4110 Exam 2
ex: survey
characteristic to be measured: attitude towards gun control scaling: scale on seven-point scale (favor or oppose gun control) measurement: .1-3 favorable attitude .4 neutral .5-7 oppose
recomputing cluster centroids
cluster centroid simply the average of all the points in that cluster
tradeoff matrix
different people could have different tradeoff matrices
simple random sampling
each population element has an equal chance of being selected -simplest method observations: x could be attitudes, market share, value of customer, voting... law of large numbers- given enough time, the house always wins
modern conjoint analysis
establishes contribution of each attribute to a customer's overall evaluations to: -predict market share of a proposed new product, given the current offerings of competitors -predict the impact of a new competitive product on the market share of any given product in the marketplace -determine consumers' willingness to pay for a proposed new product -quantify trade-offs customers or potential customers are willing to make among the various attributes or features that are under consideration in the new product design -identify consumer segments
one stage cluster sampling
ex: we want beer consumption of students -choose blocks of dorms/apartments -find beer consumption of all students in sampled blocks
two stage cluster sampling
ex: we want candy preferences of all kids in US -treat zipcodes as clusters -first stage: random sampling of zipcodes -second stage: random sampling of households in each zipcode selected in first stage
why firms use conjoint
firms want to increase profits by providing the product features that consumers value
randomized response approach
flip the coin...
when not to use exploratory research
for conclusive results
how surveys are answered
interpret question: -retrieve info form judgment -map judgments onto response scale
Summary of scales
look at chart
communication methods of data collection
mail, fax, telephone, in home interview, mall intercept, internet
convenience sample
-accidental, based on ease of accessibility -mall intercept -radio station asked people to call in disadvantage: not representative
Profiling Clusters
-how big is the segment -how fast is the segment growing -where do they shop -how do they buy -how often do they buy -how much do they buy -how do they use the product
profile analysis (snake plot)
-application of semantic differential scale -overall comparison of brands hard to grasp with many brands and attributes -not all attributes are independent -plot mean ratings for each object on each scale for visual comparison -a snake plot connects the mean value on each attribute -typically place positive descriptors on the right
actionability
-are customers in the segment and the marketing mix necessary to satisfy their needs consistent with goals and core competencies of the firm -will segmentation help develop effective marketing message?
Cluster Limitations
-assumes that the clusters are in circular way, if they are thin and long then it tends not to work well -comparing python clustering algorithms
k means clustering
-belongs to non-hierarchical class of clustering algorithms -aims to partition n observations into K clusters in which each observation belongs to the cluster with the nearest mean, serving as a prototype of the cluster
accessibility
-can managers reach the identified segments -how to reach through promotion and distribution
identifiability
-can managers recognize segments (profiling) -through demographics
number of clusters: elbow criterion
-choose a number of clusters so that adding another doesn't add sufficient information -minimize the within cluster variance and maximize the between cluster variance -ratio of within to between cluster variance is plotted against the number of clusters -the point at which a sharp bend occurs indicates the number of clusters
setting up focus groups
-define research objective -communication with moderator -define target participants
types of primary data
-demographics -psychographics -awareness -intentions -motivations/benefits -behavior
deciding on order, wording, and layout
-do any words have vague or ambiguous meanings -are there any questions double barreled -leading or loaded questions -potentially confusing instructions -applicable to respondents -appropriate length -order bias
sustainability
-do the segments represent a large enough portion of the market -are segments big enough for marketing programs to be profitable
ideal characteristics of cluster sampling
-each cluster is a small scale representation of target population -each cluster is as heterogeneous as the population -clusters should be as similar to each other as possible
probability samples
-each unit has a known probability of being included in the sample -simple random sample -sampling error can be inferred often more expensive
conjoint: new products
-estimate market share of brands that differ in attribute levels -determine attribute composition of most preferred brand
qualitative research characteristics
-exploratory research -less structured than other types of research --high degree of flexibility -number of respondents is typically small --usually not representative of population
non-probability sampling: circumstances
