Proctor Exam 2- Secondary Data Syndicated and Big Data

Question 1

Machine learning can be applied anywhere there is a

need for quick automatic decisions based on ongoing feedback from patterns in the environment.

Answer 1

Machine learning can be applied anywhere there is a

need for quick automatic decisions based on ongoing feedback from patterns in the environment.

Question 2

Problems that lend themselves to machine learning:

1. The data causes problems for traditional analytic techniques, such as where variables
   are highly correlated, data is non-linear, or where there are far more variables than
   records (so called “wide and shallow” datasets)
2. Accuracy is more important than understanding
3. Potential outputs are defined, but the action is dependent on conditions which themselves cannot be easily predicted or identified before the event happens.
4. Rules and associations might be perceived or deduced, but are not easily described by logical rules

Answer 2

Problems that lend themselves to machine learning:

1. The data causes problems for traditional analytic techniques, such as where variables
   are highly correlated, data is non-linear, or where there are far more variables than
   records (so called “wide and shallow” datasets)
2. Accuracy is more important than understanding
3. Potential outputs are defined, but the action is dependent on conditions which themselves cannot be easily predicted or identified before the event happens.
4. Rules and associations might be perceived or deduced, but are not easily described by logical rules

Question 3

What makes this a good machine-learning problem is that the decisions and
the variables constantly change and the value of one variable and the right decision may depend
on the values of many more variables. Humans instinctively make these assessments, but it is
impossible to discretely list every rule and situation for a computer to look up and evaluate.

Answer 3

What makes this a good machine-learning problem is that the decisions and
the variables constantly change and the value of one variable and the right decision may depend
on the values of many more variables. Humans instinctively make these assessments, but it is
impossible to discretely list every rule and situation for a computer to look up and evaluate.

Question 4

Machines learn by studying data to detect patterns or by applying known rules (algorithms) to:
Categorize like or unlike people or things
Identify patterns and relationships that were unknown before analysis
Predict likely outcomes or actions based on identified patterns
Detect anomalous or unexpected behaviors

Answer 4

Machines learn by studying data to detect patterns or by applying known rules (algorithms) to:
Categorize like or unlike people or things
Identify patterns and relationships that were unknown before analysis
Predict likely outcomes or actions based on identified patterns
Detect anomalous or unexpected behaviors

Question 5

Machines learn through essentially an exhaustive process of trial and error, sifting through information, comparing the information to a goal, making adjustments, and trying again

Answer 5

Machines learn through essentially an exhaustive process of trial and error, sifting through information, comparing the information to a goal, making adjustments, and trying again

Question 6

Within Machine Learning The traditional advanced analytic techniques you will learn about later in this course are not well
suited for the unstructured nature of some big data

Answer 6

Within Machine Learning The traditional advanced analytic techniques you will learn about later in this course are not well
suited for the unstructured nature of some big data

Question 7

Machine learning, however, takes advantage of a computer’s ability to
follow rules and execute swift comparisons as a fast way to understand patterns and meaning in data.

The algorithms automatically sort data, testing and comparing what it has seen in the past to what it is seeing in the present. The learning may lead to a new understanding of behavior or it might serve as automatic input to an action executed by another computer process.

Answer 7

Machine learning, however, takes advantage of a computer’s ability to
follow rules and execute swift comparisons as a fast way to understand patterns and meaning in
data.

The algorithms automatically sort data, testing and comparing what it has seen in the past to
what it is seeing in the present. The learning may lead to a new understanding of behavior or it might serve as automatic input to an action executed by another computer process.

Question 8

Supervised learning always has a predetermined outcome provided by the programmer. The machine seeks faster, more efficient, or more accurate ways to meet the goal based on the data and the programmers input.

Answer 8

Supervised learning always has a predetermined outcome provided by the programmer. The machine seeks faster, more efficient, or more accurate ways to meet the goal based on the data and the programmers input.

Question 9

Supervised Learning

Process
Machine is given pre-classified data and discovers a pattern associated with the classification. As more data becomes available, the machine adjusts its associations and gets better at classifying

Example
Sorting out junk emails from wanted content

Limitations
Only works on one task at a time.
User may not be able to interpret the associations behind the sorting.

