MGT 453 - Marketing Research Design Lecture 3 PDF

Marketing Research Lecture 3: Marketing Research Design Professor Landry  Reminders/Announcements Next week: group project proposals due In two weeks: online class (pilot surveys) zoom link on quercus In three weeks: midterm  Online Pilot Surveys In two week...

Marketing Research Lecture 3: Marketing Research Design Professor Landry  Reminders/Announcements Next week: group project proposals due In two weeks: online class (pilot surveys) zoom link on quercus In three weeks: midterm  Online Pilot Surveys In two weeks, each group will give an online pilot survey during class Pilot survey is used to test potential questions for your actual survey Each survey should take 5 minutes or less to complete Come ready with a shareable link to paste in the chat box Platforms: Google Forms, Microsoft Forms, Qualtrics, SurveyMonkey…  Online Pilot Surveys You will be asked to summarize/tabulate the results of your pilot survey in the appendix of the final group project report Review pilot survey results before you design your final survey – this will help you write better survey questions Unlike the final survey (which will held to a high standard), the pilot survey can be “rough”!  Marketing Research Design Research design: a plan that specifies the methods and procedures for collecting and analyzing the information that is needed to accomplish the research objectives 1. Gain background information and develop hypotheses exploratory 2. Measure the state of a variable of interest descriptive 3. Test hypotheses about relationships between variables causal  Exploratory research is usually conducted early in the marketing research process, when the researcher may not know that much about the problem at hand Vacation Package A: New Brunswick, Canada Vacation Package B: Orlando, FL, USA 1. Magic Kingdom (Cinderella's Castle) 2. Epcot Center 3. Animal Kingdom 4. Hollywood Studios fly 1st class, direct, 4-star hotel fly economy, 2 layovers, 2-star hotel avg. May temp: 10 C avg. May temp: 26 C “Laddering” Interviews… Laddering Aim: to uncover the reasons underlying the preferences and choices of consumers Exploratory research is usually conducted early in the marketing research process, when the researcher may not know that much about the problem at hand Laddering: an interview technique that aims to understand why individuals choose certain products and/or services by identifying the relevant attribute(s) of the product/service, the benefit(s) of those attribute(s), and the individual value(s) underlying the benefit(s) Other Exploratory Research Methods Secondary data analysis: Experience surveys: gathering searching for and interpreting information from those thought information from secondary to be knowledgeable on issues data sources (e.g. using google) relevant to the research problem Focus groups: small groups Case analyses: reviewing any assembled and guided by a available information about a moderator through unstructured previous situation similar to the and spontaneous discussion current research problem Descriptive research: used to answer questions that begin with who, what, where, when, and how (…but not why) In longitudinal research, a collection of sample units who have agreed to answer questions at periodic time-intervals is referred to as a panel  Types of Panels in Longitudinal Research: Omnibus: subjects asked different questions at each measurement Also called “discontinuous panels” Continuous: subjects asked same questions at each measurement Goal: track changes over time in attitudes, behaviors, etc. Brand-tracking used to detect shifts in brand preference over time  Brand-Tracking Example Preferred Sweater Brand (%) Suppose this is continuous panel data Brand 2015 2025 Should A&F be more worried about the shift in preference towards AE 55 20 or the shift towards Polo? How might A&F’s interpretations change if the data came from two 25 38 separate, cross-sectional studies? 20 42  Brand-Tracking Studies with Continuous Panels Brand-tracking: Advantage versus repeated cross-sectional studies: form of longitudinal if brand preferences vary between measurements, research that uses it’s not due to random variation in different samples, we know individuals’ preferences actually changed continuous panels; frequently applied to Disadvantage: changes in brand preference over detect shifts in brand time could be attributable to age or changes in other preference over time characteristics correlated with age (e.g. income) Attrition in Longitudinal Research Attrition: the loss of subjects from a panel Attrition rate in longitudinal studies ≈ 17% (Capaldi & Patterson, 1987) Common reasons: subjects’ unwilling to continue, can’t reach subjects for follow-up (e.g. subject changes address or phone #), death Why attrition is a big problem in longitudinal research… Reduces sample size → reduces statistical accuracy (unless replaced) Can make the sample less representative of the relevant population Q: Should we be more concerned about attrition from continuous or omnibus panels? A: continuous panels… but why? Early observations from a continuous panel subject are less useful if later observations can’t be collected With omnibus panels, however, a subject who drops out can be replaced (although representativeness is still an issue) and earlier observations are just as useful because they generally aren’t compared with later observations Q: If we want to collect different info from the same population, but for different purposes at different times, why would we ever use an omnibus panel when we could do separate