-exploratory research -pre-testing a questionnaire -when probability samples are almost impossible --difficulty in acquiring a suitable sampling frame -generalizability of research
Horizontal attributes
-features that can't be ordered in an objective way -ex: design
conjoint overview: objective
-find importance of attributes and levels of attributes -conduct what-if analysis (sales, market share)
when to use exploratory research
-gain background information -problem formulation aid -identify alternative courses of action -develop hypotheses -isolating key variables -defining terms -establishing priorities for future research -help interpret conclusive research -for dependable estimates of market characteristics -for establishing cause and effect relationships
cluster analysis is a means of segmenting markets
-goal: find groups of similar consumers -group them together -describe each group for subsequent targeting and positioning
takeaways: sampling
-a badly chosen big sample is much worse than a well-chosen small sample -be careful about selection bias and nonresponse bias
qualitative research purposes
-access consumers' mind --experience, concerns, perceptions, feelings, thoughts, intentions, motivations -understand range and complexity of consumer thoughts --develop consumer vocab
qualitative approach unstructured approach: advantages
-individual's organization of a relatively unstructured stimulus indicates individual's basic perception/feelings -the more unstructured the stimulus, the more individual has to project his/her own emotions, needs, attitudes, values -questionnaire: highly structured (choose from a set of alternatives) -record feelings as you watch this ad: unstructured (choose own interpretation)
focus groups v. quantitative research
-insights not rules -social not individuals -homogenous not diverse -flexible not standardized -words not numbers
survey
-involve questioning -yield primary data -all primary data don't come from surveys
disadvantages: observation
-may not be representative -subjective interpretation -inability to learn why
one-on-one interviews
-more in-depth than focus groups -enhances discussion and independent thought -criteria for success --establishing relaxed sympathetic relationship --ability to probe/clarify without biasing results
advantages: observation
-natural environments -no recall error -sometimes it's the only way
conjoint overview: how it works
-new (hypothetical) products are created as profiles (combo) of attributes and attribute levels -consumer respondents rate new products in a survey -regression analysis applied to ratings
interval scale
-numbers used to rank objects also represent equal increments of the attribute being measured -differences can be compared -entire range of statistical operations can be employed for analysis what can you identify -distribution of brand preferences -most popular brand -second most preferred item for each person -mean rating of each brand
nominal scale
-objects are assigned to mutually exclusive, labeled categories -no necessary relationships among categories -only possible arithmetic operation is a count of each category -which of the following brands do you prefer what can you identify -distribution of brand preferences -most popular brand -no ordering or spacing is implied
formulating questions
-open ended v. close ended open ended -important to measure the salience of an issue to a respondent -there are too many responses to be listed or cannot be foreseen -verbatim responses are desired to give the flavor of people's answers or to cite examples close ended -information requirement is clear and known
disadvantages of asking
-people may not know what group they belong to -people may believe they belong to multiple groups -researcher often doesn't know what groups exist in the market
survey characteristics
-population has been defined correctly -sample is representative of the population -respondents selected are able and willing to cooperate -questions are understood by the respondents -respondents have the knowledge, opinions, attitudes or facts required -interviewer correctly understands and records the response
non-probability samples
-probability of units being included in the sample is unknown -mall intercept -sampling error cannot be inferred often less expensive
judgment sample
-purposive -elements are hand picked to obtain better results -swing states in election -classify types of ads -identify strengths and weaknesses
ordinal or rank scale
-ranks objects or arranges them in order by some common variable -does not provide info on how much difference there is between objects -arithmetic operations are limited to statistics such as median or mode what can you identify -distribution of brand preferences -most popular brand -have ordering but no spacing is implied
focus groups: disadvantages
-recruiting problems -group peer pressure -dominant personalities -hard to quantify results -small sample size -can't generalize to target population
focus groups: advantages
-relatively quick info -stimulate new ideas -relatively inexpensive -very flexible -very spontaneous -group interaction
nonresponse bias
-respondents differ in meaningful ways (systematically) from nonrespondents (the differences are not random)
planning what to measure