Traditional stat tool
Regression
Classification
Decision Trees
Random Forests
Bayesian statistics

Answer 9

Supervised Learning

Process
Machine is given pre-classified data and discovers a pattern associated with the classification. As more data becomes available, the machine adjusts its associations and gets better at classifying

Example
Sorting out junk emails from wanted content

Limitations
Only works on one task at a time.
User may not be able to interpret the associations behind the sorting.

Traditional stat tool
Regression
Classification
Decision Trees
Random Forests
Bayesian statistics

Question 10

Unsupervised Learning

Here the data determines the outcome. The algorithm’s mission is to extract structure from the data, and to present the structure in a way that is useful to us. Data is segmented and scored based on what the computer itself decides is relevant or related.

Answer 10

Unsupervised Learning

Here the data determines the outcome. The algorithm’s mission is to extract structure from the data, and to present the structure in a way that is useful to us. Data is segmented and scored based on what the computer itself decides is relevant or related.

Question 11

Unsupervised Learning

Process
Machine is given a lot of data and told to hunt for patterns and “clusters” of things related to each other. It draws its own conclusions about relationships.

Example
Recommendation engines
Loyalty card targeting

Limitation
Usually requires human input after the fact

Traditional Statistics Tool
Factor Analysis
Cluster Analysis
Multidimensional Scaling
Principle Component

Answer 11

Unsupervised Learning

Process
Machine is given a lot of data and told to hunt for patterns and “clusters” of things related to each other. It draws its own conclusions about relationships.

Example
Recommendation engines
Loyalty card targeting

Limitation
Usually requires human input after the fact

Traditional Statistics Tool
Factor Analysis
Cluster Analysis
Multidimensional Scaling
Principle Component

Question 12

Reinforcement Learning

This type of learning has no supervisor, but instead it has a reward signal that defines success. Similar to human learning, when success is rewarded, the machine tries to learn the patterns that result in receiving the reinforcement signal. The machine’s decisions affect the subsequent data it receives.

Answer 12

Reinforcement Learning

This type of learning has no supervisor, but instead it has a reward signal that defines success. Similar to human learning, when success is rewarded, the machine tries to learn the patterns that result in receiving the reinforcement signal. The machine’s decisions affect the subsequent data it receives.

Question 13

Reinforcement Learning

Process
Machine not only analyzes data but uses the output to improve efficiency or create new strategies. Learns how to apply a set of rules toward an outcome in the most efficient way.

Example
Game playing bots
War Game Simulations

Limitation
Strategies may not be understandable by humans so may be limited to one situation.

Traditional Statistics Tool
Game Theory
Linear Programming

Answer 13

Reinforcement Learning

Process
Machine not only analyzes data but uses the output to improve efficiency or create new strategies. Learns how to apply a set of rules toward an outcome in the most efficient way.

Example
Game playing bots
War Game Simulations

Limitation
Strategies may not be understandable by humans so may be limited to one situation.

Traditional Statistics Tool
Game Theory
Linear Programming

Question 14

Unsupervised Learning works well if we have little or limited knowledge of the data. The best examples of this application are so-called targeting engines or recommendation engine . When a
supermarket checkout machine issues you a coupon at checkout

Answer 14

Unsupervised Learning works well if we have little or limited knowledge of the data. The best examples of this application are so-called targeting engines or recommendation engine . When a
supermarket checkout machine issues you a coupon at checkout

Question 15

The best examples of reinforcement machine learning are machines that play games. Typically the
machine is taught the rules of the game and given a goal to win.

Answer 15

The best examples of reinforcement machine learning are machines that play games. Typically the
machine is taught the rules of the game and given a goal to win.

Question 16

Big Data is “The record of all interactions with people, institutions, and things recorded and stored digitally.”
Big data, then, is the digital trail left by humans and their connected machines.

Answer 16

Big Data is “The record of all interactions with people, institutions, and things recorded and stored digitally.”
Big data, then, is the digital trail left by humans and their connected machines.

Question 17

The 7 V’s of Big Data

Volume 
Velocity
Variety
Variability
Visualization
Veracity
Value

Answer 17

The 7 V’s of Big Data

Volume 
Velocity
Variety
Variability
Visualization
Veracity
Value

Question 18

Within the Healthcare industry, pharmaceutical syndicated services tracks sales, price, and
distribution of most pharmaceuticals.

Answer 18

Within the Healthcare industry, pharmaceutical syndicated services tracks sales, price, and
distribution of most pharmaceuticals.

