Professional Documents
Culture Documents
MR Finals
MR Finals
M.R finals
Contents
Chapter 8 – Surveys...................................................................................................................................1
Chapter 11 – construct measurement.....................................................................................................5
Chapter 12 – Questionnaires and supplements....................................................................................16
Lecture 7: Coding, Editing and Preparing Data, and Preliminary Data Analysis; Preparing the
Marketing Research Report and its Presentation................................................................................20
Lecture 8: Testing for difference............................................................................................................30
Lecture 9: Testing for association..........................................................................................................37
Lecture 10: Testing for interdependence.................................................................................................46
Lecture 11: Observation and its variants & experiments & test markets................................................51
Lecture 12: Information-driven Technology and the Research Process..................................................63
Chapter 8 – Surveys
Evaluating survey research design
- advantages
o accommodate large sample sizes so results can be
generalised onto target population
o produce precise enough estimates to identify
even small differences
o easy to administer and record answers to
structured questions
o facilitate advanced statistical analysis
o concepts and relationships not directly
measurable can be studied
- Disadvantages
o Questions that accurately measure respondent
attitudes and behaviour can be challenging to
develop
o Richness of detail and in depth data difficult to obtain
o Timeliness of data collection is a challenge
o Low response rates can be a problem
- Sampling error
o The difference between the findings based on the sample and the true values of the
defined target population
o Sampling error is caused by the method of sampling used and the size of the sample
o It can be reduced or controlled by increasing the sample size and using appropriate
sampling method
- Non Sampling error
o Survey research errors not related to sampling
o Common characteristics include:
imperfections in the survey design or execution of the research process
nonsampling errors are controllable (human mistakes)
nonsampling errors cannot be directly measured
one type of error can potentially allow other types of errors to enter the data
collection process
o Most nonsampling errors stem from:
1. respondent errors
- Misinterpreting market performance factors as being the problem rather than the symptom
- Researcher’s inaccurate transformation of the problem into research questions
- Person-administered surveys
o Trained interviewer asks questions and records answers
In home – takes palce in respondents home or office
Executive interview – business executive is interviewed in person
Mall intercept – shopping patrons stopped and asked for feedback during
shopping
Purchase intercept – stopped and asked for feedback at point of purchase
- Telephone-administered surveys
o Telephone interviews and computer-assisted telephone interviewing (CATI)
Telephone interview – interview over phone
Computer assisted telephone interview – computer used to conduct telephone
interview, respondents opush buttom on their phone to give answers
Completely automated telephone surveys – survey completely administered by
computer
Mobile phone survey – can use sms for simple polling and short surveys
- Self-administered surveys
o Respondent reads the survey and records his responses without the presence of an
interviewer
Mail panel survey – surveys mailed to representative sample
Dropp off survey – questionnaire left with respondent to do at later time
Direct mail survey – distribution and return via post
- Online-administered surveys
o Web-based and email surveys, online panels, computer assisted etc
Fax survey –
Email survey
Internet survey
Construct development - An integrative process in which researchers identify the subjective properties
of a research variable for which data should be collected.
- For each construct being investigated, the researcher must determine its dimensionality traits
(i.e. single versus multidimensional) before developing appropriate scales.
Construct properties:
Construct dimensionality:
- Domain observables - The set of identifiable and measurable components associated with an
abstract construct.
Construct validity:
- Discriminant validity to exist, researchers must be able to establish the fact that the construct
being investigated does not significantly correlate with other constructs that are operationalised
as being different.
- the use of untested or inappropriate scale indicators to measure the construct creates a
measurement artefact that could lead to misinterpretations of the true components as well as
the true dimensionality traits making up the investigated construct.
- convenience sampling
Construct operationalisation:
Scale development
- Scale development can be defined as the process of assigning a set of descriptors to represent
the range of possible responses to a question about a particular object or construct
- Within this process, the focus is on measuring the existence of various properties or
characteristics of a person’s response.
- Scale points - Designated degrees of intensity assigned to the responses in a given questioning
or observation method.
Scale properties:
-
-
-
-
-
-
Scale levels:
- Ordinal - used to depict non-mathematical ideas such as frequency, satisfaction, degree of pain,
happiness etc.
o Ordinal = order (relative position of labels) + description (label)
- Interval - A scale where order is known as well as the difference between these variables.
o Helps to indicate the distance between two variables
o Contains ordinal properties, except an addition of distance between variables
o Values are established, constant and measureable
o Equal distance between attributes
o Attitude scales can be interval
o Can do subtraction and addition of scores
o Can be likert scales
But scales cannot contain a neutral response
Each attribute is a label for a meaningful amount of something
i. E.g – not at all = 2/10 and very high = 8/10
- Ratio - A scale that allows the researcher not only to identify the absolute differences between
each scale point but also to make comparisons between the responses.
o Allows a true natural zero to be a valid response
o Has fixed number responses only
o Can calculate most descriptive statistics with this measure
o E.g - age
Scale formats
Likert scales - An ordinal scale format that asks respondents to indicate the extent to which they agree
or disagree
- with a series of mental belief or behavioural belief statements about a given object.
