
background (e.g., use lighter colors).
● When detail is not needed, summarize instead.
The other lines are needed to contextualize the South Korea line but
they are less important and fading them helps it stand out.
Performance Metrics

● Are measurable.
● Have a target (could be a range or minimum
threshold).
● Drive action.
● Can be influenced.
● Can be used to prove/disprove a hypothesis.
● Will advance objectives if achieved.

Your work is meaningless with the wrong metrics!

 Key Signs of Vanity Metrics 🔑

● It’s not related to an agreed upon goal.

● It’s difficult to reproduce
● It’s difficult to influence.
● It’s often overly simplistic
● It lacks nuance and context.
● It’s misleading.
● It can be easily manipulated.
● It doesn’t help you improve in any way.

You need context to spot these signs!

Decisions get made with data at three different
levels of abstraction.
1. Discrete
2. Operational
3. Strategic

Understanding the goal of an analysis will help you

understand which level you’re working at, and what the
scope of the task is.
 Know Your Audience

Peers Managers Executives

Speak shorthand. Prove your point. Get to the point.

Your message gets more and more refined as you move through these audiences.
A Strong Point of View:

● Is specific and data-informed.

● Describes what action needs to be taken.
● Describes what outcome should be expected.
● Explain the stakes (costs and opportunity costs).
● Surfaces underlying assumptions.
● Explores and considers alternatives.
● Includes the final statistical outcome you
● Can be summarized into one sentence.
Story Structure
The dramatic arc of a story refers to the rise and fall of tension.

Drop
n
nsio
te

ping
s i ng
Ri

tens
io
n
Setup Conflict Resolution
Executive Summaries
When you put the beginning, middle, and end of your data story
together, you have an executive summary!

Our goal was to double the growth of Product X within six months but
only 3% of the sales team has downloaded the material from the portal
in the first two months. Modifying the sales compensation structure to
incentivize Product X sales will increase downloads of the material so that
the sales team can promote the product more effectively.
CHART titles follow a simple format.

chart title = noun you measured + when you measured it

Examples
● 2019 Monthly Profit by Percentage
● Share of U.S. Jobs Requiring AI Skills
● Sales History in Millions (2013–2018)

Note: How you measured it (units) is usually on the y-axis.

Data Story
The slides make up the bulk of your presentation.
● Your POV (setup, conflict, and resolution) and the sub-actions
required for a solution.
● Each subaction should have its own slide with what, why, how.
● Slides should be high signal but may have more detail than the
executive summary.
● Always include sources at the bottom of the slide or in the notes.

An ideal slide in this section should have a headline with the key
insight, a visualization to support it, and a recommendation.
Interest in the Four Seasons doubles when a new season of White Lotus airs.
The Andaz should attempt to secure one of it’s locations as the Season 4
venue to grow brand awareness.

Source: Residual Thoughts

Every slide should have one point or goal.
Write the headline first.
● This will help you clarify what
you want to say and need to
show.
● Everything in the slide
should be in service of the
headline.
Context is how we know how to interpret data.

● What does this mean?

● Is this normal?
● Is this good?
● Is this bad?
● How should I feel about this?

To answer all these questions, we need context!

Comparisons are crucial for context.

● Between time periods

● Between segments
● Against averages
● Against targets
● Against benchmarks
● Between related metrics
Annotating Charts With Context

Use text or callout boxes to add explanations.

Use arrows, colors, magnification and visual callouts to

point things out and highlight what’s important.

Add lines or big numbers for targets and benchmarks.

Add visual for initiatives that have impacted the data.

Data label annotation
Shading and callout box annotation
Circle and text annotation
Simple number annotation
Arrow and color annotation
How to Critically Assess
Your Work
It’s crucial to be able to critically assess your own work.

● Fresh eyes and peer review are ideal but not always
possible.
● Access to fresh eyes and peer review is not a replacement
for quality work.
● The better you are at assessing your own work, the better
you will be at providing fresh eyes and peer review for
others.
Avinash Kaushik’s Quality Assessment Algorithm
1. Time to the most important insight.
2. Effort to understand the whole graphic.
3. Contains trust marks.
4. Rank ordering of key messages.
5. Explanation of the logic powering the graphic.
6. Exposes nuance.
7. Visualizer isn’t trying to be too clever.
8. Likely to recommend to influential leaders.
Quality Assessment Scorecard
3 2 1

Perfect Could be Reboot

optimized

Time to the most important insight.

Effort to understand the whole infographic.

Contains trust marks.

Rank ordering of key messages.

Explanation of the logic powering the graph.

Exposes nuance.

Visualizer isn’t trying to be too clever.

Likely to recommend to influential leaders.