Question 19

The majority of pharmaceutical data comes from patient billing and processing at every point of
the drug supply chain.

Answer 19

The majority of pharmaceutical data comes from patient billing and processing at every point of
the drug supply chain.

Question 20

Unlike most CPG categories however, the pharmaceutical sales and
distribution chain involves many government regulations, physicians, other providers, and
insurance companies as mediators of sales to the patient as the ultimate consumer. As we have
said, drug products historically have not been easily tracked using industry standard digital codes
such as the UPC .

Answer 20

Unlike most CPG categories however, the pharmaceutical sales and
distribution chain involves many government regulations, physicians, other providers, and
insurance companies as mediators of sales to the patient as the ultimate consumer. As we have
said, drug products historically have not been easily tracked using industry standard digital codes
such as the UPC .

Question 21

Big Data Researcher Skills
These are the most common skills found on big data analytic teams:

Programming 
Data Manipulation
Exploratory Data Analytics 
Mathematics Statistics 
Business Skills domain Expertise
People Skills communication Skills

Answer 21

Big Data Researcher Skills
These are the most common skills found on big data analytic teams:

Programming 
Data Manipulation
Exploratory Data Analytics 
Mathematics Statistics 
Business Skills domain Expertise
People Skills communication Skills

Question 22

The evolution of big data has produced an entirely new field called data science, an interdisciplinary field about scientific methods, processes, and systems to extract knowledge or insights from data in various forms.

Answer 22

The evolution of big data has produced an entirely new field called data science, an interdisciplinary field about scientific methods, processes, and systems to extract knowledge or insights from data in various forms.

Question 23

In addition to the typical and historical skills of qualitative techniques and traditional statistical skills of quantitative surveys and analytics, modern researchers are concerned with:
Data Curation
Data Governance
Data Provenance

Answer 23

In addition to the typical and historical skills of qualitative techniques and traditional statistical skills of quantitative surveys and analytics, modern researchers are concerned with:
Data Curation
Data Governance
Data Provenance

Question 24

Data Curation

Right data is assembled for the right question

Answer 24

Data Curation

Right data is assembled for the right question

Question 25

Data Governance

Data is secure & accurate

Answer 25

Data Governance

Data is secure & accurate

Question 26

Data Provenance

Data from reputable sources & tracked through all potential uses

Answer 26

Data Provenance

Data from reputable sources & tracked through all potential uses

Question 27

Machine Learning is also being used to understand the return on investment of marketing itself
( MROI ), that is, measuring how much money is generated by investing in marketing

Answer 27

Machine Learning is also being used to understand the return on investment of marketing itself
( MROI ), that is, measuring how much money is generated by investing in marketing

Question 28

The datasets
that track marketing spending are large, have more variables than cases, and contain
relationships that are often non-linear mixed with responses that are well behaved and
straightforward.

Answer 28

The datasets
that track marketing spending are large, have more variables than cases, and contain
relationships that are often non-linear mixed with responses that are well behaved and
straightforward.

Question 29

Prediction and understanding are related but independent goals of market research; It is simply a decision of the business regarding the
problem at hand whether one goal might be favored over the other.

Answer 29

Prediction and understanding are related but independent goals of market research; It is simply a decision of the business regarding the
problem at hand whether one goal might be favored over the other.

Question 30

striking the right balance between prediction and understanding is still required

Answer 30

striking the right balance between prediction and understanding is still required

Question 31

The complexity of big data
and the algorithms that read it have increased the frequency with which predictive models take
precedence as more and more marketing occurs in digital environments where automation is
possible and desirable.

Answer 31

The complexity of big data
and the algorithms that read it have increased the frequency with which predictive models take
precedence as more and more marketing occurs in digital environments where automation is
possible and desirable.

Question 32

data that is collected for any purpose subsequently used in research other than to meet the needs of your
particular study is called “secondary” data.

Answer 32

data that is collected for any purpose subsequently used in research other than to meet the needs of your
particular study is called “secondary” data.

Question 33

secondary data in some detail:
any purpose other than to meet the needs of your particular study.
non-specific research purposes, called “syndicated” or multi-client
data.
another purpose and subsequently used in research.

Answer 33

secondary data in some detail:
any purpose other than to meet the needs of your particular study.
non-specific research purposes, called “syndicated” or multi-client
data.
another purpose and subsequently used in research.