Semantic differential –
o A bipolar scale format that captures a person’s attitudes or feelings about a given object
o Randomisation of the positive and negative pole descriptors - Bias’s may exist if positive
end is on left always. Randomisation is positive.
o Lack of magnitude expressed in the pole descriptors – Only end 2 poles express extreme
intensity. Researched will have to guess how respondent is interpreting other points.
o Use of non-bipolar descriptors to represent poles – sometimes the poles are not true
opposites
E.g - take the ‘expert/not an expert’ scale item – not an expert’ do not allow the
respondent to interpret any of the other scale points as being relative
magnitudes of that phrase. Other than that one endpoint described as ’not an
expert’, all the other scale points would have to represent some intensity of
‘expert’, thus creating a skewed scale towards the positive pole.
10
Behaviour intention scales - A scale designed to capture the likelihood that people will demonstrate
some type of predictable behaviour intent towards purchasing an object or a service in a future time
frame.
o To increase the clarity of the scale point descriptors, the researcher can attach a
percentage equivalent expression to each one.
11
Non-comparative scale - A scale format that requires a judgment without reference to another object,
person or concept.
1. a graphic rating scale - graphic rating scale A scale that uses a scale point format which
presents the respondent with some type of graphic continuum as the set of possible
responses to a given question.
2. a performance rating scale - performance rating scale. A scale that uses an evaluative scale
point format which allows the respondent to express some type of post decision or
behaviour evaluative judgment about an object.
3. a staple scale - staple scale A modified version of the semantic differential scale that takes a
single narratively expressed descriptor and centres it within a numeric set of plus and minus
descriptors.
Comparative scales
1. Rank order - A rank-order scale incorporates a scale point format that allows respondents to
compare their own responses by indicating the first preference, second preference, third preference and
so forth, until all the desired responses are placed in either a ‘highest to lowest’ or a ’lowest to highest’
rank order
12
Constant sums scale - constant sums scale. A scale format that requires the respondent to allocate a
given number of points, usually 100 among several attributes or features based on their importance to
the individual; this format requires a person to evaluate each separate attribute or feature relative to all
the other listed ones.
Single-item scale design - A scale format that collects data about only one attribute of an object or a
construct.
- An easy example to remember that illustrates this point is the need to collect age data about the
respondents. The object is ‘a person’ and the single attribute of interest is that person’s ‘age’.
Multiple-item scale design - A scale format that simultaneously collects data on several attributes about
an object or a construct.
Formative composite scale design - A multi-item scale format where each individual scale item
represents a different part of the whole construct, object or phenomenon to be measured.
13
o For example, to measure the image of the Mercedes-Benz EClass, the researchers would
have to measure the different attributes that make up that car’s image, such as
performance, resale value, styling, price, safety features, sound system and
craftsmanship. By creating a scale that measures each pertinent attribute, the
researcher can sum the parts into a complete (e.g. formative) whole that measures the
overall image held by respondents towards the E-Class.
Reflective composite scale design - A multi-item scale format where each individual scale item is an
equally valid indicator of the whole construct, object or phenomenon to be measured.
- For example, to measure the image of a Mercedes-Benz E-Class using a reflective composite
scale, the researcher can use a variety of synonyms for image as the individual scale items, such
as look, appearance, aura or persona of the E-Class.
Intelligibility of questions:
- The degree to which the questions on a scale are understood by the respondents.
- The researcher should try to eliminate all guessing on the part of the respondent. Respondents
should be able to understand what types of data are being asked for and how to respond. The
intelligibility factor thus promotes the use of ’respondent instructions’ in scale designs,
especially in self-administered surveys.
- The extent to which the scale point elements match the data being sought.
Discriminatory power:
- The scale’s ability to differentiate significantly between the categorical scale responses.
- Moreover, the researcher must decide how many scale points are necessary to make up the
relative magnitudes of a desired response scale.
- Usually between 3-7 points
Scale reliability:
- The extent to which a scale can produce the same measurement results in repeated trials.
- Test-retest technique:
o A technique of measuring scale reliability by administering the same scale to the same
respondents at two different times or to two different samples of respondents under
similar conditions.
- Problems with the test-retest approach:
o First, some of the students who completed the scale the first time might be absent for
the second administration of the scale.
14
o Second, students might become sensitive to the scale and therefore alter their
responses in the second measurement.
o Third, environmental or personal factors may change between the two administrations,
thus causing changes in student responses in the second measurement.
- Balancing positive/negative scale point descriptors
o If wanting a balanced/unbiased response – balance positive and negatives
o Some respondents wont have enough knowledge or experience with the given topic to
be able to accurately assess their thoughts or feelings.
o An alternative approach to handling situations in which respondents may feel
uncomfortable about expressing their thoughts or feelings about a given object,
because they have no knowledge of or experience with it, would be to incorporate a
’not applicable’ response choice that would not be part of the actual scale
- In deciding what scale level should be developed, the researcher must take into consideration
what kind of data analysis will be conducted after the data are collected.
- The basic sample statistics that are generated through analysing data; these are the mode, the
median and the mean.
Measures of dispersion:
- Relate to how all the data are actually dispersed around a given central tendency value.
- The sample statistics that allow a researcher to report the diversity of the data collected from
scales; they are the frequency distribution, the range and the estimated sample standard
deviation.