Subtotals 0

Score /40 ( %)
XKCD: COVID Risk Chart 2 1 0

Perfect Could be Reboot

optimized

Time to the most important insight. 1

Effort to understand the whole infographic. 1

Contains trust marks. 0

Rank ordering of key messages. 1

Explanation of the logic powering the graph. 0

Exposes nuance. 2

Visualizer isn’t trying to be too clever. 1

Likely to recommend to influential leaders. 2

Subtotals 4 4 0

Score 8/16 (50%)

Information is Beautiful: Coronavirus Riskiest 2 1 0
Activities
Perfect Could be Reboot
optimized

Time to the most important insight. 1

Effort to understand the whole infographic. 1

Contains trust marks. 0

Rank ordering of key messages. 1

Explanation of the logic powering the graph. 1

Exposes nuance. 1

Visualizer isn’t trying to be too clever. 0

Likely to recommend to influential leaders. 1

Subtotals 0 6 0

Score 6/16 ( 38%)

COVID-19 Recovery Consulting: COVID-19 Risk 2 1 0
Index
Perfect Could be Reboot
optimized

Time to the most important insight. 2

Effort to understand the whole infographic. 1

Contains trust marks. 2

Rank ordering of key messages. 1

Explanation of the logic powering the graph. 2

Exposes nuance. 1

Visualizer isn’t trying to be too clever. 2

Likely to recommend to influential leaders. 2

Subtotals 10 2 0

Score 13/16 ( 81%)

Texas Medical Association: Know Your Risk 2 1 0
During COVID-19
Perfect Could be Reboot
optimized

Time to the most important insight. 2

Effort to understand the whole infographic. 2

Contains trust marks. 1

Rank ordering of key messages. 2

Explanation of the logic powering the graph. 2

Exposes nuance. 2

Visualizer isn’t trying to be too clever. 2

Likely to recommend to influential leaders. 2

Subtotals 0

Score 15/16 (94 %)

What Is a Good Dashboard
What is a dashboard?

“A dashboard is a visual display of the most

important information needed to achieve one or
more objectives, consolidated and arranged on a
single screen so he information can be
monitored at a glance.”

- Stephen Few, Information Dashboard Design

Dashboards are visual displays.
● Information on a dashboard is presented visually.
● Combination of text and graphics, with emphasis on
graphics.
● Graphical representations of data can communicate
more efficiently and comprehensively than text
alone.
Dashboards display the information needed to
achieve specific objectives
● Achieving an objective usually requires access to a
collection of facts that not otherwise related.
● Information needed often comes from diverse
sources.
● It isn’t a specific type of information, just whatever
information is needed to do a particular job.
Anyone trying to complete an outcome uses
dashboards.

● A driver operating a vehicle

● An executive making strategic decisions for a
huge corporation
● A manager running the daily operations of a
team
● An individual contributor completing tasks.
Dashboards are used to monitor information at a
glance.
● A dashboard must be able to quickly point out that
something deserves attention and might require
action.
● It doesn’t have to problem all the information needed
to take action but it should make getting that
information as seamless as possible.
○ This might involve navigating to a different
through linking or drilling down.
Dashboards present information using small,
concise, direct, and clear display media.
● Function over form.
● Information should be displayed efficiently using
easiest to understand charts.
Dashboards are customized.
● Information should be tailored to the requirements
of a person, group, of function who will use the
dashboard.
A dashboard is not…
● A display that is used for data exploration and
analysis.
● A portal.
● A scorecard.
● A report that people use to look up specific facts.
Dashboards help users maintain situation
awareness.
Being aware of what’s happening and understanding what
that information means now and into the future.
○ Level 1: Perception of the elements in the
environment.
○ Level 2: Comprehension of the current situation.
○ Level 3: Project of future state.

Pilots don’t wait for an alarm to go off before looking at their

cockpit displays!
Information in and of itself is useless.
Businesses use information to maintain awareness of
performance and keep performance on track or improve it.

Any form of information display that doesn’t help with

performance monitoring is useless.
Dashboards should be designed to support the
performance monitoring process.
1. Update high-level situation awareness.
2. Identify and focus on particular items that need attention.
a. Update awareness of this item in greater detail.
b. Determine whether action is required.
3. If action is required, access traditional information that is
needed, if any, to determine an appropriate response.
4. Respond.
Information that cannot support action to maintain or
improve performance does not belong on a dashboard.
Effective dashboards are about communication.

If you have not mastered data visualization and

storytelling, you cannot make a great dashboard.
Collecting Dashboard
Requirements
Assessing what’s needed in a
dashboard is requires knowledge.

Effective dashboard designers:

● Understand the relevant
domain.
○ Sales, post-surgical infection
prevention, airplane
manufacturing, classroom
teaching, etc
● Know how to get into the
heads of those who will use the
dashboard.
How to Collect Dashboard Requirements
1. Clearly define the dashboards purpose.
2. Focus on the Goals Not the Means
3. Get into stakeholders heads.
4. Ask the Right Questions
5. Identify What Really Matters
1. Clearly define the dashboards purpose.
● People’s definitions of what a dashboard is vary widely.
● Create alignment on how it will be used and what success looks
like.
2. Focus on the goals, not the means.
● Most people don’t know how a well-designed dashboard looks and
functions.
● Keep discussions focused on what your stakeholders want to
accomplish, not how it will look.
3. Get into stakeholders heads.

