Module 5:

The Art of Analytics

By Kevin Hartman
The Art Of
Analytics
What we will discuss

• Planning for your analytics •

expedition
• Identifying your business
objective
• Data, data everywhere
• Finding insights
• Getting your story straight
Lesson 1: A Plan for Your Analytics Expedition
A proper approach to data analytics requires a set thorough steps
to keep complicated analysis robust yet manageable
Five things discussed in this lesson:

• Getting to a big idea is hard, but following a process map will help analysts be
effective and efficient

• Analysts’ first critical step is to set a goal for their analysis which will guide
their analytics journey

• Analysts can collect data using three tried-and-true techniques (although one
among them stands out as the most capable technique)

• Finding patterns in data requires analysts to be purposeful in how they

investigate the data they collect

• Reporting results is a critical exercise that separates solid analysis from big
ideas
 BIG
YOU
IDEA

(Shuttleworth et al., 2010)

 Data

Knowledge
Stimuli
Data

Data BIG
YOU Data
IDEA
Data

Tension

Outsiders
Pain

Fold A to meet B to illustrate

(Shuttleworth et al., 2010) A the hard way to get a big idea B
Source: David McCandless
Source: David McCandless
Source: David McCandless
Source: David McCandless
Source: David McCandless
Source: David McCandless
Analytics Planning
A simple four-step process that balances thinking with action can ensure a robust
approach to data collection and produce successful analysis

Plan (Goal) Collect (Information) Analyze (Story) Report (Visual Form)

• Establish the campaign / brand / • Locate sources that house the required • Produce “tidy”, analysis-ready datasets • Leverage preattentive attributes in
company’s clear, singular objective to data identified in the planning step to ensure your analysis is error-free visual perception to quickly and
be addressed by your analysis effective communicate your meaning
• Utilize data mining tools and techniques • Proactively address data-quality issues
• Define key questions you will be asking necessary to collect required data and concerns • Ensure that recommendations are as
of the data clear and concise as possible
• Select a data management system that • Perform analysis techniques that lead
• Identify the type of analysis you will be balances your needs for power and you to draw conclusions from collected • Follow simple rules of design to
conducting and the resultant data you simplicity data visualize insights with impact
will need • Ensure the effectiveness of future • Compress learnings into easy-to- • Connect to your audience with passion
• Plan your approach to collecting your analysis by limiting bias in the data understand snippets by constantly to ensure your story is memorable
data — the data you require and the asking yourself “what’s the 60 seconds
sources (and tools) you will use story?”
1 Plan 2 Collect

3 Report 4 Analyze
Reference List

• McCandless, D. (2014). Information is beautiful. New York, NY: HarperCollins.

Lesson 2: Identifying Your Business Objective
The objective you adopt is your guiding North Star, dictating the
type of data you will collect and the analyses you’ll conduct
Five things discussed in this lesson:

• Identifying an advertiser’s Marketing Objective is critical for successful

analysis, but this process is frequently poorly executed and often goes awry

• Analysts can choose from five primary Marketing Objectives when planning for
analysis (along with a rarely applied sixth MO)

• Finding your Marketing Objective can be simple when using the CDJ and
asking key questions

• Prioritizing objectives is an important exercise, as the success of an analysis

hinges on the objective’s clarity and effect

• Design a plan for analysis using a modified scientific research approach

Analytics Planning
A simple four-step process that balances thinking with action can ensure a robust
approach to data collection and produce successful analysis

Plan (Goal) Collect (Information) Analyze (Story) Report (Visual Form)