15
16
17
Questionaire components
- Words
- Questions
o Unstructured questions - Open-ended questions formatted to allow respondents to
reply in their own words.
o Structured questions - Closed-ended questions that require the respondent to choose
from a predetermined set of responses or scale points.
o Bad questions - Any questions that prevent or distort the fundamental communication
between the researcher and the respondents.
Incomprehensible
Leading or loaded
Unanswerable
Double barrelled – more than 1 issue of a time
- Format
- Hypothesis
Design steps:
18
19
Cover letter:
Supplementary documents:
- Supervisor instructions
o A form that serves as a blueprint for training people on how to execute the
interviewing process in a standardised fashion; it outlines the process by which to
conduct a study that uses personal and telephone interviewers.
- Interviewer instructions
o The vehicle for training the interviewer on how to select prospective respondents,
screen them for eligibility and conduct the actual interview.
o The main importance of interviewer instructions is to ensure that all the
interviewers conduct the interviews in basically the same fashion.
- Screening forms
o A set of preliminary questions that the interviewer uses to determine the eligibility
of a prospective respondent for inclusion in the survey.
o
- Quota sheets
20
o A simple tracking form that enhances the interviewer’s ability to collect data from
the right type of respondents; the form helps ensure that representation standards
are met.
o May need quota sample from particular segment
- Rating cards
o Cards used in personal interviews that represent a reproduction of the set of actual
scale points and descriptions used to respond to a specific question/set-up in the
survey. These cards serve as a tool to help the interviewer and respondent speed up
the data collection process.
- Call record sheets
o A recording document that gathers basic summary information about an
interviewer’s performance efficiency (e.g. number of contact attempts, number of
completed interviews, length of time of interview).
Data editing
- Ensure that mistaken omissions by interviewer are filled in – e.g, didn’t tick a box
- Interviewer asking the proper questions
- Proper recording of answers
- Proper screening of respondents – manually check if person fitted into screening rules
- Proper recording and coding of open-ended questions
21
Data coding
Data entry
o Those tasks involved with the direct input of the coded data into some specified software
package that ultimately allows the research analyst to manipulate and transform the data
into useful information.
o Can be entered directly into the computer
o Touch screen, light pen, kiosks, computerised /online surveys
o Entered manually
o Scanned through ocr software
E.g – multiple choice tests
Error detection
o This is the process of ensuring that the data entered are correct and error free.
o Customised software can prevent errors by making it impossible for data entry personnel to
make certain types of mistakes.
o Data analysis tools can run error check procedures to identify wrong data, incorrect data,
missing data.
o Frequency count, outlier detection, etc
o Computer and internet-based surveys can accommodate vast and complex arrays of data,
greatly increasing the capacity for data collection and substantially reducing confusion and
errors by interviewers and respondents.
22
Data description
o The final step in preparing the data for further analysis is to describe the data sample in such
a way so as to reveal the general pattern of responses and the general profile of a ‘typical’
respondent.
o Why is it required – Almost all data sets are disaggregated (just rows and columns) – Every
set of data needs some summary information developed that describes the numbers it
contains
o Statistical techniques used – Data tabulation (frequency distribution, cross-tabulations, etc.)
– Measures of central tendency (mean, median, mode)
Data tabulation
o The simple process of counting the number of observations (cases) that are classified into
certain categories.
o One-way tabulation - Categorisation of single variables existing in a study.
is the categorisation of single variables existing in the study. In most cases, a one-
way tabulation shows the number of respondents who gave each possible answer to
each question on the questionnaire.
When a research team performs a ‘one-way’ tabulation they focus on a single
variable operating in the research study
Research analysts use one-way tabulations to:
determine the degree of non-response to individual questions
locate simple blunders in data entry
calculate summary statistics on various questions, for example averages,
standard deviations and percentages
communicate the results of the research project
Profile sample respondents
Distinguishing characteristics between groups
Establishing percentage of respondents who respond differently to different
situations –
Frequency table is constructed to illustrate one-way tabulations
23
Cross-tabulation - Simultaneously treating two or more variables in the study; categorising the number
of respondents who have answered two or more questions consecutively.
o This simultaneously treats two or more variables in the study
by categorising the number of respondents who
have responded to two or more consecutive
questions
Helps to analyse relationships among and
between variables
To quickly compare how different groups of
respondents answer survey questions
Provide a valid description of both
aggregate and subgroup data
Merging of frequency distributions of two or more
variables
o As always, the analyst needs to base the selection of variables on the objectives of the
research project.
o Two key elements of cross-tabulation are:
how to develop the cross-tabulation
how to interpret the outcome
o It is widely used
o Normally the main form of data analysis in most marketing research projects
Easily understood and interpreted by managers
o Simple to conduct and appealing to less sophisticated managers
Challenges
- The analyst should take care to construct cross-tabulations that accurately reflect
information relevant to the objectives of the project
- Certain survey approaches can lend themselves to the construction of an endless variety of
cross-tabulation table
- More than three variables can be cross-tabulated but interpretation complex
- Cross-tabulations are not efficient when examine relationships among several variables
- Cross-tabulations can interpret associations not causation
24
- Mean
o The arithmetic average of the sample; all values of a distribution of responses are
summed and divided by the number of valid responses.
o Survey data should display some central tendency with most of the
o responses distributed around the mean
o For interval or ratio data, researchers generally use the mean
o It is fairly insensitive to adding or deleting data values. The mean can be subject to
distortion, however, if the data contain extreme values.