✔ Understand how they think about activities the dashboard

will help them monitor.

✔ Use a whiteboard to explore their mental models.

✔ Determine how much complexity they can rapidly

understand.
A stakeholders understanding of
their work exists as mental
models.

● Everything we learn from

making a sandwich to
calculating the trajectory of a
rocket is represented in our
brain as mental models.
● The better we understand
something, the more detailed
our mental model.
Dashboards must closely match mental models to be effective.
Use a whiteboard to understand stakeholders mental models.

● Use simple drawings to sketch concepts.

○ Different shapes like circles and squares can be used to
represent different parts of the process.
○ Lines (with or without arrows) can be used to connect parts
of that are related.
● As the model is emerging ask questions.
○ How does this relate to that?
○ If a problem occurs here, what does it affect and how?
● It may take several revisions to refine the model.
Novices may not have strong
mental models.

● It takes experience to develop

strong mental models.
● Help them think through the
parts and relationships of the
processes that need
monitoring to develop a simple
model.
● This will help them deepen
their expertise.
Experts may not be able to
communicate their mental
models.

● Model becomes deeply

ingrained, automatic, and
intuitive.
● Ask them to teach you what
they do in simple terms to help
coax information to the
surface.
● Pull each part of the model
into focus one by one.
Determine how they process complexity.

● People vary significantly in how they process complexity.

● People benefit most from dashboards that are
well-matched to their abilities.
● Experts with complex mental models may process a lot of
information quickly if it’s presented well.
● Novices may process much less information at one time
and need simpler dashboards.
4. Ask the right questions

● Use questions to guide your conversation.

● Don’t be afraid to ask as many questions as you need.
● People appreciate being heard.
● You will different questions each time you build a
dashboard but some are consistently useful.
● Ask follow up questions based on the answers you get.
● New questions will emerge during the dashboard design
process itself.
Useful Questions to Inform the Dashboard Design Process
● How frequently should the information be updated?
● Who will use the dashboard? Is it for a single person, a single group, or people in
several different departments?
● What will the dashboard be used to monitor, and what objectives will it support?
● What questions should the dashboard answer? What actions will be taken in
response to these answers?
● What specific items of information should be displayed on the dashboard? What
does each of these items tell you, and why is that important? At what level of
summary or detail should the information be expressed to provide the quick
overview that’s needed?
● Which of these items of information are most important for achieving your
objectives?
● What are the logical groupings that could be used to organize items of information
on the dashboard? In which of these groups does each item belong?
● What are the useful comparisons that will allow you to see these items of
information in meaningful context? For instance, if one of the measures that your
dashboard displays is revenue, do you have targets or historical data that could also
be displayed to make current revenue more meaningful?
5. Identify what really matters.

● A dashboard should only include information that actually

influences performance.
● People are often used to dashboards that include
information they never use but are afraid to give up.
● Useless information steals attention from the data that
actually matters every time the dashboard is used.
Ask questions to filter out useless information:

● “Describe a situation when this information would lead

you to do something.”
● Give me an example of the actual data that would appear
and the action that you would take in response.”

If no examples come to mind, it doesn’t need to be included.

Dashboards create maintenance debt! It’s important
to make sure they will actually get used.
Dashboard Fundamentals
How dashboards will be used affect how they should be designed.
Feature Items

Update Frequency Daily

Hourly
Real time

User Expertise Novice

Journeyperson
Expert

Audience Size One person

Multiple people with the same requirements
Multiple people who need to monitor different data subsets

Technology Platform Desktop/laptop

Web server/browser
Mobile device

Screen Type Extra-large screen

Standard screen
Small screen
Variable screens

Data Type Quantitative

Non-quantitative
Update Frequency

● If a dashboard is updated more frequently than daily it

should include the time it was last refreshed.
● Real-time dashboards need to be simpler and contain less
information because the need for constantly updating
situational awareness implies the need for speed.
Frequently Updated Dashboard
Real-Time Dashboard
User Expertise

● Know your audience.

● What is displayed and how it is displayed should be
influenced by the people who will be using it.
● More experience people in a role will be able to handle
more complexity than less experienced people.
● Individuals vary!
Audience Size

One person:
● A dashboard for one individual can be highly customized to their
needs.
Multiple people with the same requirements:
● A group of people all looking at the same data are still different.
● You either have to design multiple versions of a dashboard or
make compromises so that it works better for some than others.
Multiple people who need to monitor data subsets:
● E.g., sales managers responsible for different regions.
● You can make multiple views but they add maintenance debt, or
add filters but they add friction. Ideally, your tech save settings
for each individual.
Technology Platform

● Dashboards that run on individual computers are often

fastest and will generally look consistent from device to
device.
● Browser based dashboards may have performance
limitations and vary based on which browser is being used
to access the dashboard.
● Mobile dashboards will need to be manipulated using
touchscreen.
Screen Type

● Screen size, distance of viewer from screen, aspect ratio,

resolution all affect design.
● Often designing for multiple screens.
● Smaller screens can’t accommodate as much information,
but neither can larger screens viewed from further away.
Data Type

● Dashboards primarily displayed quantitative data but may

also display qualitative data.
● Quantitative data must consider:
○ Variations in timeframe
○ Enrichment through comparisons
○ Anticipation through prediction
○ Enrichment through evaluation
○ Enrichment through other qualitative data.
Timeframe should be dictated by how the dashboard will be used.

