Data Storytelling 1 Day

Data Storytelling
www.thecads.com
Welcome!
Copyright © 2019 CADS and/or its affiliates. All rights reserved. CADS Confidential – Internal/Restricted/Highly Restricted 2
Introductions
Why are you interested in Data Storytelling?
What kind of work do you do with data?
What kind of Data Analysis applications do you use?

Excel? Tableau? Power BI? Python Programming? R? VBA? Etc.
Module Goals
Make stories that highlight value from data-

driven insight.
Design intuitive visuals that are easy for

the audience to interpret.
Present findings through stories.
Program Agenda
The Data Storytelling Data Storytelling: Activity

1 Process 3 (2 hours) Go through the process and
present from a real dataset
(2 hours) Who? What? How? And
storyboarding
Bad Data Storytelling Graphical Perception

&
( 1 hour) Recognize what is going wrong
to begin the process of change 2 Choosing a Visual 4
( 2 hours) Our 6 standard visuals in depth
Data Storytelling is done in 3 distinct related and sequential
parts.
Story
1 What do you want to show you audience
3
of meaning from the data?
Data Visual
You must have it. It should be reputable. It
should be about what you want to know.
2 Communication of the story beyond
words for maximum: comprehension,
speed, and interest
Bad Data Storytelling
Notice that the next few examples will have “okay” visuals with bad/non-existent stories.
The core questions of any visual
Key Questions
1. What do they want you to know?

2. What do they want you to do? (now that you know #1)
Sub-questions
3. What technical errors do you notice? (list)
4. How/Why was this made? Can you empathize with 1-3?
Beautiful Visual – but what exactly is the story here?
Hint: Notice the slices!
Copyright © 2019 CADS and/or its affiliates. All rights reserved. CADS Confidential – Internal/Restricted/Highly Restricted Source: https://twitter.com/yougov/status/838720989991223297 9
Just because we CAN fit everything on one page..
Doesn’t mean you SHOULD
Copyright © 2019 CADS and/or its affiliates. All rights reserved. CADS Confidential – Internal/Restricted/Highly Restricted Source: https://www.forbes.com/sites/ywang/2017/08/08/xiaomi-is-once-again-worlds-biggest-smartphone-brand-but-could-it-go-further 10
Distorted Geographic Map
How would you interpret this map of America?
Copyright © 2019 CADS and/or its affiliates. All rights reserved. CADS Confidential – Internal/Restricted/Highly Restricted Source: http://livingqlikview.com/the-9-worst-data-visualizations-ever-created/ 11
Data Storytelling to push an agenda
Is global warming real? Not everyone agrees.
Copyright © 2019 CADS and/or its affiliates. All rights reserved. CADS Confidential – Internal/Restricted/Highly Restricted
Source: https://www.mediamatters.org/blog/2009/12/08/fox-news-fiddles-with-climate-change-polling/157839
12
Data Storytelling to push an agenda
The math just doesn’t add up
Copyright © 2019 CADS and/or its affiliates. All rights reserved. CADS Confidential – Internal/Restricted/Highly Restricted Source: http://livingqlikview.com/the-9-worst-data-visualizations-ever-created/ 13
A good Data Story combines BOTH Story & Visual
Show the numbers but with purpose
What is the visual trying to tell us?
BP 2017 Annual Report
Beautiful Visual without a cohesive story
Too much information, too small a space
Cognitive overload
Uninformative title
Proportions are not

comparable
Axes are not clear
Suggested re-design – Refine the message
Make “Renewables” the focus of the visual & build a visual hierarchy
What is this visual trying to tell us?
New Zealand Oil & Gas 2015 Annual Report
What is this visual trying to tell us?
New Zealand Oil & Gas 2015 Annual Report
What exactly is this analyst trying to tell us
We must think of the story before plotting any visuals
Assuming we are presenting
Core Visual
Expenses are greater than income due to our change in strategy.
The Who, What, How method
Build a profile of your audience,
Think about what to tell them,
Consider how you can tell them this story.
Example – Who is your audience?
Vice – President of Strategy at our Toy Company
1. What are they like? 4. How can you solve their problem?
Older / Previously from technical background but

Suggest a viable business strategy to curb losses /
has been more client facing recently / Detail-
Back strategy up with appropriate data
oriented / Time-poor
2. Why are they here? 5. How can you best reach them?
Dislikes lengthy meetings / Tends to rush out last-