- Mode
o The most common value in the set of responses to a question; that is, the response
most often given to a question.
o For nominal data, researchers generally use the mode.
- Median
o The middle value of a rank-ordered distribution; exactly half of the responses are
above and half are below the median value.
o For ordinal data, researchers generally use the median
Measures of dispersion
25
- Measures of dispersion describe how close to the mean or other measure of central
tendency the rest of the values in the distribution fall.
o how spread apart the scores of the distribution are or how much the scores vary
from each other
Range
- The distance between the smallest and largest values in a set of responses.
Standard deviation
Variance
- Variance between the mean and the observed value is called the deviation from the mean
- Variance is the mean squared deviation of all the values from the mean
- As soon as researchers determine the sum of the squared deviations, they divide it by the
number of respondents minus 1. They subtract the number 1 from the number of
respondents to help produce an unbiased estimate of the standard deviation. The result of
dividing the sum of the squared deviations is the average squared deviation. To get the
result back to the same type of units of measure as the mean, they simply take the square
root of the answer.
- Since the estimated standard deviation is the square root of the average squared deviations,
it represents the average distance of the values in a distribution from the mean. If the
estimated standard deviation is large, the responses in a distribution of numbers do not fall
very close to the mean of the distribution. If the estimated standard deviation is small, you
know that the distribution values are close to the mean.
- Can represent how reliable and strong the central tendency is.
- Often, quantitative researchers are so concerned about statistics, computer output and
questionnaires that they forget to provide a clear, logical interpretation of their findings.
- Should include;
o Executive summary
26
Credibility
- The quality of a report that is related to its accuracy and professional organisation.
- A professional report does not tolerate any degree of carelessness in handling data,
reporting statistics or describing outcomes.
- Errors in mathematical calculations, grammatical errors and incorrect terminology are just a
few types of inaccuracies that can also serve to diminish the credibility of the report.
Believability
- The quality of a report that is based on clear and logical thinking, precise expression and
seamless presentation.
- If readers do not understand what they read, they do not believe what they read.
o The literature review and relevant secondary data may be integrated in the analysis of
findings in qualitative data analysis, rather than being presented separately from other
findings
o The objectives and questions in qualitative research tend to be broader more general, and
more open-ended than in quantitative research
o Goal is comparison to previous research findings
o Qualitative researchers rarely present statistics
o The literature review and relevant secondary data analysis is usually presented in the body
of the report (i.e. research background or research issues section)
o Presenting statistics is crucial for quantitative research reports
27
Format
28
Reporting results
No matter how complicated the statistical analysis, the challenge for researchers is to
summarise and present the analysis in a way that makes it easy to understand for non-
specialists.
o Frequencies
o tables, bar charts or pie charts
o
29
Presentation guidelines
o Begin with a slide showing the title, the individual presenters, the client and the marketing
research firm
o A sequence of slides showing the research objectives, the research questions, the research
methodology and a description of the sample
o Additional slides to highlight the research findings or the results
o Conclusion slides with recommendations, conclusions and research implications
30
o Hypothesis: any stated difference or change exists in reality and is not simply due to random
error in sampling – e.g, just drawn the correct sample that validates the hypothesis
o From educated guess – e.g, young people are more likely to use fintech
o Very difficult to prove a hypothesis 100%, because causation could be due to unseen
variables
o That is why we test to reject the null hypothesis
o Researchers have preconceived notion of how variables relate to each other –
o Marketing research
o Prior studies
o Unless a hypothesis is empirically verified, it remains an unproven statement in a testable
format
o Null hypothesis: anything that is thought to be true, or that has been observed in a
sample, is due to random error
- Rejection of the null hypothesis leads to an acceptance of the hypothesis
- Researchers rarely state the null hypothesis explicitly.
o Examples of marketing-related hypotheses: –
o Heavy and light users of a brand differ psychographic ally
o Ski resort is visited by more than 30% of the advanced skiers in the region
o A hotel has a more upmarket image than its nearby competitors
o Concepts of sampling distribution, standard error of the mean or the proportion, and the
confidence interval are fundamental to construction and interpretation of hypotheses tests
Statistical significance
2 Types of errors
o Type I error: when sample results lead to the rejection of the null hypothesis, when it is in
fact true – false positive -
31
o Univariate t–test:
o a method to compare whether the mean for a dependent variable in a sample is
different from the population mean (or mean of interest) of that variable
o Will compare mean value to a test value mean – a hypothesised mean value
- Test mean is the level that they are interested in or hope for
- Test the certainty that the discovered mean is not equal to the test value at
95% confidence level
o Test whether the difference is statistically significant]
o Relatively simple test to conduct
o Practical example
o Testing a sample against a certain benchmark to determine if it would hold true and
high enough score for the population
o E.g – determine if retirement age of a population is 65 (which may be the age of
another country, and we want to determine if our countries average is different)
which may have some practical value with how we compare to them, and what we
can do about it or copy from them, as we may also hypothesis that they are doing
something better than us and are looking for evidence of this.
o Interpretation and assessing significance – use these 3 different values
o Absolute value of T is greater than 2 = statistically significant – reject null
o Confidence interval of difference does not include 0
- means dependant variabe and test variable are always significantly different
from 0 difference (or statistically different than the null hypothesis) – reject
null hypothesis that dependant variable is not differnt from test
- if included 0 (null), then could accept the null
o P value - Sid (2tailed) less than 0 - statistically significant – reject null
- If less thank .95% have to accept null hypotheses – insignificant
1. Read null statement and use this to determine if accept or reject the
null
2. If it is insignificant, then it is a null statement, null hypothesis
32
2.