● If the user needs to respond immediately when things go

wrong, they will need real-time data.

● If they need to respond to major changes yesterday, they

will need daily updates.

● If patterns are over longer periods of time are more

meaningful, they may need week-to-date or
month-to-date summaries.
 Comparisons are crucial for adding context that make measures useful.
Comparative Measure Example

The same measure at the same point in time in the past The same day last year

The same measure at some other point in time The end of last year

The current target for the measure A budgeted amount for the current period

Relationship to a future target Percentage of this year’s budget so far

A prior prediction of the measure Forecast of where we expected to be today

Relationship to a future prediction of the measure Percentage of this quarter’s forecast

Some expression of the norm for this measure Average, typical range, or a benchmark, such as the number of
days it usually takes to ship an order

An extrapolation of the current measure in the form of a probable Projection into the future, such as to the coming year end
future, either at a specific point in the future or as a time series

Someone else’s values for the same measure A competitor’s measure, such as revenues

A separate but related measure Order count compared to order revenue

Comparisons can be expressed graphically or using text alone.

Year over year sales Week over week sales

comparison is comparisons are
expressed as text. expressed graphically
Comparisons can be made between measures through segmenting.

● Categorical segmentation of sales

might include regions, sales
channels, product categories, or a
frequency distribution of order
value ranges.

● Temporal segmentation might

include daily, weekly, or monthly
sales.
Ask questions to confirm that the right context is being provided.

● In looking at this information, can you tell me how what

what you’re monitoring is performing?
○ Do you see anything abnormal?
○ Are there relative magnitudes of these measures (for
example, sales revenues among several regions) what
you would expect?
○ Can you describe what might be different that would
cause you to think that something’s abnormal or
wrong?
○ What is it that tells you that it’s good (or bad)? How
good (or bad) is it? Is it getting better or worse?
The best dashboards help stakeholders predict what will happen
based on what has already happened.

● The past is more informative when it can be used to

predict the future.
● Optimal performance management prevents bad things
from happening and improves opportunities for good
things to happen.
● Incorporating predictive models into dashboards can help
stakeholders make better decisions about the future.
The best dashboards help stakeholders predict what will happen
based on what has already happened.
Explicitly declaring if something is good or bad can update
situational awareness faster.
● Qualitative evaluation can be encoded
visually such as objects (e.g., a traffic
light) or attributes (e.g., displaying
measures shades of green or red).
● These visual indicators can effectively
alert users to the stake of a measure
without altering the overall design of a
dashboard.
● Must be done based on defined
criteria that stakeholders are aligned
on.
Other qualitative data such as simple lists can be used to enrich
quantitative data.

● Top 10 customers
● Issues that need to be investigated.
● People who need to be contacted.
● Etc.
Annotated context can also be used to enrich quantitative data.

● Explanations of previous investigations.

● Changes to the data source.
● Campaigns that launched
● Etc.
Dashboard Design
Principles
Dashboard Design Principles

● Everything you’ve learned so far about data visualization

and storytelling applies to dashboards.
● Dashboards also have some specific design principles to
unique to their needs.
8 Dashboard Design Principles

1. Organize information to support its meaning and use.

2. Maintain consistency to enable quick and accurate
interpretation.
3. Put supplementary information within reach.
4. Make the experience aesthetically pleasing.
5. Expose lower-level conditions.
6. Prevent excessive alerts.
7. Keep viewings in the loop.
8. Accommodate real-time monitoring.
 1. Organize information to support its meaning and use.

● Information shouldn’t just be added to a dashboard

randomly.
○ Organize groups according to activities, entities, and
use.
○ Co-locate items that belong to the same group.
○ Delineate groups using the least-visible means.
○ Support meaningful comparisons.
○ Discourage meaningless comparisons.

1
 Organize groups according to activities, entities, and use.

● Form groups that align with activities within the organization.

○ Order entry, shipping, budget planning, etc.
● Form groups the way the viewer use the information.
○ E.g., Comparing revenue and expenses to determine
profitability.
● Ordering information to support efficient scanning.
○ E.g., Each step in the marketing funnel.

Remember differ users will have different needs and use dashboards
differently.

1
 This group of charts follows a natural sequence from sending the email,
to opening it, to clicking on it, to unsubscribing.

1
 Co-locate items that belong to the same group.

● Use positioning to group related items visually.

● Delineation from other groups should be subtle.

1
 Delineate groups using the least-visible means.

● Borders and background fill

colors qualify as non-data pixels
and should be minimized as
much as possible.
● Blank space is the optimal way to
delineate groups when possible.
● Because dashboards are often
high defensive, there isn’t always
enough space to spare subtle
borders can be used.