She is looking for a solution to our problem of
minute / PowerPoint with Executive Summary &
falling income & rising expenses
appendix will be good
3. What keeps them up at night? 6. How might they resist?
She is new to the role and needs a win / The proposal must fit within budgetary
Company Shareholders are critical of her ability / constraints / needs to be done with readily-
Operational losses unsustainable in long-run available resources
Your turn! Use the template on someone you present to
E.G. Your Boss or a client
Cultural Dimensions can help us understand our audience
Nationality has a big effect on the stories we tell and visuals we create
Copyright © 2019 CADS and/or its affiliates. All rights reserved. CADS Confidential – Internal/Restricted/Highly Restricted Source: https://www.hofstede-insights.com/country-comparison/malaysia/ 25
We all have different associations for color
Color usually differs on culture
Source: https://www.six-degrees.com/an-international-guide-on-the-use-of-color-in-marketing-advertising/
26
The direction we read is also influenced by culture
How will their eyes naturally process a page?
Now we know Who the audience is.
What do we want to tell them?
Show them pearls not shells.
Always consider the audience – how is this specifically relevant to them.
X O
This is all the data. What’s the pearl?
So what?
Student Interest Science Before & After

attending a Summer Science Camp
Source: Storytelling with Data pg. 5
This is all the data. What’s the pearl?
So what?
“I want the audience to know the camp was a success.”

Student Interest Science Before & After
attending a Summer Science Camp
Copyright © 2019 CADS and/or its affiliates. All rights reserved. CADS Confidential – Internal/Restricted/Highly Restricted Source: Storytelling with Data pg. 5 31
Data Stories generally boil down to 1 of these 6 types.
You should be able to state which you are presenting.
Distribution
Ranking Proportions
Change over
Correlation
Time
Comparison
Rankings are a quick way to know results
Who won? - Leads to questions like how? & why? & how much?
Our regional sales were similar this year, with West bringing in the most
Distribution profit.
RM23,520
RM22,828
RM21,662 1st
RM19,891 3rd 2nd
Ranking Proportions 4th
Sales in RM 2016
Change over
Correlation
Time
Comparison
North East South West
Sales Region
Distributions generate descriptive stats to generate insight.
Don’t get fooled by averages, and buckets. Great for outliers
Distribution
Ranking Proportions
Change over
Correlation
Time
Comparison
Proportion stories identify a part of a whole.
Area Maps and 100% Stacked are the usual standard.
Data Science Projects, on the whole, fail to launch. Before

Distribution we begin one, we should research best practices.
Ranking Proportions
Data Science Projects

68% 32%
Change over
Correlation
Time
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Comparison
Correlations stories establish relationships between known variables
Is profit related to distance?
There is a positive correlation between the distance of a delivery and the
profit we made. We should test if this is causal relationship.
RM450.00
Distribution RM400
RM400.00
RM350.00
Ranking Proportions
RM300.00 A wedding of
over 300 pax
RM250.00
RM200
Profit
RM200.00
Change over RM150.00

Correlation
Time RM100.00
This order used
RM110 more gas due to
dirt roads.
RM50.00
Comparison
RM0.00
0 2 4 6 8 10 12 14
Distance of Delivery in KM
Unit to Unit Comparisons are essential for good decision making.
Bar charts, area maps, scatter plots, line charts, and even heat maps are effective.
Distribution
Ranking Proportions Profit Average

Product Category
Margin Price
A Furniture 15% 123
B Bedding 20% 198
Change over C Office Supplies 10% 212
Correlation
Time D Home Decor 25% 433
Comparison
Change-over-time is essential for trends and forecasting.
Line charts are the most common, but bar charts and scatter plot work too.
Distribution
Ranking Proportions
Change over
Correlation
Time
Comparison
Ok. So now we know what we want to tell
them. But, what do I put in the tile?
Hybrid Titles – the multifunctional title of the future.
Hybrid Titles
Insight, Call-to-Action & Legend all rolled into 1
Insight
“The pilot summer science program was a success. More children are interested
in science after the camp as compared to before. Please approve the budget
required for next year’s camp.”
Legend
Call-to-action
Call-to-action
“Please consider restarting sales of Product A & B while Product C gains

traction in the market. Marketing & R&D Expenses of Product C are not
sustainable without Income streams from Product A & B.
Insight
Legend
What exactly is a call-to-action?
Don’t be afraid to start the conversation with these words below.
Example: What do I want to tell the audience
Consider our Toy Company VP
1. What do I want to show the audience?
Change-over-time for both income & expenses / Absolute value of losses incurred
at current date
2. What is the context