Bivariate t-test:
33
o e.g. researcher may want to compare the average number of cups of coffee
consumed per day by male students (the one group) to the average number of cups
of coffee consumed by female students (the other group)
o E.g. 2 – management may think there is a difference in customer age between males
and females
34
o Choose level of significance for rejecting null hypothesis and accepting the hypothesis
o Typically, p< 0.05 for rejecting the null hypothesis
o Probability or critical value
o The P value or calculated probability is the estimated probability of rejecting the null
hypothesis of a study question when the null hypothesis is true
o Test statistic
o A statistic on which the decision can be based whether to accept or reject a
hypothesis is called test statistic – Examples: z, t and Chi–square
o If these groups are derived from one independent variable, as in the above case
‘education level’ (albeit broken down into three groups: MBA, undergraduate and
PhD)
- one independent variable which has a number of different levels or
conditions (i.e. groups of students)
– Independent variable must be categorical (nominal variable)
- One dependent variable (i.e, consumption of beer bottles)
– Dependent variable must be metric (either interval or ratio)
o This compares the variance (variability in scores) between the different groups (due to the
independent variable) - with the variability within each of the groups
o Tukey and Duncan output for ANOVA –
o Below shows only 1 subset, which is a subset which does not reject the nul. (no Sig.)
o If had 2 subsets, would indicate different groups, one significant, the other not
35
o F test -
o Larger F ratio indicates more variability between the groups than there is within
each group
o A significant F-test indicates that we can reject the null hypothesis, which states that
population means are equal
o Variance between groups > variance within groups = statistical significance
-
o Null hypothesis in ANOVA – all groups are the same.
36
N way ANOVA
MANOVA
o Analysts can analyse multiple dependent variables and three or more groups,
simultaneously, using a related procedure called multivariate analysis of variance
(MANOVA).
o Assumptions
- D.V - Two or more dependant variables – Interval or ratio level
– E.g – study time, test scores
- I.V – two or more categorical, independent groups
– E.g – ethnicity, physical activity (low, moderate, high)
o –The objective in MANOVA is identical to that in ANOVA—to examine group
differences in means—except that the analysts consider comparisons for multiple
dependent variables.
- For example, a researcher might want to measure customers’ use of several
types of related products, such as golf balls, golf shoes, golf clubs and golf
clothing (the multiple dependent variables). Since use of one of these types
of products may be related to other factors, such as education levels (the
different groups), MANOVA would be a good choice
37
38
- Using Pearson Chi Square ( p value) (2 tailed significance level less than
0.05)
- Person value either positive or negative
- Either significant or not
Scatter diagram
o Allows researchers to not only understand the presence of an association, but also the
strength and direction of the association –
o If researchers know that CD purchases are related to age, then they might want to
know the extent to which younger persons purchase more CDs, and ultimately
which types of CDs
o Covariation is the amount of change in one variable that is consistently related to the
change in another variable of interest (how much one variable varies in relation to another
variable varying)
o Scatter diagrams used to visually understand the covariation between variables
o Whether changes in Y are often systematically related to changes in X
o Relationships can be positive or negative
- Negative covariance – When X increases, Y decreses/moves in the opposite
direction
– Negative – e.g, the more expensive ice cream is, the less you’re
going to buy it.
- No relationship – amount of ice-cream is not dependant on your income
level, because people will buy it regardless
- Curvilinear – e.g where price signals quality (wine, hairdresser, car) – as
price goes up, sales goes up, but at a certain price level, sales goes down as
the price gets too high
o Strong or weak – look at variance
- Strong - when all dots are close to line
- Weak – when dots are spread away from the line
o Can be linear or curvilinear
o Use dependant value on y axis (what you are trying to predict), independent
variable on x axis – e.g, is sales of ice cream related to temperature outside and
what is the relationship
- sales of ice-cream (is dependant and what trying to predict), temperature
outside is independent (predictor – thing that using to predict)
39
o -1<Pearson coefficient<1
o Also includes level of significance
o Tells of the strength of the correlation between variables
o If strong and significant, you can be confident of a linear relationship between variables
o R Value in regression analysis
o This allows a research team to determine the sturdiness of a linear (not curvilinear)
relationship between two variables
o interval and/or ratio- scaled only
o Can only show association; doesn’t allow researcher to make causal statements
o E.g – if know that price and sales are highly correlated – you know there’s a
relationship between the two, but need extra theory to know that its causal
o Pearson coefficient values
o Positive coefficient -
positive linear
relationship
o Negative coefficient –
negative linear
relationship
40
E.g – Significant association because of Pearson coefficient and significance (2 tailed) score.