1
 Support meaningful comparisons.

● Comparisons are crucial for providing context that adds meaning

to measures.
○ Combining items in a single table or graph.
○ Placing items close to one another.
○ Linking items in different groups using a common color.
○ Including comparative values (e.g., ratios, percentages, or
variance) whenever useful for clarity and efficiency.

1
 This dashboard uses small multiples to support meaningful comparisons.

1
 This dashboard co-locates sales and profit and provides year over year
variance figures to support meaningful comparisons.

1
 Discourage meaningless comparisons.

● Not everything on a dashboard should be compared or

linked to everything else.
● Inadvertently linking data can cause inappropriate
comparisons.
○ Separate items especially when they’re not related.
○ Don’t use like colors for different categories.

1
 Inconsistent use of color creates meaningless comparisons on this
dashboard.

1
2. Maintain consistency to enable quick and accurate
interpretation.

● We search for the significance of differences in appearance,

whether consciously or unconsciously.
● Thing that mean the same thing or function the same way should
look the same wherever they appear on the dashboard.
● Even something as subtle as using a darker axis on one chart can
cause the view to perceive the difference as meaningful.
● Charts that are intended for similar use should use the same
chart type.
● Nothing in a dashboard should change from day to day besides
the data itself.

2
This dashboard maintains consistent colors for each energy type.

2
This dashboard uses bar charts consistently and doesn’t introduce
variety for variety's sake.

2
3. Put supplementary information within reach.

● Dashboards can never provide all the information needed

to performance a job.
● They provide the overview but need to be supplemented
with additional information for a more comprehensive
understanding.
● Supplementary information should be easily accessible:
○ Pop-up windows invoked by hovering
○ Temporarily altered states invoked by a sustained
click.
○ Through access to a separate screen when more more
than a little information is needed.

3
Pop-Up Windows or Tooltips

● Pop-up windows, also known as tooltips, appear when the

mouse hovers over an area with supplemental
information.
● These work best for small amounts of information such as
one chart.
● Nothing about the dashboard itself changes, but
supplemental information is temporarily overlaid.

3
This dashboard uses a tooltip to drill down into subcategories.

3
This dashboard uses a tooltip to provide more granularity.

3
Altered States

● Parameters can be applied to alter the data across a

dashboard (e.g., all metrics filtered for one region).
● Altered states can also be invoked using a sustained click.
When the click is released the altered state reverts so
there’s never confusion about what data is being viewed.
● This should only done for supplemental views that need to
be accessed frequently.

3
The same data is displayed as absolute volumes or percentages.

3
Independent Screens

● When something on the dashboard requires a response,

additional information is usually needed.
● Additional is best displayed on a separate screen that can
be easily accessed from the original dashboard.

3
Clicking on a product exposes the customer list on this
dashboard.

3
The first screen on this dashboard provides of overview of location
data, while clicking on one of the locations opens a second screen
with more detail.

3
4. Make the experience aesthetically pleasing.

● An aesthetically unpleasing experience undermines the

dashboard’s ability to communicate.
● Aesthetic thinking should applied directly the information
being displayed, not to meaningless and distracting
ornamentation.
● Aim for simple elegance.
○ Choose appropriate and meaningful colors.
○ Use high resolution text and images.
○ Align content whenever appropriate.
○ Use a legible font.

4
Choose appropriate and meaningful colors.

● Poor use of color is the most common aesthetic issue.

● Bright or dark colors visually demand attention and fatigue
the viewer.
○ Keep bright colors to a minimum, using them only to
highlight information that requires attention.
○ Except for content that demand attention, use less
saturated colors such as those that are predominant
in nature.
● Stick with a consistent palette and use your brand color
palette when appropriate.

4
This dashboard has several issues related to color.

● Inconsistent used of color.

● Excessive use of bright and
highly saturated colors.
● Unflattering color palette.
● Inappropriate background
colors.

4
This dashboard uses color effectively

● Consistent neutral color

use. Darker color used to
draw attention
appropriately.
● Colors aren’t too bright.
● Red and green are used to
encode context
● Charts are easy to read on
light backgrounds.

4
Use high resolution text and images.

● Poor resolution charts and text slow down the process of

scanning and annoy the user.
● In the worst cases, the charts become too difficult to
interpret.
● Fortunately, this is rarely an issue with most dashboarding
software today.

4
Align content whenever appropriate.

● Poorly aligned dashboards create discomfort even when

we’re not consciously aware of it.
● Most dashboard tools make this easy to fix quickly so it
shouldn’t be ignored.

4
Use a legible font.

● Some fonts are design for legibility while others are designed for
effects.
● Always use a font with good legibility on dashboards.
● Fonts with variable widths are easier to read.
● San-serifs fonts are easier to read on screens.

4
5. Expose lower-level conditions.

