BIG DATA FOR MARKETING
ANALYTICS
Tilburg University
2024-2025
MINOR MARKETING
,WEB CLIPS MODULE 1: INTRO & CHARACTERISTICS OF BIG DATA
What are big data made of? -> Digital traces (exhaust) = A record created and stored of
some behavior
è For example a click on a website, like or share, watch, text
Big data:
Ø Explosion of the informaUon storage and compuUng power (used to analyze the
data), it becomes larger and larger
Two noUons of big:
1. Columns / width / dimensions = variables: total # of variables = p
2. Rows / length = observa0ons: total # of observaUons = n
Ø Tall data (n >> p) = Many observaUons, relaUvely few variables
Ø Wide data (n << p) = Few observaUons, many variables
è These noUons tend to go together, you need lots of observaUons to learn about
many variables
Big data:
- Early definiUon = Too large to be loaded into one machine, ‘distributed-data big’
(data had to be distributed across different computers), domain of computer
engineering
- More our focus: Big data is a shorthand label that typically means applying the tools
of arUficial intelligence, like machine learning, to vast new troves of data beyond that
captured in standard databases.
Primary data (custommade) -> Collected to answer a specific research quesUon
Secondary data (readymade) -> Collected for non-research purpose
Types of business data:
Structured Unstructured
Internal Human- generated (survey raUngs) Human- generated (emails, voicemails,
Machine-generated (web matrics leaers)
External Human generated (raUngs on yelp) Human generated (comments on yelp,
Machine generated (Ume of tweet) comments in online forums)
Structured data: Data with a clear scale such that it can immediately be analyzed
Unstructured data: No preplanned scale, not immediately quanUfiable -> 80% of data is
unstructured
Internal: Created within the firm
External: Created outside the firm
1
,Uses of big data:
1. Personaliza0on -> Neelix provides their members with personalized suggesUons
(automated recommendaUons) based on the big data of customers preferences
2. Boos0ng engagement -> Facebook tests different versions to see when customers
are more engaged. With primary data they chose the best layout.
3. New product development -> Big data is always on, look at social media etc. to
develop new products
4. Reducing churn -> Customer churn = customer quits some service, a firm wants to
reduce this. Use past data to esUmate model that predicts churn and use that model
to predict probability of churn on current customers. A firm can intervene on those
most likely to churn
5. Public policy and economy -> Mobile phones has become a primary source of public
data intelligence
10 characterisUcs of big data:
è Advantages:
1. Big
Ø When the event is rare of small. For example when analyzing click
through rate, because the average CTR is 0.35%.
Ø When there is heterogeneity (= customers respond differently to the
same thing), we can make a disUncUon between different types of
customers
Ø When the relaUonship is complex -> A firm may be able to predict beaer
when a customer is likely to quit by designing a heat map
2. Always on -> The data is being collected 24/7. It is important when we need to
know and respond to answers quickly. For example trend spojng, monitoring
compeUUon and economic acUvity
3. Nonreac0ve -> With big data users are typically not aware they are being
recorded. In other situaUons, people usually change their behavior when they
know they are being observed, but with big data this is thus not the case.
è Disadvantages:
4. Incomplete -> It records what happened, but not why. It predicts for example
that customers will quit a service, but not why they quit.
5. Inaccessible
Ø From outside the organizaUon: Legal, business or ethical barriers to giving
outside researchers access to data.
Ø From inside: Databases are not integrated, lacking variables to match,
different coding schemes. A customer journey consists of a lot of different
touchpoints, but it can be difficult to link the data from different
databases that come from the same user
6. Nonrepresenta0ve -> If your sample is representaUve, you can make inferences
about the populaUon based on your sample. You have to take in account that not
everyone uses a plaeorm, so you can’t generalize it.
Ø Underlying opinion is different from the expressed opinion, only posUng
your opinion when it is really posiUve or negaUve. The expressed opinion
is not representaUve for the underlying opinion
2
, 7. DriLing -> The measure is changing (if you want to measure change, don’t
change the measure). Big data sources can suffer from problems, because:
Ø The users can change
Ø How they use it can change
Ø The plaeorm itself changes
8. Algorithmically confounded -> How the plaeorm is designed can influence
behavior, introducing bias or noise into what you’re trying to study. For example
Facebook makes recommendaUons to become friends with friends of friends.