Question 34

The second most important question in your research design, just behind the original purpose of
the research, is “What is already known about your research goal?”

Answer 34

The second most important question in your research design, just behind the original purpose of
the research, is “What is already known about your research goal?”

Question 35

To reiterate and re-emphasize: All market research designs for every research project should
begin with an assessment of what is already known about your research problem. The answer to
that question almost always involves the search for and use of secondary data in its many forms,
whether you are merely searching the Internet or buying needed data from a broker.

Answer 35

To reiterate and re-emphasize: All market research designs for every research project should
begin with an assessment of what is already known about your research problem. The answer to
that question almost always involves the search for and use of secondary data in its many forms,
whether you are merely searching the Internet or buying needed data from a broker.

Question 36

the advantage of each imagined example of secondary research is that it costs
less time, money and effort.

Answer 36

the advantage of each imagined example of secondary research is that it costs
less time, money and effort.

Question 37

No
single organization outside of governments could fund entire population studies such as censuses
or large -scale public health studies, for example

Answer 37

No
single organization outside of governments could fund entire population studies such as censuses
or large -scale public health studies, for example

Question 38

Secondary data may be older than you like or might be an annual estimate rather than the
monthly trend information that would be best for your project; or the best information that you
can nd may be at a regional level and you might need data about a city in the region

Answer 38

Secondary data may be older than you like or might be an annual estimate rather than the
monthly trend information that would be best for your project; or the best information that you
can nd may be at a regional level and you might need data about a city in the region

Question 39

Two key elements to review to assess the potential for bias are the original purpose of the
research or data and the methodology of the original research.

Answer 39

Two key elements to review to assess the potential for bias are the original purpose of the
research or data and the methodology of the original research.

Question 40

If possible, it is always a good idea to evaluate the original sources rather than a summary or an
article describing the research

Answer 40

If possible, it is always a good idea to evaluate the original sources rather than a summary or an
article describing the research

Question 41

The best way to evaluate accuracy is to examine multiple sources of the same estimates

Answer 41

The best way to evaluate accuracy is to examine multiple sources of the same estimates

Question 42

Finally, you should evaluate the credibility and reputation of the source of your data

Answer 42

Finally, you should evaluate the credibility and reputation of the source of your data

Question 43

One part of the syndicated data industry builds these pro les and makes them
available to buyers.

Answer 43

One part of the syndicated data industry builds these pro les and makes them
available to buyers.

Question 44

The bene t from knowing more about your customer is pretty obvious, and there are many ways
to know them. One part of the syndicated data industry builds these pro les and makes them
available to buyers.

Answer 44

The bene t from knowing more about your customer is pretty obvious, and there are many ways
to know them. One part of the syndicated data industry builds these pro les and makes them
available to buyers.

Question 45

Psychographics pertain to people’s values, attitudes, lifestyles , and personalities. These deeper,
more personal and less obviously observed criteria are helpful in understanding consumer
motivation.

Answer 45

Psychographics pertain to people’s values, attitudes, lifestyles , and personalities. These deeper,
more personal and less obviously observed criteria are helpful in understanding consumer
motivation.

Question 46

Demographics: Grouping people by age, gender,
ethnicity, income, education, geographic or other
physical and structural characteristics.

Answer 46

Demographics: Grouping people by age, gender,
ethnicity, income, education, geographic or other
physical and structural characteristics.

Question 47

Behavioral: Grouping people by common behaviors,
e.g., products purchased, websites or stores visited,
shopping behaviors.

Answer 47

Behavioral: Grouping people by common behaviors,
e.g., products purchased, websites or stores visited,
shopping behaviors.

Question 48

Psychographics: Grouping people by shared values,

principles, personality, interests, and lifestyle.

Answer 48

Psychographics: Grouping people by shared values,

principles, personality, interests, and lifestyle.

Question 49

Market Share is one of the oldest and simplest forms of measuring performance and seems so
basic that many do not realize that Arthur Nielsen, Sr. actually invented it as a tracking measure in
his first Nielsen syndicated database in the 1930s

Answer 49

Market Share is one of the oldest and simplest forms of measuring performance and seems so
basic that many do not realize that Arthur Nielsen, Sr. actually invented it as a tracking measure in
his first Nielsen syndicated database in the 1930s

Question 50

Simply tracking the sales of a product relative

to total category sales is still a simple and powerful measure of performance

Answer 50

Simply tracking the sales of a product relative

to total category sales is still a simple and powerful measure of performance

Question 51

A Development Index compares the sales of a product to either population (as sales per capita)
or to total possible ACV in a market.