● Sometimes details must be understand at a lower level

than what’s displayed.
○ E.g., Europe sales activity is Europe overall may look
like fine but the UK is tanking.
● Simple alerts can help highlight that something is amiss so
that viewers know to take a deeper look.
○ A text-based message such as, “All is fine at this regional
level, but something deeper is amiss.”
○ An icon base system with variations for top level and
lower level alerts.

5
6. Prevent excessive alerts.

● Don’t allow a dashboard to create too many alert or they

will get ignored.
● Make sure the threshold of activity (e.g., a reduction in
sales) that triggers alerts draws attention to real problems
that require real response.
● More signal less noise.

6
7. Keep viewings in the loop.

● Situational awareness is only maintained when people are

kept in the loop.
● These systems cease to work when we over-automate.
● There should always be at least one simple step required
by the person who needs to maintain situational
awareness in highly automated systems.

7
8. Accommodate real-time monitoring.

● Dashboards that need real-time monitoring have different

design requirements
○ Reduce information to what’s essential.
○ Provide a means to temporarily halt updates.
○ Provide audio alerts.
○ Time-stamp alerts.
● Use need to be able to spot situations that need attention
and respond quickly.

8
Reduce information to what’s essential.

● Real-time dashboard users can’t handle as much

information on a screen because the data is constantly
changing.
● Non-essential information can still be easily accessible.

8
This dashboard helps users spot sudden increases or decreased in
traffic and immediately identify key drivers. Additional
information is only a click away.

8
Provide a means to temporarily halt updates.

● If the information on a dashboard is changing very quickly,

the user may need to pause updates to take a closer look.
● Design features should make it clear that the dashboard is
paused and the data is going stale so users don’t forget.
○ Can be subtle or increasing loud as needed.

8
Provide alerts and alert escalation.

● While real-time dashboards are monitored

frequently, they’re not typically monitored constantly.
People have other tasks and need to eat lunch.
● When an alert hasn’t been acknowledge after a set
amount of time, it can escalate if needed.
○ Popping up a notification.
○ Sending a text message.
○ Setting off an audio alarm.

8
Time-stamp alerts.

● Alerts need time stamps so that the people viewing them

know how fresh they are.

8
 Day 3 Activity 1
You will be graded on completion.
Identify a well designed dashboard.

Identify a well-designed dashboard and indicate what you

think it does well. Bullet points are fine.

Upload a PDF that includes a screenshot of the dashboard and

your response.

You can find dashboards on Tableau Public but you’re

welcome to use other sources, too.
Common Mistakes in
Dashboard Design
Dashboards post unique data visualization challenges.

● Squeezing all the necessary information in the allotted

space while keeping the display easily and immediately
understandable is hard!
● Vendor websites are full of poorly designed dashboards so
it’s crucial to always look at examples with a critical eye
while remembering fundamentals.
13 Common Mistakes to Avoid in Dashboard Design

1. Exceeding the boundaries of a single screen.

2. Supplying inadequate context for the data.
3. Displaying excessive detail or precision.
4. Expressing measures indirectly.
5. Choosing inappropriate display media.
6. Introducing meaningless variety.
7. Using poorly designed display media.
8. Encoding quantitative data inaccurately.
9. Arranging information poorly.
10. Highlighting important information ineffectively or not at all.
11. Cluttering the display with visual effects.
12. Misusing or overusing color.
13. Designing an attractive visual display.
 1. Exceeding the Boundaries of a Single Screen.

● We can only a few chunks of information at a time in

working memory and we can’t reliably recall information
that is no longer visible.
● Having to move between visible parts of a dashboard
negates our ability to compare data and spot insights.
● We tend to assume that anything beyond our immediate
field of vision is less important.

1
 Data is fragmented across many screens on this dashboard.

❌ Most of the screen is wasted on empty space and large navigation.

❌ The user must visit 10 different screens to update their situational
awareness.

1
 Scrolling is required to gain insights from this dashboard.

❌ Multiple scroll bars must be navigated to understand the chart.

❌ The most important information is not prioritized within a single
screen.

1
2. Supplying Inadequate Content for the Data

● Measures are useless context to determine if they are

good or bad.
● Supplying the right context is the difference between
numbers that just sit there or numbers that enlighten and
inspire action.
● More is not necessarily better; the right amount depends
on the needs of viewers.

2
Crucial context is missing from this dashboard.

❌ Radial scale is not labelled so tics are meaningless.

❌ Gauges neglect to indicate how good or bad these numbers are.

2
This gauge incorporates context more effectively but gauges should
usually be avoided.

2
3. Displaying Excessive Detail or Precision

● Measures that are expressed with too much detail or

precision slow viewers down with providing any benefit.
● Every unnecessary piece of information results in time
wasted as the viewer has to filter out what’s important.

3
This level of precision in this dashboard only slows down the user.

❌ Top right highlighted section displays time down to seconds.

❌ Lower‐right highlighted section displays from 4 to 10 decimal
digits for each measure.

3
4. Expressing Measures Indirectly

● For a measure to be meaningful we must know what is

being measured and the units of measure.
● A measure is deficient if it’s not the one that most clearly
and efficiently communicates the meaning the dashboard
viewer needs.
● Indirect measures create more work for viewers to extract
the right meaning and slow them down.

