Customer marketing and analytics
Lecture 1
Marketing decision problem
Asks what the decision-maker needs to do/action oriented/focuses on symptoms
What logo design do we use?
Marketing research problem
Asks what information is needed and how it can be best obtained/information oriented/focuses on the
underlying causes
How much do people like the different logos?
Qualitative research
Not concerned with numbers
Quantitative research
Focus on numbers, amendable to statistical analysis
Research design
Exploratory: major emphasis is on gaining ideas and insights -> qualitative
Descriptive: often guided by initial hypothesis -> quantitative
Causal: research in which the major emphasis is on determining a cause-and-effect relationship ->
Quantitative and primary data
- Descriptive research reveals associations between variables
- Causal research reveals associations between changes in variables
Secondary data: data previously collected for purposes other than the research at hand
- Syndicated research: large-scale marketing research that is undertaken by a research firm to
be sold, often on a subscription basis, to a number of clients
Primary data: data collected specifically to answer the questions posed by the current research
objectives
Lecture 2
Univariate statistics (one variable):
- Describe and summarize data – a complete run down analysis of all the variables in your
dataset one-at-a-time
- Inferential analysis – learning about the world
Bivariate statistics (two variables):
- Differential analysis
- Associative analysis
Screening the dataset
1. Check for missing data
a. Assign missing values
b. Delete missing values
i. Listwise -> if a person has a missing value for any variable, then they are
excluded from the whole analysis
ii. Pairwise -> a person’s data is excluded only for analyses for which she has
missing data
2. Find strange codes and errors (consistency checks, analyzing frequencies and plots, reverse
coding)
,Nominal: assigns numbers to identify subjects or objects, nothing is implied by the numbers other
than identification (student number, gender, region etc.)
Ordinal: ranking of objects. Numbers indicate relative positions, but amount of difference between
numbers is unknown. (Preference of brands)
Interval: order is important, and distances are known and fixed. No meaningful zero point. (Likert
scales, attitudes, liking, satisfaction, temperature)
Ratio: highest order of number. Contains most information, order is important, distance is identical,
zero is meaningful. (Weight, age, household size, units sold)
Univariate analysis -> bivariate analysis -> multivariate analysis
Univariate analysis – descriptive analysis
- What is the average age of our customer base?
- What percentage of our customers is female?
- What is the average liking of our product?
- How variable are people’s liking ratings?
One question at a time
Nominal and ordinal -> frequency tables/bar charts/pie charts
Interval and ratio -> central tendency or measure for dispersion/histogram
Measures of central tendency:
- Mean -> interval and ratio data
- Median: middle number when data is sorted in size order -> ordinal, interval and ratio
- Mode: the most frequent number -> all data
Measures of variability/dispersion:
- Standard deviation: variance
- Variance: (actual number – sample average)^2 / n
- Range
Measures of shape:
- Skewness: indicator of distribution symmetry
- Kurtosis: indicator of flatness or peakness of distribution
2
, Sample error/standard error: the standard deviation of sample means
Inferential statistics
Generalize sample results to a population
Central limit theorem (CLT): the distribution of sample means approximates a normal distribution
as the sample size gets larger, regardless of the population’s distribution. As n increases, the
distribution of the sample mean approaches the normal distribution.
Confidence interval: interval that, at a specified confidence level, includes the true population value
- 90% confidence interval > sample mean +/- 1.64 * SE
- 95% confidence interval -> sample mean +/- 1.96 * SE
- 99% confidence interval > sample mean +/- 2.58 * SE
“We are 95% sure that the true (unknown)
population mean is contained in the interval
42.91-47.53”
Two-sided tests
- H0: mu = 40
- H1: mu 40
One-sided tests
- H0: mu smaller or equal to 40
- H1: mu is bigger than 40
Two independent samples
- T test independent samples
- One way anova
Independent samples: compare one variable between different samples
Related samples: compare 2 or more variables within respondents
The larger the test statistic and the smaller the p-value, the less likely is H0!
Two methods to decide whether to reject the null hypothesis:
- Test statistic >2
- P-value <5%
Lecture 3
Latent constructs
Latent constructs: less objective and more
abstract constructs
3
Lecture 1
Marketing decision problem
Asks what the decision-maker needs to do/action oriented/focuses on symptoms
What logo design do we use?