Answer 51

A Development Index compares the sales of a product to either population (as sales per capita)
or to total possible ACV in a market.

Question 52

The Developmental Indices allow you to quickly spot performance gaps and detect any regional
issues that might result from retail performance, sales personnel performance or even product preference

Answer 52

The Developmental Indices allow you to quickly spot performance gaps and detect any regional
issues that might result from retail performance, sales personnel performance or even product preference

Question 53

High Category Indices show reasonable demand for products similar to ours,
but for some reason, we are not appealing to the market or there is a barrier to our sales that
bears further investigation.

Answer 53

High Category Indices show reasonable demand for products similar to ours,
but for some reason, we are not appealing to the market or there is a barrier to our sales that
bears further investigation.

Question 54

The BDI/CDI chart can help visualize where our new product might
have high potential (High CDI) but face strong competition (High Competitive BDI). When over 50
markets are commonly available, this matrix helps organize and visualize the information.

Answer 54

The BDI/CDI chart can help visualize where our new product might
have high potential (High CDI) but face strong competition (High Competitive BDI). When over 50
markets are commonly available, this matrix helps organize and visualize the information.

Question 55

Advanced analytic techniques can precisely
estimate and predict the effects of the change for every item, product, brand, or category, but
often simple observational techniques and arithmetic can provide much real information to
inform price decisions

Answer 55

Advanced analytic techniques can precisely
estimate and predict the effects of the change for every item, product, brand, or category, but
often simple observational techniques and arithmetic can provide much real information to
inform price decisions

Question 56

Every company generates and retains data as part of

conducting their business and much of it is useful in answering market research questions.

Answer 56

Every company generates and retains data as part of

conducting their business and much of it is useful in answering market research questions.

Question 57

Most modern businesses compile information in customer databases, data warehouses, or
enterprise decision support systems. Increasingly this information is augmented by external
information obtained from sources we will discuss later so that massive amounts of information
are available internally.

Answer 57

Most modern businesses compile information in customer databases, data warehouses, or
enterprise decision support systems. Increasingly this information is augmented by external
information obtained from sources we will discuss later so that massive amounts of information
are available internally.

Question 58

Modern computing power is enabling businesses to track and manage
millions of customers, whether consumers or businesses, in a practice called
Customer Relationship Management (CRM

Answer 58

Modern computing power is enabling businesses to track and manage
millions of customers, whether consumers or businesses, in a practice called
Customer Relationship Management (CRM

Question 59

Data mining is the application of usually automated analytic techniques to nd patterns in data
that can be used to grow a business

Answer 59

Data mining is the application of usually automated analytic techniques to nd patterns in data
that can be used to grow a business

Question 60

two types of

external data: Syndicated Services and Big Data.

Answer 60

two types of

external data: Syndicated Services and Big Data.

Question 61

Technologies that promise to automate and discover new
insights about customers and consumers are transforming the market research industry as well.

Technology and the Internet have now automated many market research processes that replace what was considerable human effort. Data coding, text analysis, sample selection,
and questionnaire management and creation are all examples

Automated facial recognition algorithms can now detect emotional response to advertising
real time.

Advanced data fusion techniques interconnect and link attitudes and opinions from
thousands of surveys that may share only the minimum of actual questionnaire content

Answer 61

Technologies that promise to automate and discover new
insights about customers and consumers are transforming the market research industry as well.

Technology and the Internet have now automated many market research processes that replace what was considerable human effort. Data coding, text analysis, sample selection,
and questionnaire management and creation are all examples

Automated facial recognition algorithms can now detect emotional response to advertising
real time.

Advanced data fusion techniques interconnect and link attitudes and opinions from
thousands of surveys that may share only the minimum of actual questionnaire content

Question 62

So it makes sense that primary and secondary research often work together in the same
project

Answer 62

So it makes sense that primary and secondary research often work together in the same
project

Question 63

The use of multiple sources validates your data and investigation by cross
verifying the same ideas and information. This process is sometimes known as
data triangulation , taken from the navigation process of locating an unknown point in space through geometric relationships among other known points. You can validate by triangulating
data sources, research methods, or even theories

Answer 63

The use of multiple sources validates your data and investigation by cross
verifying the same ideas and information. This process is sometimes known as
data triangulation , taken from the navigation process of locating an unknown point in space through geometric relationships among other known points. You can validate by triangulating
data sources, research methods, or even theories

Question 64

Data triangulation is simply using evidence from many di erent
types of data sources, such as interviews, documents, public records, social media conversations,
or observations.