4
This chart is using the wrong measure for its intended purpose.

❌ Chart encodes dollar amounts rather than variance.

❌ Users must do their own calculations to determine variance.

4
This chart is uses the measure that matches its intended purpose.

4
5. Choosing Inappropriate Display Media

● One of the most common design mistakes in all data

visualization.
● Choosing the wrong chart type for the data and meaning
that needs to be displayed.

5
This chart choice doesn’t make sense with the data.

❌ These pie chart slices don’t represent parts of a whole!

5
This chart needlessly obscures the data.

❌ It’s impossible to accurate understand the relationship between

revenue and costs using this bubble chart.
❌ It’s very hard for users to estimate area differences in area
between circles.

5
The same data is much more useful in a bar chart.

5
This chart should have been a table.

❌ ✅

5
6. Introducing Meaningless Variety

● Always choose the chart that will work best even if that
means the dashboard is filled with many versions of the
same chart.
● Viewers won’t be bored because multiple data is displayed
the same way but they will be aggravated if they struggle
to extract meaning due to poor chart choices.
● Consistency in means of display allows viewers to use the
same percentual strategy for interpreting all of the data
which saves time and energy.

6
This dashboard is a visual jumble of charts.

❌ Many of the charts are difficult to understand on their own.

❌ The same measures are presented multiple ways, adding to the
complexity.

6
7. Using Poorly Displayed Media

● Individual components of charts are crucial for their

overall effectiveness.
● Like charts themselves, they need to be designed to
communicate clearly and efficiently without distraction.
● They should aid, not inhibit quick and easy interpretation.

7
This chart is difficult to follow.

❌ Using a legend instead of

labelling the slices requires users’
eyes to move back and forth.

❌ The legend order doesn’t

correspond with the slices.

❌ The bright colors also aren’t

serving any purpose and are overkill.

7
This chart needs to be decluttered.

❌ Gridlines and 3D effect needlessly distract the viewer.

❌ Different colors don’t convey any meaning.

7
The design of this chart fails to help the user with its main purpose.

❌ The order of the bars makes it difficult to compare actual and

budget.
❌ The 3D effect and color brightness are also distracting.

7
8. Encoding Quantitative Data Accurately

● Graphical representations of quantitative data can be

erroneously designed in ways that misrepresent values.
● This leads to too much time being spent to uncover
insights at best and incorrect interpretations of the data at
worst.

8
This dashboards creates a misleading interpretation by not starting the
scale at zero.

❌ Revenue appears to be 4X costs in January when it’s actually less

than 2X costs.

8
9. Arranging Information Poorly

● Dashboards often need to present a large collection of information

in a limited amount of space; if it’s not well organized, the result is a
cluttered mess.
● Data should be place based on importance and proper viewing
sequence.
○ Most important information should be prominent.
○ Information that requires immediate attention should stand out.
● Data should be framed within a visual design that segregates data
into meaningful groups.
○ Data that should be compared should be visually design to
encourage comparisons.
● Even a dashboard with little information can look cluttered.

9
This dashboard is poorly arranged.

❌ Most prominent region given

to a logo and navigation controls
.
❌ Order size and order size trend
graphs should be grouped
together.

❌ Larger graphs and title draws

attention to tangential returns
section.

9
10. Highlighting Important Information Ineffectively or Not at All

● The viewers eyes should be immediately drawn to the

information that is most important.
○ This should be true even if it is not in the most visually
prominent area.
● All the information on a dashboard should be important
but that doesn’t mean it’s equally important.
● If everything shouts for attention, nothing will be heard.

10
It’s hard to know where to look on this dashboard.

❌ Navigation controls are too

prominent due to placement and
borders.

❌ All the charts are equally bold

and colorful creating a wash of
sameness.

❌ Chart legends attract undue

attention due to background
color.

10
11. Cluttering the Display with Visual Effects

● Artistry and decorative flourishes should be avoided.

● Even if viewers initially appreciate them, they will grow
weary of them over time because they don’t aid rapid
understanding.
● There is nothing wrong with plain. Plain is good.

11
This dashboard design only makes sense if the purpose was to
train users on an actual control panel.

11
This dashboard doesn’t benefit from being made to look like a
spiral bound notebook.

11
This dashboard is full of distracting ornamentation.

11
Inept graphics also flourish because many graphic artists
believe that statistics are boring and tedious. It then
follows that decorated graphics must pep up, animate,
and all too often exaggerate what evidence there is in
the data... If the statistics are boring, then you've got the
wrong numbers.

– Edward Tufte

11
12. Misusing or Overusing Color

● Color should not be used haphazardly.

● Should be used thoughtfully with an understanding of how
people perceive color and the significance of color differences.
○ When a color appears as contrast, our eyes are pay
attention.
○ When colors are the same, we relate them to one another.
○ We associate meanings with green, yellow, and red
● Keep in mind, people who are colorblind don’t see these
differences.