Marketing research problem
Asks what information is needed and how it can be best obtained/information oriented/focuses on the
underlying causes
How much do people like the different logos?
Qualitative research
Not concerned with numbers
Quantitative research
Focus on numbers, amendable to statistical analysis
Research design
Exploratory: major emphasis is on gaining ideas and insights -> qualitative
Descriptive: often guided by initial hypothesis -> quantitative
Causal: research in which the major emphasis is on determining a cause-and-effect relationship ->
Quantitative and primary data
- Descriptive research reveals associations between variables
- Causal research reveals associations between changes in variables
Secondary data: data previously collected for purposes other than the research at hand
- Syndicated research: large-scale marketing research that is undertaken by a research firm to
be sold, often on a subscription basis, to a number of clients
Primary data: data collected specifically to answer the questions posed by the current research
objectives
Lecture 2
Univariate statistics (one variable):
- Describe and summarize data – a complete run down analysis of all the variables in your
dataset one-at-a-time
- Inferential analysis – learning about the world
Bivariate statistics (two variables):
- Differential analysis
- Associative analysis
Screening the dataset
1. Check for missing data
a. Assign missing values
b. Delete missing values
i. Listwise -> if a person has a missing value for any variable, then they are
excluded from the whole analysis
ii. Pairwise -> a person’s data is excluded only for analyses for which she has
missing data
2. Find strange codes and errors (consistency checks, analyzing frequencies and plots, reverse
coding)
,Nominal: assigns numbers to identify subjects or objects, nothing is implied by the numbers other
than identification (student number, gender, region etc.)
Ordinal: ranking of objects. Numbers indicate relative positions, but amount of difference between
numbers is unknown. (Preference of brands)
Interval: order is important, and distances are known and fixed. No meaningful zero point. (Likert
scales, attitudes, liking, satisfaction, temperature)
Ratio: highest order of number. Contains most information, order is important, distance is identical,
zero is meaningful. (Weight, age, household size, units sold)
Univariate analysis -> bivariate analysis -> multivariate analysis
Univariate analysis – descriptive analysis
- What is the average age of our customer base?
- What percentage of our customers is female?
- What is the average liking of our product?
- How variable are people’s liking ratings?
One question at a time
Nominal and ordinal -> frequency tables/bar charts/pie charts
Interval and ratio -> central tendency or measure for dispersion/histogram
Measures of central tendency:
- Mean -> interval and ratio data
- Median: middle number when data is sorted in size order -> ordinal, interval and ratio
- Mode: the most frequent number -> all data
Measures of variability/dispersion:
- Standard deviation: variance
- Variance: (actual number – sample average)^2 / n
- Range
Measures of shape:
- Skewness: indicator of distribution symmetry
- Kurtosis: indicator of flatness or peakness of distribution
2
, Sample error/standard error: the standard deviation of sample means
Inferential statistics
Generalize sample results to a population
Central limit theorem (CLT): the distribution of sample means approximates a normal distribution
as the sample size gets larger, regardless of the population’s distribution. As n increases, the
distribution of the sample mean approaches the normal distribution.
Confidence interval: interval that, at a specified confidence level, includes the true population value
- 90% confidence interval > sample mean +/- 1.64 * SE
- 95% confidence interval -> sample mean +/- 1.96 * SE
- 99% confidence interval > sample mean +/- 2.58 * SE
“We are 95% sure that the true (unknown)
population mean is contained in the interval
42.91-47.53”
Two-sided tests
- H0: mu = 40
- H1: mu 40
One-sided tests
- H0: mu smaller or equal to 40
- H1: mu is bigger than 40
Two independent samples
- T test independent samples
- One way anova
Independent samples: compare one variable between different samples
Related samples: compare 2 or more variables within respondents
The larger the test statistic and the smaller the p-value, the less likely is H0!
Two methods to decide whether to reject the null hypothesis:
- Test statistic >2
- P-value <5%
Lecture 3
Latent constructs
Latent constructs: less objective and more
abstract constructs
3