Answer 64

Data triangulation is simply using evidence from many di erent
types of data sources, such as interviews, documents, public records, social media conversations,
or observations.

Question 65

Theory triangulation is a bit more
complicated because the theories
are helping you understand your
data better rather than the sorts of
integration required in data and
methodology triangulation. Here you
are applying different theories to the
data to help make sense of it. One theory might support your data and
another might undermine it.

Answer 65

Theory triangulation is a bit more
complicated because the theories
are helping you understand your
data better rather than the sorts of
integration required in data and
methodology triangulation. Here you
are applying different theories to the
data to help make sense of it. One theory might support your data and
another might undermine it.

Question 66

It is important to remember when working with secondary data that it often contains personal
data; that is, data that can be used either directly or indirectly (e.g., by combining it with other
data) about a specific individual.

Answer 66

It is important to remember when working with secondary data that it often contains personal
data; that is, data that can be used either directly or indirectly (e.g., by combining it with other
data) about a specific individual.

Question 67

Two key responsibilities that
apply to secondary data:
1. Researchers must ensure that personal data used in research is thoroughly protected from
unauthorized access and never disclosed without the consent of the data subject.
2. Researchers must always behave ethically and not do anything that might cause harm to a
data subject or damage the reputation of market, opinion, and social research.

Answer 67

Two key responsibilities that
apply to secondary data:
1. Researchers must ensure that personal data used in research is thoroughly protected from
unauthorized access and never disclosed without the consent of the data subject.
2. Researchers must always behave ethically and not do anything that might cause harm to a
data subject or damage the reputation of market, opinion, and social research.

Question 68

This data is collected
or purchased by research companies and then the curated data is sold to multiple buyers to help
track or interpret their businesses. The companies collect the data once, but sell it many times to
multiple buyers as a subscription. The research companies also leverage the data for research
products and consulting services that may be customized for their customers.

Answer 68

This data is collected
or purchased by research companies and then the curated data is sold to multiple buyers to help
track or interpret their businesses. The companies collect the data once, but sell it many times to
multiple buyers as a subscription. The research companies also leverage the data for research
products and consulting services that may be customized for their customers.

Question 69

Many research companies syndicate what are called “ omnibus surveys ”, where questions on
many topics are conducted during the same interview

Answer 69

Many research companies syndicate what are called “ omnibus surveys ”, where questions on
many topics are conducted during the same interview

Question 70

By keeping track of unique but
anonymous identi ers and using advanced data fusion techniques, CivicScience builds a massive
database of opinions, preferences, attitudes, and demographic information that it provides to its
customers by subscription.

Answer 70

By keeping track of unique but
anonymous identi ers and using advanced data fusion techniques, CivicScience builds a massive
database of opinions, preferences, attitudes, and demographic information that it provides to its
customers by subscription.

Question 71

Household Panel Sales Data

The types of syndicated sales tracking data discussed so far are very accurate estimates of sales
and share in a market, but they hide some very important dynamics of purchasing necessary to
understand trends, diagnose performance, and grow products with the right marketing plans

Answer 71

Household Panel Sales Data

The types of syndicated sales tracking data discussed so far are very accurate estimates of sales
and share in a market, but they hide some very important dynamics of purchasing necessary to
understand trends, diagnose performance, and grow products with the right marketing plans

Question 72

Market Measurement

What the business does not know, however, is how much its competitors are selling, and the need
for that knowledge drives the largest scale type of syndicated data: sales tracking for consumer
goods.

Answer 72

Market Measurement

What the business does not know, however, is how much its competitors are selling, and the need
for that knowledge drives the largest scale type of syndicated data: sales tracking for consumer
goods.

Question 73

The markets for consumer package goods (CPG) known as Fast Moving Consumer Goods
(FMCG) outside the United States, is particularly well suited for this kind of tracking. The
distribution channels are well known, products are identi ed easily, and there are lots of CPG
businesses that can pay for the information and support the costs of collecting the data.