9. Dirty -> Big data sources can be loaded with junk or spam.
10. Sensi0ve -> Some of the informaUon that companies have is sensiUve. Under the
GDPR, Cookies that are not strictly necessary for the basic funcUon of your
website must only be acUvated aqer your end-users have given their explicit
consent to the specific purpose of their operaUon and collecUon of personal data
IN CLASS SLIDES MODULE 1
Big data = Large, diverse, complex, longitudinal and / or distributed data sets generated
from instruments, sensors, internet transacUons, e-mail, video, click stream and / or all
other digital sources available today and in the future (White House)
è Shorthand term that refers to several things: Size of data, nature of the collecUon
process, type of data, purpose-built or data exhaust, passively or acUvely collected
è Boaom line: No precise definiUon for big data
Analy0cs = Discipline that applies logic and mathemaUcs to data to provide insights for
making beaer decisions
Bias = How far the esUmate is off from the truth on average
Variance = How much noise or uncertainty there is in the esUmates
è Roughly add these up to create overall measure: Mean squared error
3
, Simple random sampling:
- Each sample of size n from a populaUon of size N has the same probability of
occurrence (sample is representaUve of populaUon and unbiased)
- As you increase the sample size, the variance goes down
Formula’s:
𝑠𝑑
𝐶𝑜𝑛𝑓𝑖𝑑𝑒𝑛𝑐𝑒 𝑖𝑛𝑡𝑒𝑟𝑣𝑎𝑙 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑚𝑒𝑎𝑛: 𝑥̅ ± 1.96
√𝑛
(1 − 𝑓)
𝑀𝑒𝑎𝑛 𝑠𝑞𝑢𝑎𝑟𝑒𝑑 𝑒𝑟𝑟𝑜𝑟: 𝑀𝑆𝐸 = 𝐸 [(𝑋D − 𝜇)! ] = 𝐸I𝜌",$
!
K× × 𝜎$!
𝑓
READINGS
1.2 Welcome to digital age
• Researchers must combine tradiUonal methods with modern data science
• Internet of things
2.2 Big data
• Big data is characterized by its volume, variety and velocity
• Big data were not originally designed for research purposes
• Repurposing big data requires know of context behind data
Ø Social scienUsts: Focus on limitaUons
Ø Data scienUsts: Focus on advantages
2.3 Characteris0cs
• Big: Can enable specific kinds of research, like rare events of heterogeneity across
different regions. Increases the need to consider how data was generated
• Always on: Enables study of unexpected events and supports real-Ume measurement
• Incomplete: 3 forms -> 1. Missing demographic data 2. Missing behavior across other
plaeorms 3. Missing data to operaUonalize theoreUcal constructs
• NonrepresentaUve: Can be useful for within sample comparisons but poses
challenges for broader generalizaUons
• Driqing: Making it difficult to study long-term trends. Three main forms: 1.
PopulaUon driq (who uses the system) 2. Behavioral driq (changes in how people use
the system) 3.System driq
• Dirty: Cleaning data is challenging because it is oqen not collected with research in
mind
SELF TEST 1
Example of inaccessibility: Data is not integrated across systems and disjointed
4
ANALYTICS
Tilburg University
2024-2025
MINOR MARKETING
,WEB CLIPS MODULE 1: INTRO & CHARACTERISTICS OF BIG DATA
What are big data made of? -> Digital traces (exhaust) = A record created and stored of
some behavior
è For example a click on a website, like or share, watch, text
Big data:
Ø Explosion of the informaUon storage and compuUng power (used to analyze the
data), it becomes larger and larger
Two noUons of big:
1. Columns / width / dimensions = variables: total # of variables = p
2. Rows / length = observa0ons: total # of observaUons = n
Ø Tall data (n >> p) = Many observaUons, relaUvely few variables
Ø Wide data (n << p) = Few observaUons, many variables
è These noUons tend to go together, you need lots of observaUons to learn about
many variables
Big data:
- Early definiUon = Too large to be loaded into one machine, ‘distributed-data big’
(data had to be distributed across different computers), domain of computer
engineering
- More our focus: Big data is a shorthand label that typically means applying the tools
of arUficial intelligence, like machine learning, to vast new troves of data beyond that
captured in standard databases.
Primary data (custommade) -> Collected to answer a specific research quesUon
Secondary data (readymade) -> Collected for non-research purpose
Types of business data:
Structured Unstructured
Internal Human- generated (survey raUngs) Human- generated (emails, voicemails,
Machine-generated (web matrics leaers)
External Human generated (raUngs on yelp) Human generated (comments on yelp,
Machine generated (Ume of tweet) comments in online forums)
Structured data: Data with a clear scale such that it can immediately be analyzed
Unstructured data: No preplanned scale, not immediately quanUfiable -> 80% of data is
unstructured
Internal: Created within the firm
External: Created outside the firm
1
,Uses of big data:
1. Personaliza0on -> Neelix provides their members with personalized suggesUons
(automated recommendaUons) based on the big data of customers preferences
2. Boos0ng engagement -> Facebook tests different versions to see when customers
are more engaged. With primary data they chose the best layout.
3. New product development -> Big data is always on, look at social media etc. to
develop new products
4. Reducing churn -> Customer churn = customer quits some service, a firm wants to
reduce this. Use past data to esUmate model that predicts churn and use that model
to predict probability of churn on current customers. A firm can intervene on those
most likely to churn
5. Public policy and economy -> Mobile phones has become a primary source of public
data intelligence
10 characterisUcs of big data:
è Advantages:
1. Big
Ø When the event is rare of small. For example when analyzing click
through rate, because the average CTR is 0.35%.
