uva msc business administration: digital marketing track | semester II - period IV (2023-2024) [by gycc]
MARKETING METRICS
[6314M0343Y] - lecture notes
Week 1-2: product design
Week 2-4: predict
Week 4-6: promotion and targeting
2024.02.06 | WEEK 1 | CONSUMER PREFERENCES
Before designing a new product – the need to first understand your customers:
● Goals / needs / preferences
○ Underlying motivation – what is it that people want
○ Psychological needs (e.g. the social status associated with products)
○ Functional needs (e.g. does it fulfil its intended purpose?)
● How do they trade off these preferences for money?
○ Would they rather keep their money or buy an alternative product?
○ Using a scale to rate the possible range wherein customers are willing to buy
(this course measures the value that customers give to its products after consumption)
(various ways to measure customer value, e.g. constrained measure: CONJOINT ANALYSIS)
1
, uva msc business administration: digital marketing track | semester II - period IV (2023-2024) [by gycc]
A historical perspective.
Creator of conjoint analysis – Thurstone
● Presented the idea to measure consumer preferences in 1931
● “Perhaps the simplest experimental method that comes to mind is to ask a subject to fill in the
blank space in a series of choices”
● E.g. eight hats and eight pairs of shoes VS. six hats and ___ pairs of shoes
● NOTE: in this context, the combination eight hats and eight pairs of shoes is chosen as the
standard, and each other combination is compared directly with it
Heavy criticism from the academic crowd:
● Opposed to the idea that non-market (or non-actual; hypothetical analysis) behaviour could
teach us something about consumers
● Only from the 1960s onwards, academics really continued to try to extract valuable info from
stated preferences (aka they accepted Thurstone’s concept)
(people often say they would buy sth when in reality they often mispredict their behaviour)
2
, uva msc business administration: digital marketing track | semester II - period IV (2023-2024) [by gycc]
“Is there any value in asking people how they want an existing OR new product to look like?”
● Consumers face different types of decisions daily: consumers are good at making trade-offs
(decisions) between benefits and costs
● CONJOINT ANALYSIS → a survey-based statistical technique designed for consumers to
make trade-offs, forcing them to reveal their true preferences
○ Producing a mathematical system of their preferences
○ Conjoint analysis reveals what people really want
○ Evaluation of products (the need to make trade-offs)
○ Product → (from a very functional perspective: a collection of attributes)
Example of different measures – scanners.
In conjoint,
● All products are considered a bundle of attributes
● Examples of attributes for a new phone
○ E.g. size of phone, cameras, CPU, price, colour
● Examples of levels of attributes
○ Price: low, medium, high
○ Camera: 12 megapixels, 32 megapixels
○ Size of phone: 5.8 inch, 6.5 inch, 7.2 inch
Why does conjoint work?
Respondents evaluate bundles!!
● Relate the rating or choice
on the attribute levels
○ Decomposing the ‘overall utility’ in partworths
○ Uncovering the relative importance of each attribute
3
, uva msc business administration: digital marketing track | semester II - period IV (2023-2024) [by gycc]
CONJOINT ANALYSIS: the process
(through the example application: pizza)
Step 1. Attributes
⇒ how to decide on which attributes to include
● Through conducting focus groups, interviews with the new product design team, analysis of
the current market, pre-tests
● Make sure to not pick too many attributes
○ Pick the most important ones
○ Consumers may get confused with too many
● Attributes = what you can customise/alternate
● E.g. pizza
○ Toppings / waiting time / crust / price / cheese type or amount
Step 2. Attribute levels
⇒ always need to cover the whole range of attribute levels
● E.g. price ranging from low to high (low/medium/high)
● E.g. is there pepperoni on the pizza? (yes/no)
● Best to limit the # of levels
● Best to keep the # of levels roughly the same for different attributes
○ It is more useful to keep your levels consistent/similar
○ This would lead to more natural responses and less biassed answers (e.g. focusing
more on toppings as there are more choices to consider)
Step 3. Profiles
⇒ developing (hypothetical) profiles – factorial design: all possible combinations of all levels and
alternatives of the created attributes
● In this example, 3x3x3x4x3 = 324 theoretical bundles
● The need to carefully select profiles and as little as possible,
while also maximising the results/information
4
MARKETING METRICS
[6314M0343Y] - lecture notes
Week 1-2: product design
Week 2-4: predict
Week 4-6: promotion and targeting
2024.02.06 | WEEK 1 | CONSUMER PREFERENCES
Before designing a new product – the need to first understand your customers:
● Goals / needs / preferences
○ Underlying motivation – what is it that people want
○ Psychological needs (e.g. the social status associated with products)
○ Functional needs (e.g. does it fulfil its intended purpose?)