-revisit the research objectives -decide on the research issue of your questionnaire -get additional information on the research issue from secondary data sources and exploratory research -decide on what is to be asked under the research issue
K means advantages
-robust to different types of variables -appropriate for large data sets common in mktg -less sensitive to some customers who are outliers
disproportionate samples
-sample sizes are not proportional to stratum sizes -more heterogenous group --> more samples
drawbacks: non-probability samples
-sampling units elected by interviewer, usually those who are most accessible, at home, with time, acceptable appearance -sample may not represent target population (convenience) -difficult to project sample results to target population
selection bias
-selection of individuals groups or data for analysis such that proper randomization is not achieved -sample obtained is not representative of the population intended to be analyzed
Vertical attributes
-several levels that are present can be ordered according to their objective quality from highest to lowest -it's possible to say in this case that one good is "better" than another ex: MPG, Price etc
typical focus group
-size 8-12 -length 1.5-2 hrs -data collection 1. transcribe 2. audiotape -moderator (flexible yet focused) (use interview guides) -reporting formats (selected quotations) (analysis of repeated themes)
focus groups
-solicit opinions, ideas, and emotional responses to particular topics or issues -discussion between small groups of people selected from a demographically relevant population (target market) -moderator facilitates and guides discussions -need video recording for further analysis
semantic differential scale (opposites)
-subjects asked to check the cell between a set of bipolar adjectives that best describes their feelings -most popular way in marketing research to obtain attitude toward an object -anchored at each end by an adjective --often bipolar (inconvenient-convenient) --sometimes monopolar (sweet-not sweet) -sometimes difficult to find bipolar opposites -colorful (colorless), emotional (rational), calm (excitable)
how to study attitudes
-surveys: measure of attitudes -measurement: assigning numbers to reflect the relative or absolute amount of some property --scales: creation of a continuum on which measured objects are located --scales are measurement tools -Goal is to achieve objectivity: "when description gives way to measurement, calculation replaces debate"
key concepts: sampling
-target population: entire body of interest -sampling error: from sampling -non-sampling error: systematic error
analysis of features considered jointly
-technique that enables a researcher to estimate consumers' valuations of different attributes -allows us to understand how consumers make trade-offs among attributes/characteristics of products and services -how much are consumers willing to pay/give up to get/avoid different attributes
quota sample
-to ensure representativeness (chi square) -sample is similar to population on a number of variables -age and gender disadvantages: units selected or discarded by interviewer to fit quota`
ratio scale
-type of interval scale with meaningful zero point -possible to say how many times greater or smaller one object is than another -only scale that permits comparisons of absolute magnitude what can you identify -distribution of brand preferences -most popular brand -second most preferred item for each person -mean purchase intention quantity
basis for stratification
-use criteria that are related to whatever you are measuring -ex: gender as strata for average height -income of tulane grads, by major
conjoint: pricing/valuation
-use info about consumers valuation of attributes to guide pricing strategy for a product line (iPod) -brand name equity: what's in the name?
sensitive questions
-use less threatening terminology -use counterbalancing statements -ask about others -ask after a chain of diffusing questions -ask from a range of responses (frame of reference)
conjoint overview
-useful marketing research technique to develop new products -experimental, using surveys
segmentation fails
-using same segmentation scheme for different objectives (dif. advertising or dif. products) -too much focus on techniques -using only psychographic or demographic variables -not focusing on differences in customer needs -static segmentation schemes -lack of senior managment Judgement and speculation = critical because you will never have all the info you want -usually not enough time to collect info -costs a lot to collect all info -someone will often prevent you from collecting it -consumers can't give it to you -consumers won't give it to you -collecting may tip off competitors
how to get consumer evaluations
-very nice but we don't know consumers valuations of attributes -consumers prbly don't know own values -Solution: force consumers to rank different bundles of attributes
implications for big data
-very reason why they are available is technology -users of technology are typically not representative sample of the population
trade-offs
-would consumers pay 35 cents more for a plastic bottle? -what is more important: texture or flavor? -who would be willing to have a heavier laptop in order to double the processing speed? -what do I care more for in a potential boyfriend: looks or brains?