Ø When there is heterogeneity (= customers respond differently to the
same thing), we can make a disUncUon between different types of
customers
Ø When the relaUonship is complex -> A firm may be able to predict beaer
when a customer is likely to quit by designing a heat map
2. Always on -> The data is being collected 24/7. It is important when we need to
know and respond to answers quickly. For example trend spojng, monitoring
compeUUon and economic acUvity
3. Nonreac0ve -> With big data users are typically not aware they are being
recorded. In other situaUons, people usually change their behavior when they
know they are being observed, but with big data this is thus not the case.
è Disadvantages:
4. Incomplete -> It records what happened, but not why. It predicts for example
that customers will quit a service, but not why they quit.
5. Inaccessible
Ø From outside the organizaUon: Legal, business or ethical barriers to giving
outside researchers access to data.
Ø From inside: Databases are not integrated, lacking variables to match,
different coding schemes. A customer journey consists of a lot of different
touchpoints, but it can be difficult to link the data from different
databases that come from the same user
6. Nonrepresenta0ve -> If your sample is representaUve, you can make inferences
about the populaUon based on your sample. You have to take in account that not
everyone uses a plaeorm, so you can’t generalize it.
Ø Underlying opinion is different from the expressed opinion, only posUng
your opinion when it is really posiUve or negaUve. The expressed opinion
is not representaUve for the underlying opinion
2
, 7. DriLing -> The measure is changing (if you want to measure change, don’t
change the measure). Big data sources can suffer from problems, because:
Ø The users can change
Ø How they use it can change
Ø The plaeorm itself changes
8. Algorithmically confounded -> How the plaeorm is designed can influence
behavior, introducing bias or noise into what you’re trying to study. For example
Facebook makes recommendaUons to become friends with friends of friends.
9. Dirty -> Big data sources can be loaded with junk or spam.
10. Sensi0ve -> Some of the informaUon that companies have is sensiUve. Under the
GDPR, Cookies that are not strictly necessary for the basic funcUon of your
website must only be acUvated aqer your end-users have given their explicit
consent to the specific purpose of their operaUon and collecUon of personal data
IN CLASS SLIDES MODULE 1
Big data = Large, diverse, complex, longitudinal and / or distributed data sets generated
from instruments, sensors, internet transacUons, e-mail, video, click stream and / or all
other digital sources available today and in the future (White House)
è Shorthand term that refers to several things: Size of data, nature of the collecUon
process, type of data, purpose-built or data exhaust, passively or acUvely collected
è Boaom line: No precise definiUon for big data
Analy0cs = Discipline that applies logic and mathemaUcs to data to provide insights for
making beaer decisions
Bias = How far the esUmate is off from the truth on average
Variance = How much noise or uncertainty there is in the esUmates
è Roughly add these up to create overall measure: Mean squared error
3
, Simple random sampling:
- Each sample of size n from a populaUon of size N has the same probability of
occurrence (sample is representaUve of populaUon and unbiased)
- As you increase the sample size, the variance goes down
Formula’s:
𝑠𝑑
𝐶𝑜𝑛𝑓𝑖𝑑𝑒𝑛𝑐𝑒 𝑖𝑛𝑡𝑒𝑟𝑣𝑎𝑙 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑚𝑒𝑎𝑛: 𝑥̅ ± 1.96
√𝑛
(1 − 𝑓)
𝑀𝑒𝑎𝑛 𝑠𝑞𝑢𝑎𝑟𝑒𝑑 𝑒𝑟𝑟𝑜𝑟: 𝑀𝑆𝐸 = 𝐸 [(𝑋D − 𝜇)! ] = 𝐸I𝜌",$
!
K× × 𝜎$!
𝑓
READINGS
1.2 Welcome to digital age
• Researchers must combine tradiUonal methods with modern data science
• Internet of things
2.2 Big data
• Big data is characterized by its volume, variety and velocity
• Big data were not originally designed for research purposes
• Repurposing big data requires know of context behind data
Ø Social scienUsts: Focus on limitaUons
Ø Data scienUsts: Focus on advantages
2.3 Characteris0cs
• Big: Can enable specific kinds of research, like rare events of heterogeneity across
different regions. Increases the need to consider how data was generated
• Always on: Enables study of unexpected events and supports real-Ume measurement
• Incomplete: 3 forms -> 1. Missing demographic data 2. Missing behavior across other
plaeorms 3. Missing data to operaUonalize theoreUcal constructs
• NonrepresentaUve: Can be useful for within sample comparisons but poses
challenges for broader generalizaUons
• Driqing: Making it difficult to study long-term trends. Three main forms: 1.
PopulaUon driq (who uses the system) 2. Behavioral driq (changes in how people use
the system) 3.System driq
• Dirty: Cleaning data is challenging because it is oqen not collected with research in
mind
SELF TEST 1
Example of inaccessibility: Data is not integrated across systems and disjointed
4