12
Color choice makes this chart is difficult to interpret.

❌ Colors are too similar so it’s difficult to figure out which label
corresponds which which slice.

12
Color choice makes this chart is difficult to interpret.

❌ Bright, high-contrast colors are distracting.

❌ Important lines could be greyed out to highlight the important
ones.

12
13. Designing An Unattractive Visual
Display

● Some dashboards are just ugly.

● This can lead to questions of skill
and professionalism.
● It can also cause stakeholders to
resent using them.
● Avoiding useless embellishment
and artistry does not mean
avoiding clean, pleasant
aesthetics.

13
The designer of this dashboard designed with intention but lacked skill.

❌ Graphs are stretched to fill up space.

❌ Grey used to fill in background is too dark.
❌ Image under the title is distracting and wastes space.

13
 Day 3 Activity 2
You will be graded on completion.
What’s Wrong With This Dashboard?

Visit this dashboard and identify some of the issues with its
design.

Risk & Compliance

Add your answer in Courseworks. Bullet points are fine.

Introduction to Data
Visualization Tools
Data Visualization Tool Access

At Work
● You’ll likely be limited to using the tools available at your
company.
● Importing data into other tools is typically not allowed and
creates a security risk.

Personal Projects
● You can use any tool you want with your own data or publicly
available datasets.
● Most tools have a free option to encourage more people to learn
how to use it.
Spreadsheets for Data Visualization

Excel Google Sheets

- Create simple charts from small - Create simple charts from small
datasets quickly, especially if the data datasets quickly, especially if the data
is already in Excel. is already in Google Sheets.

- Easy to customize but design - Easy to customize but design

features are limited. features are limited. Can handle
waterfall charts better.

- Seamless integration with Copilot for - Seamless integration with Google

AI assist and Power BI for advanced Presentations for dynamic updates
visualization. and Google Charts for website
publishing.

- Requires a license. - Always free and easier to collaborate.

Chart Creation Tools for Web Publishing

Google Charts Datawrapper

- Create basic, interactive charts that - Create basic, highly-customizable

can be interacted with on a website. charts that can be interacted with on
a website.

- Charts can be combined into - Charts can be combined into

dashboards. dashboards.

- Can be publish anywhere that - Can be publish anywhere that

supports embed code supports embed code and has a
Substack integration

- Some coding needed - No coding needed

Top Data Visualization Tools

Power BI Tableau

- Robust data visualization features that - Much better data visualization

are easier to learn. features with a slightly steeper
learning curve.

- Seamless integration with Microsoft - Integrates with a wider range of data

stack including Excel and SQL Server sources including cloud-based data
sources and web services.

- Only available on the web and on - Available on the web and on Windows
Window. Not available on Mac. and Mac.

- Built-in collaboration features - Large and active online community

including co-authoring and makes it easier to learn and solve
commenting. problems.
AI As a Data Visualization Tool

● Right now AI works best a copilot.

○ Have you favorite chatbot guide you as you create data
visualizations in a new tool.
■ Copilot can be used directly inside Excel
■ Gemini can be used directly inside Google Sheets
○ This can dramatically speed up onboarding to a new tool
or help you get unstuck.
● AI can also be used to create charts and dashboards.
○ Just like any data visualization tool, you will need to
instruct on how to design effective charts and
dashboards.
○ Watch out for hallucinations. They are everywhere.
Tableau Basics
Connecting Data to Tableau

● If you’re using Tableau in a company, you’ll probably have

a server connection for most data.

● You can also upload data in many different formats.

● Make sure you select the right format or it won’t work.

There are many public datasets you can use to practice data
visualization.

● Tableau Public

● Kaggle

You can also ask chatbots to help you generate fake datasets for you!
Connect to the ieor4532 cloud site and use the Superstore Datasource.
Connect to the ieor4532 cloud site and use the Superstore Datasource.
Connect to the ieor4532 cloud site and use the Superstore Datasource.
Tableau has a robust knowledge base and large online community to help
you learn.

● Tableau Knowledge Base

● Tableau Community Forums

● Tableau on Reddit

If you’re new to Tableau, you’ll probably need to use AI as a

copilot to today so you can find the settings you need faster.
Day 3 In-Class Assignment
Day 3 Project

Create a dashboard based on the following requirements.

You’ve been hired as a data analyst by Superstore, a national

retailer. Your job is to create one dashboard that helps all the
regional sales managers monitor performance and uncover
opportunities to improve.

Connect to the ieor4532 cloud site and use the Superstore

Datasource. When your dashboard is finished, publish it to the
Project 3 folder.
Project 2 Grading

● 25% Clarity and Visual Design

○ Is the dashboard clear, well-structured, and nicely
designed?
● 25% Data Accuracy and Representation
○ Does the dashboard show the right data, in the right
way?
● 25% Interactivity and User Experience
○ Is the dashboard easy to explore and use?
● 25% Analytical Depth and Insights
○ Does the dashboard surface relevant insights based
on the requirements?