Losses incurred is the result of our change in strategy. We stopped selling
Products A & B because we wanted to focus on Product C. It was anticipated that
Product C would do well in the market but this has not been the case. Rising R&D
/ Marketing costs to support Product C not sustainable without revenue from
Product A & B.
3. Do I have a call-to-action?
Convince the VP of Strategy to restart sale of Product A & B. Revenue streams will
be used to support the further-development of Product C in the market.
Your turn! Fill in the blanks for ‘what’ based on your audience
This will help with “so what?” and hybrid titles
How do I tell this story?
Remember that the data comes first.
Data –Does yours pass the BET test?
Check everything you know.
Bias & Assumptions
Wasthere
Was there any
any bias
biaspresent
presentwhen
whenthethe
data waswas
data collected?
collected?
B What assumptions
What assumptionswere
process?
process?
mademade
were during during
the datathe
collection
data collection
Enough data to support story
E
Correlation isisnot
Correlation notcausation
causation
Do not
Do notmake
makelargelargeclaims that
claims ourour
that datadata
cannot support
cannot support
True & Fair View
T
Graphical
Graphical Integrity achieved
Integrity achieved
Lie
LieFactor
Factor within acceptablerange
within acceptable range (0.8
(0.8 ~ 1.2
~ 1.2)
How –Biases & Assumption
Think clearly. We often aren’t aware of the assumptions of biases we have.
Are you asking people inside a KFC restaurant which fried chicken is their favorite?
Be mindful of assumptions/bias present in your data.
Person Restaurant
A KFC
B KFC
C KFC
D McDonalds
E KFC
How –Enough data to support your story
Avoid making claims with no data
Be mindful that the claims we make are supported by our story. Just because our
dog barks on days that it rains, does not make it a magic dog!
Dog barks in the Rain that evening

morning
Y Y
N N
Y Y
N N
Y Y
How –Enough data to support your story
Avoid making claims with no data
How –True and Fair View
Be mindful of not misrepresenting your data to the audience
Edward Tufte’s Lie Factor

A measure for the True & Fair View
How –True and Fair View
Suspiciously large sales figures for iPhones
Source:Wired.com
Graphical Integrity
Edward Tufte’s ‘Lie Factor’
Graphical Integrity
Misrepresentation of Data
Source: Washington Post
Graphical Integrity: Breakdown
Misrepresentation of Data
Size of effect !.#$%.#

× 100% = 44%
in graphic %.#
Size of effect %#$%&

× 100% = 4.7%
in data %&
%%
Lie Factor = 9.36
%.'
Anything below 1.0 is understating, and

anything above 1.0 is overstating
Graphical Integrity: in the media
Exaggeration of data
Scale Distortion is used to mislead
A zero baseline is expected
Original Zero baseline

42% 39.6%
40%
35.0%
40% 39.6%
30%
Top Tax Rate
Top Tax Rate

38% 20%
36% 10%
35.0%
34% 0%
NOW JAN 1, 2013 NOW JAN 1, 2013
Example: How do you want to tell this story?
Consider our Toy company VP
1. Bias & Assumptions
Assume that we only have three revenue streams (Products A, B, & C)/ Assume
that market failure of Product C can be corrected through further development.
2. Enough data to support your story?

We could supplement our story with the revenue breakdown by product as well
as the profit margin & expected sales values for each product/ Our underlying
analysis shows that this should not be a problem but we could include this
information in the appendix because the VP is detail-oriented.
3. True and Fair View
Calculated Later
Your turn! Fill in the blanks for ‘how’ based on your audience
This is essential to telling your story and your Call to Action.
Group Activity
Designing a data story from scratch
Checklist
• Tick off the items here as you complete them
⎕ Examine your dataset (one per team)
⎕ Fill-in – ‘Who’ template
⎕ What will you tell them? – ‘What’ template - Analysis
⎕ How will you tell this story? – ‘How’ template
⎕ Craft a big idea that defines your entire story
⎕ Storyboard & decide on an order – ‘Storyboard’ template
⎕ Be prepared to verbally present your Data Story.
Activity: Choose a topic
• One topic & 2 slides per team
#1 #2 #3 #4
Pollution Salary Unemployment Naps
#5 #6 #7 #8
Public Transport Online Course Exercise Work Hours
Graphical Perception
(McGill & Cleveland, 1984)
The Visual Decoding of Information Encoded

on a Graph
How much longer is B compared to A?
• Graphical Perception: Elementary Perceptual Tasks: Linear Comparison
4x
Source: Harvard CS109 – Hanspeter Pfister and Joe Blitzstein http://cs109.org
How much steeper is A compared to B? -use degrees.
• Graphical Perception: Elementary Perceptual Tasks: Angular Comparison
6x
78