o This is a statistical measure of the strength and direction of a linear relationship between
two variables where at least one is ordinal-scaled
o e.g. researchers might be interested in comparing whether a respondent’s ranking
of eight business schools in terms of perceived costs is related to how these schools
are ranked by the same person in terms of perceived prestige
41
o E.g above – higher rank in food quality correlates to lower food variety
- Higher food variety correlates to lower food quality
Regression analysis –
42
Bivariate regression –
43
Multiple regression
44
45
46
o Used to estimate the value (utility) respondents associate with different product
and/or service features –
o Do so through putting people in different choice situations and combinations and
observing choice behaviour/ratings to determine;
1. Make a lsit of key attributes and levels of company & competitors product in
question
2. Consumers then pick the features they want
3. Software takes note of different preference combinations
4. Statistical analysis estimates preference scores of different combinations
5. Can then specify any amount of features and get a score of utility/adoption
levels based on the test scores.
o Product image analysis
- The relative contribution of each product attribute can be determined for
use in marketing and advertising decisions
o Segmentation analysis
- Groups of potential customers who place different levels of importance on
the product features can be identified for use as high and low potential
market segments
Benefits
– Ability to provide utility estimates for individual levels of each product attribute – Ability to
estimate non-linear relationships among attribute levels
– Outcome allows to communicate the most preferred combination of features and price to
market to consumers
o Data is collected in realistic ways – consumers can make simple choices and preferences,
don’t have to fill out likert scales etc.
Limitations
– Researchers choose the appropriate attributes and attribute levels that will realistically
influence consumer preferences or choice
– Consumers may have difficulty making choices or indicating preferences among large
numbers of profiles
47
o Interdependence exists when no single variable or group of variables among those under
consideration can be defined as being dependent or independent. Thus, no one variable can
be predicted or explained by the others under consideration.
o The purpose of interdependence techniques is not to predict a variation in one variable
based on a variation of another variable or variables in a data set, but to summarise the
variables. Fundamentally, the objective of interdependence techniques is to group variables
together so that they can be better understood.
o Therefore, a marketing manager who wants to identify various market segments or clusters
of fast-food customers (e.g. burger, pizza or chicken customers) might utilise these
techniques (to summarise who in the market could be classified as a burger, pizza or chicken
customer).
Factor analysis
48
o Based on the patterns of data, researchers can combine these six measures into two
summary measures, or factors, called service quality and food quality
o Researchers do not draw a distinction between dependent and independent variables using
factor analysis; rather, they collectively analyse all variables under investigation to identify
underlying factors.
o Statistical purpose of factor analysis is to determine whether researchers can identify
combinations of variables that will help summarise the data and identify underlying
relationships.
o Factor loading
o Correlation between each of the original variables and the newly developed factors
o Each factor loading is the importance of the variable in measuring the factor
o From -1 to +1
- In the above example
– Waiting time, cleanliness, friendly personnel – Would have a
positive factor loading for service quality –
– Would have a 0 - factor loading for food quality
o High loading means that the variable helps to define the factor or is closely
associated to the factor
49
Naming factors
o The researcher examines (intuition and knowledge) the variables that have high loadings on
each factor.
o There often will be a certain consistency among the variables that load high on a given
factor.
o What will you name factors found in previous tables? Service quality, food quality.
Number of factors
o Deciding on how many factors to retain is a very complex process because there can be
more than one possible solution to any factor analysis problem
o Many situations can involve anywhere from one factor to as many factors as there are
variables
o Consider:
o Review the percentage of variation in the original data that each factor explains
o How much does each factor contribute to the understanding of the data
o Look at how many elgenvalues > 1
o
50
o Factor analysis can also be combined with multiple regression analysis for further analyses.
o Can use factor loading scores for factors, instead of using regression analysis that
need to consider all of the variables that are
related
1. Calculate factor scores (composite scores
of each respondent in the derived factors)
2. Save factor scores into data base
3. Use those factor scores as independent
variables in a regression
SPSS APPLICATION
Cluster analysis
o The purpose of cluster analysis is to classify or segment objects (e.g. customers, products,
market areas) into groups so that objects within each group are similar to one another on a
variety of variables.
o Examples
o Using 1 Variable – will be classified according to the response options
o Using 2 variables – will be clustered according to clusters where responses are same
in both variables –
o Example – Depression and anxiety scores – may have a cluster with high depression
high anxiety, cluste with low depression, low anxiety, maybe that’s it, because they
usually come together
o Cluster analysis seeks to classify segments or objects such that there will be as much
likeness within segments and as much difference between segments as possible.
o This method strives to identify natural groupings or segments among many variables
without designating any of the variables as a dependent variable.
o Cluster analysis pinpoints what is homogeneous/similar within groups but
heterogeneous/different between them.
o Similar to factor analysis, researchers most frequently use cluster analysis with metrically
measured variables. Non-metric adaptations are also possible.