• Establish the campaign / brand / • Locate sources that house the required • Produce “tidy”, analysis-ready datasets • Leverage preattentive attributes in
company’s clear, singular objective to data identified in the planning step to ensure your analysis is error-free visual perception to quickly and
be addressed by your analysis effective communicate your meaning
• Utilize data mining tools and techniques • Proactively address data-quality issues
• Define key questions you will be asking necessary to collect required data and concerns • Ensure that recommendations are as
of the data clear and concise as possible
• Select a data management system that • Perform analysis techniques that lead
• Identify the type of analysis you will be balances your needs for power and you to draw conclusions from collected • Follow simple rules of design to
conducting and the resultant data you simplicity data visualize insights with impact
will need • Ensure the effectiveness of future • Compress learnings into easy-to- • Connect to your audience with passion
• Plan your approach to collecting your analysis by limiting bias in the data understand snippets by constantly to ensure your story is memorable
data — the data you require and the asking yourself “what’s the 60 seconds
sources (and tools) you will use story?”
Categories Of Objectives
It is important to keep clear the relationship between metrics and a company’s goals,
which range from resultant business objectives to tactical campaign objectives

Business Objectives
Revenue Volume Profit
Grow the amount of money Increase the units sold, number of Grow the amount of money

the business brings in leads or capacity of the business the business has after costs

Marketing Objectives
Stimulate demand Build awareness Influence consideration Improve sales experience Reposition the brand Grow loyalty

Media Objectives
Reach & frequency Engagement Conversions

Campaign Objectives
Site visits / clicks Views View-through rate Click through rate Cost per point Cost per click
Source: Knob Creek (2009)
Primary Marketing Objectives
Being clear and singular on the business objective that will guide analysis is paramount, as it will
affect every step taken by the analyst

Build Influence Improve the Reposition Increase

Awareness Consideration Sales Process the Brand Loyalty
Do consumer recognize Do my products satisfy Do my point-of-sale Do the experiences I Do customers advocate
and recall my brand? consumers’ needs? efforts result in wins for deliver fulfill customer for my brand?

Asking whether If the products do not, my brand? expectations? Customer advocate for
consumers are aware of consumers will choose Leaks in the sales Unmet expectations can a brand will reveal
the brand or product will competitors’ items, process (e.g., cart lead to poor customer whether a brand can be
quantify the need for revealing the need to abandonment) point to a experience and a need satisfied with its level of
greater awareness. influence consideration. need for improvement. to reposition the brand. loyalty.
 Consider

Evaluate
Advocate

The Loyalty Loop

BUY

Source: McKinsey & Company

Do consumers Do my
recall and products satisfy
recognize my Consider consumers’ needs?
brand?
Evaluate
Advocate
Do customers
advocate
for my brand?

The Loyalty Loop

Do the experiences I
deliver fulfill customer
expectations?
Do my POS
BUY efforts result in
wins for my
Source: McKinsey & Company
brands?
 Brand Influence
Awareness Consider Consideration

Evaluate
Advocate
Grow
Loyalty

The Loyalty Loop

Reposition
The Brand

BUY
Improve Sales
Source: McKinsey & Company
Experience
 Objective

OBJECTIVE
(should be
(should viewed
be viewed as as a hypothesis
a hypothesis
before analysis is conducted)
before analysis is conducted)

Key
Key(“What”) question
“What?” question Key
Key(“How”) question
“How?” question Key
Key(“Why”)
“Why?” question
question

Data + Data + Data + Data + Data + Data + Data + Data + Data +

Source Source Source Source Source Source Source Source Source
 Objective

To increase sales, Company

(should be viewed
X must as a loyalty
grow brand hypothesis
before analysis is conducted)

What How Why

Key
has (“What”)
been question
consumer interest Key (“How”)
strongly question
have consumer groups Key (“Why”)
did some question
marketing drive
in our brand over time? advocated for our brand? sales while others did not?