How much larger is B compared to A?
• Graphical Perception: Elementary Perceptual Tasks: Area Comparison
10x
How much darker is B compared to A?
• Graphical Perception: Elementary Perceptual Tasks: Contrast or Saturation Comparison
2x
How much bigger in value is B compared to A?
• Graphical Perception: Elementary Perceptual Tasks: Multi-Step
4x
Which perceptual tasks are used by common chart types?
• McGill & Cleveland (1964)
USED BY
Scatter Plot
Bar Chart
Line Chart
Pie Chart
Heat Map
Stacked Bar Chart
Dot Plots & Bar Charts are very effective
• They use position and length to convey information
Pie Charts & Donut Charts are less effective
• The human-eye cannot perceive 2D-area and angle that well
Choosing a Visual Histograms,
• Form should follow function Scatterplots
Distribution
Ranking Proportions
Bar Charts
100% Stacked Bar

Change over
Time
Correlation charts, Area Maps
Line Charts, Bar Charts
Comparison
Bar charts, Stacked Bar, Scatterplots, Line Charts
Scatterplots,
85
Reverse causation: Countries which spend more on health care tend to have longer life expectancy
But what if countries spend more on health care because they have lower life expectancies? (Filmer &
Pritchett 1999)
Gapminder
Scatterplots – Is having a degree correlated to voting ‘Leave’
Pre-attentive attributes
Two dimensions maximum
Use color to differentiate
Bar Charts – Compare between categories
• Very common charts – easy to interpret
Single Series
8
• Rule of Thumb: Four multi-series maximum 7

6
• Multiple series can quickly overwhelm the audience 5
4
3
• Visual grouping happens as a result of the spacing in bar charts having 2
more than one data series 1
• Spacing between bars for same category must be smaller than between 0
Cat 1 Cat 2 Cat 3 Cat 4 Cat 5
categories Multiple Series
• Spacing between categories should be approx. 0.5 of bar width 8
7
6
• Categories must be ordered in a logical manner such as: 5

4
• Alphabetical order 3
• By relative importance 2
1
0
Cat 1 Cat 2 Cat 3 Cat 4 Cat 5
A B C D
Line Graphs – Single Variable Change Over Time
• Good for time series data
• The line graph can show a single series of data,

two series of data, or even multiple series
• When using multiple series its best to highlight

the series that you want your audience to notice
• Used for plotting continuous data

• Points are physically connected via the line
• Implies a connection between the points (continuity)
Line Graphs – Best Practice
• Annotation of significant events are nice to have for context
Call-to-action
Axis labels
Event
Labels
Pie Chart – Bad Example
• Too much!
Source: 2018, Stephen Few, PerceptualEdge
Tree maps – (Imperfect) Alternative to Pie Chart
• Notice what happens when we have many sub-categories
• Tree maps display hierarchical data as a set of

nested rectangles.
• Each branch of the tree is given a rectangle,
which is then tiled with smaller rectangles
representing sub-branches.
• A leaf node's rectangle has an area proportional
to a specified.
• Often the leaf nodes are coloured to show a
separate dimension of the data
• Tree maps use space efficiently compared to pie
charts but we cannot legibly read small sub-
categories
Text – Impact Metrics
• Draw attention with large font, place a semi-opaque background image for context
Text – Text ads in the wild
• Lessons from Spotify – less is more
Choosing a Visual - Table
• Never use this in presentations
u Remove/lighten borders
Product Category Profit Margin Average Price

A Furniture 15% 123.1
B Bedding 20% 198.1
C Office Supplies 10% 212.2
D Home Decor 25% 432.5
v Left align text w Right align numbers
Choosing a Visual - Table
• Never use this in presentations
A Furniture 15% 123.1
B Bedding 20% 198.1
C Office 10% 212.2