K means: How to
1. choose number of clusters K 2. generate K random points as cluster centroids 3. assign each data point to the nearest cluster centroid 4. recompute the new cluster centroid 5. repeat steps 3-4
popular scales
1. constant sum scale 2. likert scales (agree-disagree) 3. semantic differential scale (opposites) 4. profile analysis (snake plot)
6 steps of drawing a sample
1. define the target population 2. identify sampling frame 3. select sampling plan 4. determine sample size 5. select sample elements 6. collect data from sample
examples of basis of segmentation
1. demographic characteristics 2. psychographics (lifestyle) 3. desired benefits from products/services 4. past-purchase and product-use behavior
stage 2 conjoint
1. design a data-collection procedure 2. select a computation method for obtaining part-worth
steps: cluster analysis
1. formulate the problem and select the variables you want to use as the basis for clustering 2. compute the distance between customers along the selected variables 3. apply the clustering procedure to the distance measures 4. decide on number of clusters 5. map and interpret clusters and draw conclusions
effectiveness of a segmentations scheme
1. identifiability 2. sustainability 3. accessibility 4. actionability
what can go wrong with predictions
1. initial sample bias (selection bias) 2. low response rate combined with a nonresponsive bias
types of qualitative methods
1. observation 2. focus groups 3. one-on-one interviews 4. projection techniques
why stratification
1. obtain more reliable results 2. to guarantee minimum sample sizes so analyses can be done at the stratum level
designing questionnaires
1. plan what to measure 2. formulate questions to gather the needed information 3. formatting: decide on order, wording, and layout 4. pretest 5. correct the problems
stage 3 conjoint
1. segment customers based on part-worth 2. design market simulations 3. select choice rule
stage 1 conjoint
1. select attributes relevant to product/service 2. select levels for each attribute 3. develop product bundles to be evaluated
evaluating segmentation scheme
2 acid tests -is there heterogeneity between segments? is there homogeneity within segments -are these differences and similarities based on how the customers will respond to your value proposition
recruiting
3-4 sessions 8-12 people homogenous group screening interviews incentives (if you pay) avoid professional focus group participants
observation (ethnography)
4 elements 1. directness (att and wendy's ex) 2. disguise (bose ex) 3. structure (researcher previously knows what research will occur) 4. observation system (state of Florida and tourism)
Grouping consumers
Goal: find similar and different groups -what is close and what is far
snowball sampling
I go to vegas and would like to study the nefarious business of prostitution -sample prostitutes (not legal) -sample pimps -brothels
elements of strategic marketing
Segmentation: dividing market into groups of buyers who are -similar in meaningful ways -dissimilar from other groups Targeting Product positioning
why conjoint analysis
asking direct questions about preferences often leads to unenlightening answers -what load would you like to pay on your mutual fund -what annual fee would you like -would you like online access to funds -CONSUMERS WANT EVERYTHING AND WANT IT FREE
measurement scales
assigning numbers -to subjects (customers) -regarding attributes (attractiveness) -objects (brands)
cluster analysis
a class of statistical techniques that can be applied to data that exhibit natural groupings -no distinction between dependent and independent variables -sort through the raw data on customers and groups them into clusters
cluster
a group of relatively homogeneous data points -primary input is an analysis of similarity between customers -correlation coefficients -distance measures -association coefficients
sampling frame
a list of elements from which the sample is drawn (physical representation of target population) -individuals, households or institutions -every unit is uniquely identified -all units can be found in the frame -no units not in the population of interest are in the frame -data is up-to-date
part-worth model
a simple but very useful representation of preferences -suppose there are K attributes describing the different brands. The "utility" or "worth" of any given brand j is then assumed to be total utilities ex: worth of (Austin, $130K) = Worth (Austin) + Worth ($130K)
closed ended responses
advantages -easier to answer -require less effort by interviewer -tabulation and analysis is easier -less potential error in the way the question is asked and the way it is recorded -the responses are directly comparable from respondent to respondent disadvantage -answer to a closed response will be received no matter how relevant the question is -may not produce meaningful results -dichotomus questions are prone to a large amount of measurement error -provides fewer opportunities of self expression -alternative responses not considered by the respondent, leading to selection of a reasonable response
likert scales (agree-disagree)