51
o Different to factor analysis – cluster analysis creates groups out of independent variables –
e.g, high spenders, low spenders and medium spenders.
o Factor analysis – groups independent variables into a single underlying classification
o Clustering could be defined as the process of dividing items into unknown number of groups
whose members are alike in some way. A cluster is therefore a collection of items those are
similar among themselves and are dissimilar to the items belonging to other clusters.
o Cluster analysis procedures involves measuring the similarity between objects on the basis
of their ratings on the various characteristics.
o Several cluster analysis procedures are available, each based on a somewhat different set of
complex computer programs.
o Researchers determine the degree of similarity between objects often through a distance
measure.
o New-product research – Clustering brands can help a firm examine its product offerings
relative to competition. Brands in the same cluster often compete more fiercely with each
other than with brands in other clusters.
o Test marketing – Cluster analysis can group test cities into homogeneous clusters for test
marketing purposes.
o Buyer behaviour – Cluster analysis can be employed to identify similar groups of buyers who
have similar choice criteria.
o Market segmentation – Cluster analysis can identify distinct market segments on the basis of
behavioural and descriptive variables.
52
o Information condition
o Other data-collection methods might lessen the data’s accuracy and
meaningfulness due to respondents’ faulty recall
- For example, people might not accurately recall the number of times they
zap commercials while watching their favourite TV program
o Type-of-data condition
o Observation used only when a respondent’s feelings are relatively unimportant to
the research objective or believed to be readily inferable from the behaviour
o Not appropriate for ‘why’ question
53
Observation characteristics
o Direct observation
o This is the process of observing actual behaviours or events and recording them as
they occur.
o Direct observation is carried out by a person (rather than via a mechanical device),
who is observing and recording actual activities.
o Direct observation of artefacts of past human behaviour is effectively a trained
investigator interpreting a form of secondary data.
o Indirect observation
o Process of observation of the artefacts that represent behaviours from earlier time
periods.
o Indirect observation techniques can be used to capture and understand individuals’
past behaviours.
o Interpretations and inferences of past data should be viewed cautiously.
o Subjects’ awareness of being observed
o The degree to which subjects consciously know their behaviour is being observed
and recorded
o Disguised observation
- When subjects are completely unaware that they are being observed.
Example: Mystery shopper technique
o Undisguised observation
- Whenever subjects are aware that they are being watched Example: sales
representatives
o The researcher should minimise the presence of an observer to avoid response or
subject bias
54
o Structuredness of observation
o Structured observation
- Researchers use a checklist or a standardised recording form to help the
observer restrict his or her attention to prespecified behaviours or events
- The degree to which the researcher knows which activities and movements
of the observed are known.
o Unstructured observation
- No restrictions on the observer regarding what should be recorded.
- Researchers brief the trained observers on the research objectives and
information requirements and then allow them to use their own discretion
(based on interest and relevancy) in determining what behaviours are
actually recorded.
- Observer monitors all aspects of the phenomenon that seem relevant to the
problem at hand, e.g. observing children playing with new toys.
o Medium of observation
o This relates to how the behaviours or events will be observed.
o The researcher can choose between human observers and mechanical devices.
o Human observation
o Data collection by a researcher or trained observer who records human behaviours
or events.
o The observer must have a good understanding of the research objectives and strong
observation and interpretive skills.
o Mechanical observation
o Data collection using a mechanical device to capture human behaviours or events
o Based on the assumption that physiological actions and reactions are predictors of
people’s unobservable cognitive thoughts or affective emotional feelings
o Advances in technology have made mechanical observation devices very useful and
cost-effective and has improved flexibility and accuracy in the data collection
process
o Examples: video camera, traffic counter, optical scanner, eye-tracking monitor,
pupilometer, audio voice pitch analyser, galvanometer, online tracking.
o Mechanical and software based observation
o Voice pitch analysis
o Pupil analysis
o Eye-tracking analysis
o Galvanic skin response analysis
- An electronic instrument that measures a subject’s involuntary changes in
the electronic resistanceof his or her skin.
o Social media monitoring
o Listening platform
55
Observation Variants
o Ethnography
o Occurs in the consumer’s natural purchase, usage or consumption setting, and
involves combining observation with interviews to record the dynamics of human
behaviour and the influence of culture
o Study can be conducted over a lengthy time period at home, while shopping, at
leisure, in the car and the workplace
o Film or a video could be made which documents the activities, interviews and
behaviour of the consumer or the participant
o Netnography
o Technique draws on ethnography, but it uses ‘found data’ on the internet that is
produced by virtual communities
o In netnography researchers must:
- gain entry into the culture/community
- gather and analyse data
- ensure trustworthy interpretation
- provide opportunities for feedback from members of the community.
56
o investigating cause
o effect relationships then the researcher should consider using a causal research
design, such as experimental procedures or test marketing
o In causal research, the emphasis is on specific hypotheses about the effects of changes of
one variable on another variable.
o A variable is any observable and measurable element (or attribute) of an item or
event.
o Causal research involves experiment.
o The researcher attempts to identify the relationships among different variables by
manipulating one or more independent variables and measures their effect on one
or more dependent variables while controlling the effect of additional extraneous
variables
Key terms
Experimentation
57
58
Validity in experimentation
o Internal validity - Internal validity exists when the researcher can rule out rival
explanations for the observed results.
o The extent to which the conclusions drawn from the experiment are true – are there
any other alternative explanations, does it measure what is supposed to measure
- Refers both to how well a study was run (research design, operational
definitions used, how variables were measured, what was/wasn't measured,
etc.), and how confidently one can conclude that the observed effect(s)
were produced solely by the independent variable and not extraneous ones.