Answer 73

The markets for consumer package goods (CPG) known as Fast Moving Consumer Goods
(FMCG) outside the United States, is particularly well suited for this kind of tracking. The
distribution channels are well known, products are identi ed easily, and there are lots of CPG
businesses that can pay for the information and support the costs of collecting the data.

Question 74

The Universal Product Code or UPC, developed in the early 1970s to
help automate inventory tracking, proved also to be an essential
factor in automating collection of market information and almost
instantly ampli ed our ability to understand and optimize price,
promotion, and distribution.

Answer 74

The Universal Product Code or UPC, developed in the early 1970s to
help automate inventory tracking, proved also to be an essential
factor in automating collection of market information and almost
instantly ampli ed our ability to understand and optimize price,
promotion, and distribution.

Question 75

The UPC scanner that allowed quick check-out and payment at the supermarket also
automatically entered each transaction into an electronic database that grew with each
transaction

Answer 75

The UPC scanner that allowed quick check-out and payment at the supermarket also
automatically entered each transaction into an electronic database that grew with each
transaction

Question 76

By the late 1970s, a research company named Information Resources Incorporated (IRI) realized
the potential for using UPC’s as the basis for automated and powerful data collection for market
research.

Answer 76

By the late 1970s, a research company named Information Resources Incorporated (IRI) realized
the potential for using UPC’s as the basis for automated and powerful data collection for market
research.

Question 77

The basic structure of all syndicated scanner databases
is similar no matter who builds or sells it. Databases
consist of five essential elements:
Product, Markets, class of Trade, Time Periods and measures

Answer 77

The basic structure of all syndicated scanner databases
is similar no matter who builds or sells it. Databases
consist of five essential elements:
Product, Markets, class of Trade, Time Periods and measures

Question 78

The basic measure of audience size is the audience rating .

Answer 78

The basic measure of audience size is the audience rating .

Question 79

rating is de ned
as the number of households with their radio/TV sets tuned to a particular station/channel or
program for a speci ed length of time divided by the total number of households that have
radio/TV.

Answer 79

rating is de ned
as the number of households with their radio/TV sets tuned to a particular station/channel or
program for a speci ed length of time divided by the total number of households that have
radio/TV.

Question 80

The
other fundamental measure is audience share , which is the number of TV sets in use tuned to a
program or commercial.

Answer 80

The
other fundamental measure is audience share , which is the number of TV sets in use tuned to a
program or commercial.

Question 81

Nielsen has had to
expand its service to cover this now very fragmented audience and other research companies
have stepped up to o er di erent versions of so-called “ cross-platform ” audience measurement.

Answer 81

Nielsen has had to
expand its service to cover this now very fragmented audience and other research companies
have stepped up to o er di erent versions of so-called “ cross-platform ” audience measurement.

Question 82

atistically, when a sample cannot or does
not measure something that is real, it is called sampling error . Sampling error happens when a
sample is not representative of a population or is not large enough to accurately measure a
population condition.

Answer 82

atistically, when a sample cannot or does
not measure something that is real, it is called sampling error . Sampling error happens when a
sample is not representative of a population or is not large enough to accurately measure a
population condition.

Question 83

Although the datasets are
large, they may not be
representative, and therefore
not projectable to wider
situations
The datasets build so fast they
may accumulate incorrect
information. Sometimes the underlying process behind the accumulation of data makes the data in the
past less predictive of the future. A predictive process that works well enough once may not
work in the future
Sometimes the variables of big data are not the things that are closest to what we want to
measure directly

Answer 83

Although the datasets are
large, they may not be
representative, and therefore
not projectable to wider
situations
The datasets build so fast they
may accumulate incorrect
information. Sometimes the underlying process behind the accumulation of data makes the data in the
past less predictive of the future. A predictive process that works well enough once may not
work in the future
Sometimes the variables of big data are not the things that are closest to what we want to
measure directly

Question 84

We are going to focus on one
type of this research Sentiment Analysis , as an example, because of its popularity and promise.
We will also see how big data can supplement and amplify primary research, and at least
speculate on the value of the sensor data from connected devices as a secondary research
application.

Answer 84

We are going to focus on one
type of this research Sentiment Analysis , as an example, because of its popularity and promise.
We will also see how big data can supplement and amplify primary research, and at least
speculate on the value of the sensor data from connected devices as a secondary research
application.