ask respondents the extent to which they agree or disagree with a statement -usually 5-7 point scale -typically used item by item and also a summed scale comparing two options across all measurements -easily understood but can take a long time complete when a large number of attitudes are being measured -items for measuring patients' attitudes about interactions with physicians' service staff (all with same scale)
attitudes
mental states used by individuals to structure the way they perceive their environment and guide the way they respond to it -attitudes drive behavior components: -cognitive or knowledge component -affective or liking component -intention or action component
why recovering importance works
more realistic questions -if you choose left you prefer portability, right you prefer processing power -rather than ask directly whether you prefer portability over speed, we present realistic tradeoff scenarios and infer preferences from product choices -when respondents are forced to make difficult tradeoffs, we learn what they truly value
conjoint: services
mutual fund = past returns + fees + brand name + online access value of aircraft = capacity + max range + fuel efficacy + price + service contract
constant sum scale
often used to measure importance -generally do not want to have more than 5-7 features or the allocation process gets too hard
segmentation
organizing customers into groups with similar traits, product preferences, or expectations Goal: homogenous segment
laddering example
personal value benefit product attribute
cluster sampling
population is divided into "clusters" -clusters are drawn randomly and units are observed in clusters 1. parent population is divided into mutually exclusive and exhaustive subsets -same as stratified sampling -different criteria 2. random sample of subsets is selected -sometimes there is reason to draw groups (clusters) and observe units within the sampled cluster why: 1. reduce costs of interview 2. more convenient when complete sampling frame is missing, but a sampling frame of clusters is available
stratified sampling
population is divided into "strata" a simple random sample is drawn from each subset -divided into mutually exclusive and exhaustive subsets -a simple random sample of elements is chosen independently from each subset ideal characteristics: homogenous within strata and heterogeneous across strata
sampling
population, census, and sample -population = any complete group with at least one characteristic in common -population may be studied using one of two approaches: taking a census, selecting a sample -census: survey every member of population -sampling: selection of a part of the population -both provide info to draw general conclusions about entire population standard error decreases as we increase the number of samples
consumers are heterogeneous
poses a challenge for the development and marketing of profitable products and services
relative importance of attribute
range of part-worths for attribute/sum of ranges of all attributes
laddering
seeks to discover relationship between product attributes and customer benefits and personal values 1. ask questions about attribute level -what product attributes/features are important -what product attributes/features make them buy (not buy) like (not like) the brand 2. ask why these attributes/features are important (functional and psychological consequences benefits) 3. ask why these consequences are important and what are the ultimate goals that are being satisfied
probability sampling
simple random sampling stratified sampling cluster sampling
Conjoint study process
stage 1: design conjoint study stage 2: obtain data from a sample of respondents stage 3: evaluate product design options
cluster v. stratified sampling
strata- all groups included; homogenous cluster- random selection of groups; heterogeneous
structured & undisguised
typical questionnaire: what is your age +easy to store in a dataset +frame of reference is clear from alternatives +alternatives help to make question clearer -loss of info through response categories: -forced to reply -no accurate category listed -no opinion
why use laddering
ultimately, consumers want to know what's going on -enhance communication strategies -enhance product development
projective techniques
vague stimulus = unbiased consumer response -word association (brand names, slogans) -sentence completion -storytelling -respondents projected into simulated activities -useful when respondents may be hesitant/unable to answer direct questions
pretesting
variation meaning task difficulty respondent interest and attention -test flow of questionnaire for clarity and logic -ensure that skip patterns are clear and well laid out -time each section so it's not too long -capture and maintain respondent interest and attention
conjoint: directly asking can also be misleading
when people are directly asked, they want -durability + quality + reliability but from conjoint consumers really care about -Price + Design
unstructured & disguised
when respondents have reason to hide their feelings or when they cannot express them accurately 1. ask respondent about what others think 2. projective techniques: respondent is given a vague task, response is used to deduct respondents feelings