– In experimental research, internal validity answers the question,
‘Was it really the treatment that caused the difference between the
subjects in the control and experimental groups?’
- For example, in an experiment on the effects of electricity, if a subject
(person) receives a shock (experimental treatment) and jumps (observed
effect), and the subject jumps only because of the shock and for no other
reason, then internal validity exists.
o External validity – The extent to which a causal relationship found in a study can be
expected to be true for the entire target population.
- A study that readily allows its findings to generalise to the population at
large has high external validity.
- Depends on if have large enough sample
- How natural, how true to the real environment, is the experiment – is it
artificial? If so, external validity is very low.
o Construct validity
- This defines how well a test or experiment measures up to its claims.
- It refers to whether the operational definition of a variable actually reflect
the true theoretical meaning of a concept.
- Process of correctly identifying and understanding both the independent
and dependent variables included in an experimental design.
– Are we really measuring what we want to measure?
59
60
T.O.M – want to measure attitudes before seeing an advertisement and then attitudes after –
Depending on when you ask then about their attitudes, it may manipulate what is going on in their
mind.
EoNSTS – e.g – if ask old and young person to take medication, then results will be different –
randomisation or carefully choosing subjects may increase validity
Reliability of experiments
o Experimental design reliability is the degree to which the design and its procedures can be
replicated and achieve similar conclusions about hypothesised relationships.
o Reliable research findings are repeatable.
o Repeatedly same results = reliability
o The conclusions drawn can be generalised beyond the particular conditions in the initial
experiment (i.e. external validity).
61
Experimental designs
62
Test Marketing
63
o Marketing mix variables (IV) are varied in test marketing, and the sales (DV) are
monitored
o Conducted in a regular sales setting
o Objectives
o To determine market acceptance of the product
o Evaluating alternative levels of marketing mix variables
o Predicting new product/service acceptance levels
o Advertising and image effectiveness
o Risks
o Competitors may lower prices to combat your own price change
o Other unobservable factors that may impact sales but have nothing to do with 4 P’s
Types of test markets
64
o A type of laboratory experiment that aims to imitate real life, where respondents
are selected, interviewed and then observed making or discussing their purchases.
o Simulated test markets can lead to mathematical models used to forecast factors
such as awareness, trial, sales volumes, impact on other products.
o Advantages
- Internal validity
- Cost and time savings
- Can predict trial, repurchase and purchase cycle accurately – can ask
respondents questions
- Minimised exposure to competition
o Disadvantages
- External validity
- Isolation from real world
- Broad-based customer reaction is difficult to measure
o Virtual test markets
o Tests that are completely computerised, allowing the test subjects to observe and
interact with the product as though they were actually in the test store’s
environment
o Simulated stores appears in a computer screen
o Advantages
- ‘Stores’ closely resemble reality.
- Can be changed quickly – design of store, packaging, etc.
- No exposure to competition
o Field experiments
o Experimental designs that manipulate the independent variables in order to
measure the dependent variable in a natural test setting.
65
Use of portals
o Regular portals
o A portal is a website that offers a broad array of resources and services.
o Portals also can ‘drill down’ into lower level data to identify both negative and
positive business trends – e.g. Yahoo!, Excite, Mygarden.net.au or Bicycles.net.au
o Enterprise Information Portal (EIP)
66
o An internet site owned and operated by an organisation to support its total business
operation (i.e. Dashboard)
o Used for enterprise-wide data integration, tracking flow of information and
detecting problems in business operations
o Users can graphically view information – e.g. Google Dashboard
Benefits of dashboards
o Advances in technology are dramatically changing the methods used to obtain and interpret
the data
o Take transactional data and combine with other information
o Technology impacts definition and categorisation of customer data:
o Transactional customer data: the information contained in a strategic business unit
- e.g. cash withdrawals from an ATM, reservation, purchases from an online
store, clickstream
o Analytical customer data: data used in performing analyses as the basis for
managerial decision making
- e.g. market and industry trends, competitor information and
macroenvironmental changes
- Secondary data
o The success of any marketing program depends on leveraging high-quality timely data
o Transition from traditional behavioural intention measures to real-time transaction
measures
67
68
o Data warehouse
o Logical aggregation of information stored in a single data location
o Most customer data is recorded and stored in a data warehouse
o Contains and collects secondary data, competitive information, sales data,
environmental information and primary data on customers
o e.g. customer data for all Apple products produced and marketed
o Data mart
o Contains all the types of data included in a data warehouse but in a format relating
to only one product of a company
o e.g. customer data for a single Apple product
o Customer information stored in data warehouses or data marts is multidimensional
o The benefit of multidimensional data is that it enables researchers to develop
customer profiles and buying behaviours across products or brands
o Examples
- Customer attributes: age, gender, education, purchase frequency
- Product attributes: style, features and quality levels
- Brand attributes: colour, price, package size and taste
- Store attributes: discount, department store, convenience store, location,
inventory levels
69
Data mining
Blogs
o Marketing researchers use blogs as a way of capturing customer data and opinions and
analysing it for their own use
o Blogs – provide commentary or news on a particular subject
o serve as personal online diaries
70
o typically combines text, images and links to other blogs, web pages and other
media related to its topic.
o enable readers to leave comments
o micro-blogging (blogs with very short information posts)
Analysing text in blogs
71