Search Volume
Brand Awareness
Customer Call
Consumer
Customer
Twitter Sentiment
Product Sales Hashtagged
Twitter Sentiment

Over Time
Over Time
Transcripts
Segmentation
Satisfaction
(3P Sentiment Volume
Twitter Volume
(3P Sentiment
(Google) (3P Study) (CSR Database) (3P Study) (3P Study) Analysis Tool) (Company X) (Twitter API) Analysis Tool)
“On the role of the CM and the importance of objective setting”
Steve Moffat, Chief Marketing Officer of Guaranteed Rate
Reference List

• Court, D., Elzinga, D., Mulder, S., & Vetvik, O. J., (2009, June). The consumer
decision journey. McKinsey Quarterly, (3). Retrieved from https://
www.mckinsey.com/business-functions/marketing-and-sales/our-insights/the-
consumer-decision-journey
Lesson 3: Data, Data Everywhere
Trends in the ecosystem are making today’s technology, tools, and
data more accessible than ever before
Five things discussed in this lesson:

• Tying collected data to key questions – and ultimately to the identified goal of
the analysis – is critical for successful analysis

• Data can come in many forms, including raw, processed, unstructured, and
structured

• Analysts can use three techniques to collect data: accessing facilitated

downloads, using APIs, and scraping the web

• Facilitated downloads should be the primary access point to data.

• Ensuring data quality requires understanding how bias enters the data and
critical points of data handling that frequently lead to errors
Analytics Planning
A simple four-step process that balances thinking with action can ensure a robust
approach to data collection and produce successful analysis

Plan (Goal) Collect (Information) Analyze (Story) Report (Visual Form)

• Establish the campaign / brand / • Locate sources that house the required • Produce “tidy”, analysis-ready datasets • Leverage preattentive attributes in
company’s clear, singular objective to data identified in the planning step to ensure your analysis is error-free visual perception to quickly and
be addressed by your analysis effective communicate your meaning
• Utilize data mining tools and techniques • Proactively address data-quality issues
• Define key questions you will be asking necessary to collect required data and concerns • Ensure that recommendations are as
of the data clear and concise as possible
• Select a data management system that • Perform analysis techniques that lead
• Identify the type of analysis you will be balances your needs for power and you to draw conclusions from collected • Follow simple rules of design to
conducting and the resultant data you simplicity data visualize insights with impact
will need • Ensure the effectiveness of future • Compress learnings into easy-to- • Connect to your audience with passion
• Plan your approach to collecting your analysis by limiting bias in the data understand snippets by constantly to ensure your story is memorable
data — the data you require and the asking yourself “what’s the 60 seconds
sources (and tools) you will use story?”
 Strucured & Unstructured Data
Data collected will take two primary forms: highly mechanized, manageable data
known as “structured” and raw, unorganized data known as “unstructured”

Structured Data Unstructured Data

• Information that includes a data model • Information that does not have a pre-defined data model
• Typically well-defined and organized data with an expected • Typically text-heavy, but may contain data such as dates,
format as determined by the data model numbers, and facts as well
• Generally, but not always, easier to collect than unstructured • Might account for more than 70%–80% of all data in
data organizations
• Frequently can be imported directly into a data management • Frequently requires the use of a data mining tool, such as R, to
system collect

Source: Wikipedia, IDC “Digital Universe Study”

Source: http://tinyurl.com/wcswbha
Source: http://tinyurl.com/wcswbha
Source: http://tinyurl.com/wcswbha
Data Access Points
Growth in data accessibility and the techniques of tapping application program
interface (APIs) and scraping webpages for data make the entire web a large data
source for analysts

Facilitated Downloads APIs Webscraping

• Data tables are accessed through a GUI built by data • Typically just a URL and works like the way your • Data is intended for presentation, not necessarily
owner that directs query parameters browser makes a call download
• Accessible data is typically stripped of sensitive • API operation is a “machine-to-machine” process of • Unlike APIs and Bulk Download Sites, data exchange
information data exchange is not facilitated by data host
• Download limits, user-based access can be enforced • Scripting language, set of commands that meet data • Data can often be governed by copyrights and/or
by data owners as well host’s vernacular required to access data needs for special permissions
• Data access can be simplified, making data more • API keys, access tokens are used to manage data • Access to data ranges from simple (highlight, right-
broadly attainable access click, “save as”) to complicated (python scripts)
Data Access Points
Growth in data accessibility and the techniques of tapping application program
interface (APIs) and scraping webpages for data make the entire web a large data
source for analysts