Supplies
D Home Decor 25% 432.5
Choosing a Visual - Heatmap
• Use both sides of the brain for impact
Color saturation
A Furniture 15% 123
B Bedding 20% 198
C Office Supplies 10% 212
D Home Decor 25% 433
Stacked Bar Chart
• Proportion to whole & Comparison between categories
• Help compare totals across categories and see Many sub-categories can be difficult to
compare
the subcomponent proportions within a given
20
category 18
16
• This can quickly become visually overwhelming 14
12
• Hard to compare the subcomponents across 10
the various categories beyond the bottom 8
6
series 4
2
• We no longer have a consistent baseline to use to 0
compare. Cat 1 Cat 2 Cat 3 Cat 4 Cat 5
A B C
Horizontal Stacked Bar Chart
• Normalize to 100% for easier comparison
Visualisation of survey data collected along a Likert scale (a scale

• Show totals for categories summed to 100% commonly used in surveys that ranges from Strongly Disagree to Strongly
Agree)
• Good for comparison if there are uneven number
of responses for each question
• More useful for visualising proportions to a

whole on a because we get a consistent
baseline on both the far left and the far right
• Easy comparison of the left-most pieces as well as
the right-most pieces
Stacked Horizontal Bar Chart – Bad Example
• Cognitive Overload
Choosing a
Visual
Form follows
function.
Choose a visual
that supports the
‘What’
Group Activity
Designing a data story from scratch
Checklist
• Tick off the items here as you complete them
⎕ Examine your dataset (one per team)
⎕ Consider the audience – ‘Who’ template
⎕ What will you tell them? – ‘What’ template - Analysis
⎕ How will you tell this story? – ‘How’ template
⎕ Craft a big idea that defines your entire story
⎕ Storyboard & decide on an order – ‘Storyboard’ template
⎕ Create 2 visuals per team
Activity: Choose a topic
• One topic & 2 slides per team
#1 #2 #3 #4
Pollution Salary Unemployment Naps
#5 #6 #7 #8
Public Transport Online Course Exercise Work Hours
Your turn! Use the template on someone you present to
E.G. Your Boss or a client
Your turn! Fill in the blanks for ‘what’ based on your audience
This will help with “so what?” and hybrid titles
Your turn! Fill in the blanks for ‘how’ based on your audience
This is essential to telling your story and your Call to Action.
Storyboard Template
Tell them the outcome and how you go there.
#1 #2 #1
Thank you. www.thecads.com

Data Storytelling 1 Day

Uploaded by

Copyright:

Available Formats

You might also like

Data Storytelling 1 Day

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Data Storytelling 1 Day

Uploaded by

Copyright:

Available Formats

Data Storytelling

Why are you interested in Data Storytelling?

What kind of work do you do with data?

What kind of Data Analysis applications do you use?

Make stories that highlight value from data-

Design intuitive visuals that are easy for

Present findings through stories.

The Data Storytelling Data Storytelling: Activity

Bad Data Storytelling Graphical Perception

1. What do they want you to know?

Proportions are not

Axes are not clear

Build a profile of your audience,

Think about what to tell them,

Consider how you can tell them this story.

Older / Previously from technical background but

Dislikes lengthy meetings / Tends to rush out last-

3. What keeps them up at night? 6. How might they resist?

3. What keeps them up at night? 6. How might they resist?

What do we want to tell them?

Student Interest Science Before & After

“I want the audience to know the camp was a success.”

Data Science Projects, on the whole, fail to launch. Before

Data Science Projects

Change over RM150.00

Ranking Proportions Profit Average

Hybrid Titles – the multifunctional title of the future.

“Please consider restarting sales of Product A & B while Product C gains

1. What do I want to show the audience?

2. What is the context

1. What do I want to show the audience?

2. What is the context

Enough data to support story

True & Fair View

Dog barks in the Rain that evening

Edward Tufte’s Lie Factor

Size of effect !.#$%.#

Size of effect %#$%&

Anything below 1.0 is understating, and

Original Zero baseline

Top Tax Rate

1. Bias & Assumptions

2. Enough data to support your story?

3. True and Fair View

1. Bias & Assumptions

2. Enough data to support your story?

3. True and Fair View

⎕ Examine your dataset (one per team)

⎕ Fill-in – ‘Who’ template

⎕ What will you tell them? – ‘What’ template - Analysis

⎕ How will you tell this story? – ‘How’ template

⎕ Craft a big idea that defines your entire story

⎕ Storyboard & decide on an order – ‘Storyboard’ template

⎕ Be prepared to verbally present your Data Story.

Pollution Salary Unemployment Naps

Public Transport Online Course Exercise Work Hours

The Visual Decoding of Information Encoded

Source: Harvard CS109 – Hanspeter Pfister and Joe Blitzstein http://cs109.org

Stacked Bar Chart