Question 85

Sentiment Analysis tries to analytically determine the attitude of a speaker or writer with respect
to some topic. The attitude may be an opinion, or an intended or actual emotional state.

Answer 85

Sentiment Analysis tries to analytically determine the attitude of a speaker or writer with respect
to some topic. The attitude may be an opinion, or an intended or actual emotional state.

Question 86

Computers read words through a process called Natural Language Processing (NLP) and they
infer sentiment by a combination of NLP and machine learning. Both NLP and sentiment analysis
are very complex computer processes to accomplish a task that seems obvious to humans, but
the sheer scale of the data requires a machine. New algorithms are continuously improving our
ability to accurately read this data in a field of computer science that is relatively young.

Answer 86

Computers read words through a process called Natural Language Processing (NLP) and they
infer sentiment by a combination of NLP and machine learning. Both NLP and sentiment analysis
are very complex computer processes to accomplish a task that seems obvious to humans, but
the sheer scale of the data requires a machine. New algorithms are continuously improving our
ability to accurately read this data in a field of computer science that is relatively young.

Question 87

Three Sentiment Analysis Approaches
Knowledge Based
Statistically Based
Hybrid

Answer 87

Three Sentiment Analysis Approaches
Knowledge Based
Statistically Based
Hybrid

Question 88

Knowledge Based

Pre-classify text by categories based on the
presence of unambiguous words such as
happy, sad, angry, or bored. Assign scores to
more ambiguous terms based on previous
search

Compare analysis text to
pre-scored reference
database

Answer 88

Knowledge Based

Pre-classify text by categories based on the
presence of unambiguous words such as
happy, sad, angry, or bored. Assign scores to
more ambiguous terms based on previous
search

Compare analysis text to
pre-scored reference
database

Question 89

Statistically Based

Use machine learning to train algorithms
against known outcomes. For example, a
machine can sift through reviews and learn
which words in the reviews are associated with
accompanying star ratings

Compare analysis text to a
reference database built
from the statistical analysis

Answer 89

Statistically Based

Use machine learning to train algorithms
against known outcomes. For example, a
machine can sift through reviews and learn
which words in the reviews are associated with
accompanying star ratings

Compare analysis text to a
reference database built
from the statistical analysis

Question 90

Hybrid

Both methods combined, for example pre-
coding unambiguous words and modeling
more ambiguous terms against star ratings

Compare analysis text to a
reference database

Answer 90

Hybrid

Both methods combined, for example pre-
coding unambiguous words and modeling
more ambiguous terms against star ratings

Compare analysis text to a
reference database

Question 91

Given a target subject, then, sentiment analysis can analyze an enormous amount of text and
classify whether the text is generally positive, negative, or neutral about the subject. The best
techniques are pretty good at this task relative to human classi cation. Many studies have shown
that there is generally about 80% agreement between human readers in classifying the same text.
The best machines analysis achieves about 70%. That level is pretty good compared to 80%.

Answer 91

Given a target subject, then, sentiment analysis can analyze an enormous amount of text and
classify whether the text is generally positive, negative, or neutral about the subject. The best
techniques are pretty good at this task relative to human classi cation. Many studies have shown
that there is generally about 80% agreement between human readers in classifying the same text.
The best machines analysis achieves about 70%. That level is pretty good compared to 80%.

Question 92

Machines currently are not as accurate when
assessing finer degrees of positivity or negativity, as in the difference between a 4- and a 5-star
rating in a review.

Answer 92

Machines currently are not as accurate when
assessing finer degrees of positivity or negativity, as in the difference between a 4- and a 5-star
rating in a review.

Question 93

Probably the biggest limitation of sentiment analysis is that all social listening data, no matter how
much of it exists, is essentially qualitative. Most comments are by de nition non-representative.
People passionate enough to comment generate these comments, and they are people with
access to social media. The large number of comments does not change this inherent fact.

Answer 93

Probably the biggest limitation of sentiment analysis is that all social listening data, no matter how
much of it exists, is essentially qualitative. Most comments are by de nition non-representative.
People passionate enough to comment generate these comments, and they are people with
access to social media. The large number of comments does not change this inherent fact.

Question 94

Internet of Things analysis is not yet a common form of data

for marketing organizations. It is being tested and we are all learning.

Answer 94

Internet of Things analysis is not yet a common form of data

for marketing organizations. It is being tested and we are all learning.