Facilitated Downloads APIs Webscraping

• Data tables are accessed through a GUI built by data • Typically just a URL and works like the way your • Data is intended for presentation, not necessarily
owner that directs query parameters browser makes a call download
• Accessible data is typically stripped of sensitive • API operation is a “machine-to-machine” process of • Unlike APIs and Bulk Download Sites, data exchange
information data exchange is not facilitated by data host
• Download limits, user-based access can be enforced • Scripting language, set of commands that meet data • Data can often be governed by copyrights and/or
by data owners as well host’s vernacular required to access data needs for special permissions
• Data access can be simplified, making data more • API keys, access tokens are used to manage data • Access to data ranges from simple (highlight, right-
broadly attainable access click, “save as”) to complicated (python scripts)
American Fact Finder has been decommissioned and is no longer
available.

Data are now available at: data.census.gov

 VAST AMOUNTS OF DATA,
JUST A SEARCH AWAY
Google’s Dataset Search just launched as a full-
fledged search tool, collating all kinds of datasets from
all across the internet with useful info like publication
data, authors and file types available before you even
click through.
From NFL stats from the ’70s to catch records of
great white sharks in the northwest Pacific, Google
Dataset Search provides easy, facilitated access to a
vast amount of information.
There are about 25 million datasets available now
— actually just “a fraction of datasets on the
web,” Google told the Verge — but more will be
available as data hosts update their metadata.

Source: Toward Data Science / http://tinyurl.com/qoql2rg

Data Access Points
Growth in data accessibility and the techniques of tapping application program
interface (APIs) and scraping webpages for data make the entire web a large data
source for analysts

Facilitated Downloads APIs Webscraping

• Data tables are accessed through a GUI built by data • Typically just a URL and works like the way your • Data is intended for presentation, not necessarily
owner that directs query parameters browser makes a call download
• Accessible data is typically stripped of sensitive • API operation is a “machine-to-machine” process of • Unlike APIs and Bulk Download Sites, data exchange
information data exchange is not facilitated by data host
• Download limits, user-based access can be enforced • Scripting language, set of commands that meet data • Data can often be governed by copyrights and/or
by data owners as well host’s vernacular required to access data needs for special permissions
• Data access can be simplified, making data more • API keys, access tokens are used to manage data • Access to data ranges from simple (highlight, right-
broadly attainable access click, “save as”) to complicated (python scripts)
Source: www.w3schools.com
Source: www.tweetbinder.com
Data Access Points
Growth in data accessibility and the techniques of tapping application program
interface (APIs) and scraping webpages for data make the entire web a large data
source for analysts

Facilitated Downloads APIs Webscraping

• Data tables are accessed through a GUI built by data • Typically just a URL and works like the way your • Data is intended for presentation, not necessarily
owner that directs query parameters browser makes a call download
• Accessible data is typically stripped of sensitive • API operation is a “machine-to-machine” process of • Unlike APIs and Bulk Download Sites, data exchange
information data exchange is not facilitated by data host
• Download limits, user-based access can be enforced • Scripting language, set of commands that meet data • Data can often be governed by copyrights and/or
by data owners as well host’s vernacular required to access data needs for special permissions
• Data access can be simplified, making data more • API keys, access tokens are used to manage data • Access to data ranges from simple (highlight, right-
broadly attainable access click, “save as”) to complicated (python scripts)
 The Question:
To Scrape Or Not
To Scrape?
While tempting, scraping data from websites
is problematic.
Sophie Chou is a graduate student at the MIT
Media Lab focusing on machine learning and
journalism. After attending a session on web
scraping at the 2016 NICAR conference, she
wrote an enlightening piece on the technical
(and ethical) challenges surrounding web
scrapping.
Ms. Chou’s point of view was well illustrated
through a flowchart she developed for her
article, where nearly all roads lead to a
decision not to scrape.