● How do they trade off these preferences for money?
○ Would they rather keep their money or buy an alternative product?
○ Using a scale to rate the possible range wherein customers are willing to buy
(this course measures the value that customers give to its products after consumption)
(various ways to measure customer value, e.g. constrained measure: CONJOINT ANALYSIS)
1
, uva msc business administration: digital marketing track | semester II - period IV (2023-2024) [by gycc]
A historical perspective.
Creator of conjoint analysis – Thurstone
● Presented the idea to measure consumer preferences in 1931
● “Perhaps the simplest experimental method that comes to mind is to ask a subject to fill in the
blank space in a series of choices”
● E.g. eight hats and eight pairs of shoes VS. six hats and ___ pairs of shoes
● NOTE: in this context, the combination eight hats and eight pairs of shoes is chosen as the
standard, and each other combination is compared directly with it
Heavy criticism from the academic crowd:
● Opposed to the idea that non-market (or non-actual; hypothetical analysis) behaviour could
teach us something about consumers
● Only from the 1960s onwards, academics really continued to try to extract valuable info from
stated preferences (aka they accepted Thurstone’s concept)
(people often say they would buy sth when in reality they often mispredict their behaviour)
2
, uva msc business administration: digital marketing track | semester II - period IV (2023-2024) [by gycc]
“Is there any value in asking people how they want an existing OR new product to look like?”
● Consumers face different types of decisions daily: consumers are good at making trade-offs
(decisions) between benefits and costs
● CONJOINT ANALYSIS → a survey-based statistical technique designed for consumers to
make trade-offs, forcing them to reveal their true preferences
○ Producing a mathematical system of their preferences
○ Conjoint analysis reveals what people really want
○ Evaluation of products (the need to make trade-offs)
○ Product → (from a very functional perspective: a collection of attributes)
Example of different measures – scanners.
In conjoint,
● All products are considered a bundle of attributes
● Examples of attributes for a new phone
○ E.g. size of phone, cameras, CPU, price, colour
● Examples of levels of attributes
○ Price: low, medium, high
○ Camera: 12 megapixels, 32 megapixels
○ Size of phone: 5.8 inch, 6.5 inch, 7.2 inch
Why does conjoint work?
Respondents evaluate bundles!!
● Relate the rating or choice
on the attribute levels
○ Decomposing the ‘overall utility’ in partworths
○ Uncovering the relative importance of each attribute
3
, uva msc business administration: digital marketing track | semester II - period IV (2023-2024) [by gycc]
CONJOINT ANALYSIS: the process
(through the example application: pizza)
Step 1. Attributes
⇒ how to decide on which attributes to include
● Through conducting focus groups, interviews with the new product design team, analysis of
the current market, pre-tests
● Make sure to not pick too many attributes
○ Pick the most important ones
○ Consumers may get confused with too many
● Attributes = what you can customise/alternate
● E.g. pizza
○ Toppings / waiting time / crust / price / cheese type or amount
Step 2. Attribute levels
⇒ always need to cover the whole range of attribute levels
● E.g. price ranging from low to high (low/medium/high)
● E.g. is there pepperoni on the pizza? (yes/no)
● Best to limit the # of levels
● Best to keep the # of levels roughly the same for different attributes
○ It is more useful to keep your levels consistent/similar
○ This would lead to more natural responses and less biassed answers (e.g. focusing
more on toppings as there are more choices to consider)
Step 3. Profiles
⇒ developing (hypothetical) profiles – factorial design: all possible combinations of all levels and
alternatives of the created attributes
● In this example, 3x3x3x4x3 = 324 theoretical bundles
● The need to carefully select profiles and as little as possible,
while also maximising the results/information
4