Source: Sophie Chou / http://tinyurl.com/v4x758n

Source: Sophie Chou / http://tinyurl.com/v4x758n
Source: Sophie Chou / http://tinyurl.com/v4x758n
“On dats sources used in marketing analytics”
Pradeep Kumar, Chief Data Officer of FCB
Chicago, Illinois
Reference List

• Chou, S. (2016). To scrape or not to scrape: technical and ethical challenges of

collecting data off the web. Retrieved March 12, 2020, from https://
www.storybench.org/to-scrape-or-not-to-scrape-the-technical-and-ethical-
challenges-of-collecting-data-off-the-web/

• Waterman, T. (2020). Google just published 25 million free datasets. Retrieved

March 12, 2020, from https://towardsdatascience.com/google-just-published-25-
million-free-datasets-d83940e24284
Lesson 4: Finding Insights
Visualizing data can express stories we would not see when data is
in tabular form (if we recognize the patterns at all)
Five things discussed in this lesson:

• Analysts can choose from five analysis categories, each with its own degree
of difficulty and depth

• Meaningful patterns in data can take many forms, each with the ability to
unearth and (importantly) convey a unique story

• Finding patterns in data requires analysts to create views into the data using
visual techniques, often from multiple angles

• Analysts should challenge themselves to master a wide variety of visual

techniques

• R is an excellent tool used to conduct analysis efficiently on large datasets

Analytics Planning
A simple four-step process that balances thinking with action can ensure a robust
approach to data collection and produce successful analysis

Plan (Goal) Collect (Information) Analyze (Story) Report (Visual Form)

• Establish the campaign / brand / • Locate sources that house the required • Produce “tidy”, analysis-ready datasets • Leverage preattentive attributes in
company’s clear, singular objective to data identified in the planning step to ensure your analysis is error-free visual perception to quickly and
be addressed by your analysis effective communicate your meaning
• Utilize data mining tools and techniques • Proactively address data-quality issues
• Define key questions you will be asking necessary to collect required data and concerns • Ensure that recommendations are as
of the data clear and concise as possible
• Select a data management system that • Perform analysis techniques that lead
• Identify the type of analysis you will be balances your needs for power and you to draw conclusions from collected • Follow simple rules of design to
conducting and the resultant data you simplicity data visualize insights with impact
will need • Ensure the effectiveness of future • Compress learnings into easy-to- • Connect to your audience with passion
• Plan your approach to collecting your analysis by limiting bias in the data understand snippets by constantly to ensure your story is memorable
data — the data you require and the asking yourself “what’s the 60 seconds
sources (and tools) you will use story?”
 Five Categories Of Marketing Data Analysis
Each method of marketing data analysis features a varying level of sophistication
and returns a corresponding depth of insight

Descriptive Inferential Exploratory Causal Predictive

Typically first method Features the use of a Typically visual- “Gold standard” for Utilizes models to
applied, provides a small set of data to based, seeks to data analysis, utilizes predict the future
description of the infer something about discover connections correlation and value of data
data analyzed a larger set (i.e., and patterns in data regression to explore attributes when some
extrapolation) data relationships related data is moved

Source: Eric Peterson, “Web Analytics Demystified” (2004)

 PATTERNS IN DATA
Meaningful patterns in data can take many forms, each with the ability to unearth and
(importantly) convey a unique story

Change Clustering Relativity Ranking Correlation

Trend or instance of Collection of data Observations Position in a scale of Mutual relationship or

observations points with similar (or considered in relation achievement or connection between
becoming different dissimilar) values or in proportion to status two or more things
over time something else

Source: Eric Peterson, “Web Analytics Demystified” (2004)

Change Clustering Relativity Ranking Correlation

Line Bubble Pie Bar Scatter

Column Distribution Tree Map Spider Regression

Histogram Boxplot / Candlestick Word Cloud Sankey Double Bar

“On the need for diverse data and points of view”
John Kenny, Chief Strategy Officer of FCB
Chicago, Illinois
Lesson 5: Getting Your Story Straight
Your story connects the data you’ve collected to the objective you
hold and must be told in a concise, organized fashion
Five things discussed in this lesson:

• For the analyst, reporting out the insights they’ve found is a critically important
undertaking that cannot be taken lightly

• People struggle to retain information, presenting a big challenge to analysts

who want audiences to remember what they say

• Several studies illustrate how easy it is for people to become overwhelmed by

information

• Analysts must limit the amount of brain activity their stories require

• An organized story will also help the analyst most effectively align supporting
data to substantiate their key points
Analytics Planning
A simple four-step process that balances thinking with action can ensure a robust
approach to data collection and produce successful analysis

Plan (Goal) Collect (Information) Analyze (Story) Report (Visual Form)

• Establish the campaign / brand / • Locate sources that house the required • Produce “tidy”, analysis-ready datasets • Leverage preattentive attributes in
company’s clear, singular objective to data identified in the planning step to ensure your analysis is error-free visual perception to quickly and
be addressed by your analysis effective communicate your meaning
• Utilize data mining tools and techniques • Proactively address data-quality issues
• Define key questions you will be asking necessary to collect required data and concerns • Ensure that recommendations are as
of the data clear and concise as possible
• Select a data management system that • Perform analysis techniques that lead
• Identify the type of analysis you will be balances your needs for power and you to draw conclusions from collected • Follow simple rules of design to
conducting and the resultant data you simplicity data visualize insights with impact
will need • Ensure the effectiveness of future • Compress learnings into easy-to- • Connect to your audience with passion
• Plan your approach to collecting your analysis by limiting bias in the data understand snippets by constantly to ensure your story is memorable
data — the data you require and the asking yourself “what’s the 60 seconds
sources (and tools) you will use story?”
Source: David McCandless
“
Studies have shown that
80% of what people learn
is forgotten within 24
hours.

- Will Thalheimer, How Much Do

People Forget? (Dec 2010)
1 2

3 4
Source: Nova PBS Official / https://www.youtube.com/watch?v=VkrrVozZR2c :
 Objective

To increase sales, Company

(should be viewed
X must as a loyalty
grow brand hypothesis
before analysis is conducted)

What How Why

Key
has (“What”)
been question
consumer interest Key (“How”)
strongly question
have consumer groups Key (“Why”)
did some question
marketing drive
in our brand over time? advocated for our brand? sales while others did not?

Search Volume
Brand Awareness
Customer Call
Consumer
Customer
Twitter Sentiment
Product Sales Hashtagged
Twitter Sentiment

Over Time
Over Time
Transcripts
Segmentation
Satisfaction
(3P Sentiment Volume
Twitter Volume
(3P Sentiment
(Google) (3P Study) (CSR Database) (3P Study) (3P Study) Analysis Tool) (Company X) (Twitter API) Analysis Tool)
 Objective

To increase sales, Company

(should be viewed
X must as a loyalty
grow brand hypothesis
before analysis is conducted)

Consumer interest in Advocacy for our brand has Advocacy for our brand has
Key (“What”)
our brand has question
reached Key
been(“How”)
driven byquestion
specific Key (“Why”)
a clear question
and unmistakable
all-time high levels consumer segments only link to increased sales

Search Volume
Brand Awareness
Customer Call
Consumer
Customer
Twitter Sentiment
Product Sales Hashtagged
Twitter Sentiment

Over Time
Over Time
Transcripts
Segmentation
Satisfaction
(3P Sentiment Volume
Twitter Volume
(3P Sentiment
(Google) (3P Study) (CSR Database) (3P Study) (3P Study) Analysis Tool) (Company X) (Twitter API) Analysis Tool)
Reference List

• Nova PBS Official. (2011). Inside NOVA: Change blindness [Video]. Retrieved from
https://www.youtube.com/watch?v=VkrrVozZR2c