cross-sectional studies? A: While creating an omnibus panel is costly if you’re just taking one “snapshot,” it has several advantages when used repeatedly: Data collection is fast and easy once in place since members’ contact info already on file → recruitment efforts not necessary Subjects already agreed to provide info → better response rates Also worth noting: omnibus panels are often given surveys from multiple firms for unrelated marketing problems Q: How can omnibus panel data be less reliable than cross-sectional data? Assume sample sizes are the same and both samples are matched demographically to the relevant population (using same age cohorts) A: Omnibus panels may be unrepresentative in other ways since members are disproportionately willing to serve on the panel and be tracked by – and repeatedly provide info to – researchers Especially if willingness to be on panel is correlated with variable of interest e.g. Apple is deciding whether to make “on” the default for the next iOS location services, wants to gauge consumers’ willingness to share location data Omnibus panel members may be more open to sharing location data → not as representative as subjects (who can be anonymous) in a cross-sectional survey  Causal Research Causal research: used to test hypotheses of the form if x, then y i.e. whether a change in one variable causes another variable to change To uncover causal relationships, researchers use experiments  Causal Experiments: Testing “if x then y” In an experiment, an independent variable is “manipulated” to see how it affects a dependent variable, while “controlling for” extraneous variables Independent (the x’s): variables that the researcher can control and wants to see how they affect another variable; often refer to a ‘P’ Dependent (the y’s): marketer has little or no direct control over but an interest in changing; usually a measure thought to causally increase profits Extraneous: may affect dependent variable but not an independent variable  Experimental Design Experimental design: create setting where a change in dependent variable can be attributed to a change in the independent variable (and nothing else) We want to be as sure as possible that any change in y was truly caused by x “Lab” vs. “Field” experiments…  Experimental Design Before-after with control group: experimental design in which subjects are randomly divided into two groups: Treatment group (aka experimental): members are exposed to a change in the independent variable Control group: not exposed to change in independent variable By comparing how the dependent variable changes from “before” to “after” in each group, we can estimate the effect of x on y… Experimental effect: 𝑬 = 𝑻𝟐 − 𝑻 𝟏 − 𝑪𝟐 − 𝑪 𝟏 Dependent variable measurements: 𝑻𝟏 – treatment group before change in independent variable 𝑻𝟐 – treatment group after change in independent variable 𝑪𝟏 – control group before change in independent variable* 𝑪𝟐 – control group after change in independent variable* Due to unforeseen circumstances, the vacation packages can now only be offered in August instead of May  NB FL May 10 26 Avg. Temp. (C) Aug. 17 29 May 58 18 Avg. Precip. (mm) Aug. 35 94 Group O fly 1st class, direct, 4-star hotel fly economy, 2 layovers, 2-star hotel avg. temp. in August: 17 C avg. temp. in August: 29 C Group E fly 1st class, direct, 4-star hotel fly economy, 2 layovers, 2-star hotel avg. temp. in August: 17 C avg. temp. in August: 29 C In what we just did… 1. What’s the independent variable of interest? 2. What’s the dependent variable of interest? 3. What are the extraneous variables (if any)? 4. Who is in the control group? 5. Who is in the treatment group? Experimental effect: 𝑬 = 𝑻𝟐 − 𝑻 𝟏 − 𝑪𝟐 − 𝑪 𝟏 Dependent variable measurements: 𝑻𝟏 – treatment group before change in independent variable 𝑻𝟐 – treatment group after change in independent variable 𝑪𝟏 – control group before change in independent variable* 𝑪𝟐 – control group after change in independent variable*  Internal validity asks if experimentally- observed change in dependent variable was actually caused by the independent variable In doubt when control and treatment groups Internal Validity systematically differ on some extraneous variable of Experiments (possibly unknown) that could be responsible for observed change in the dependent variable Can never guarantee internal validity, but random group assignment goes a long ways towards alleviating potential concerns of internal validity Lab experiment > Field experiment  External validity asks to what extent does the experimentally-observed relationship between independent and dependent variables generalize to real-world settings? External Validity Field experiment > Lab experiment of Experiments Can refer to generalizability across… Population units (were subjects representative?) Time Situations…  Construct Validity Construct validity: does the dependent variable predict the real-world variable that it is supposed to? e.g. do preferences in blind taste tests predict brand preferences in purchasing behavior? New Coke beat original Coke 55-45, Pepsi 54-46 in blind taste tests in 13 cities with nearly 200k subjects Conducted in the field (grocery stores & other vendors) New Coke replaced original Coke in 1985, but failed miserably and quickly was taken back off the market  Construct Validity Were the New Coke taste tests valid? Internal validity? Yes, new recipe was responsible for preferences in blind taste tests (huge sample size too: 200k) External validity? Yes, field experiments in 13 cities with nearly 200k consumers Construct validity? No, preferences in blind taste tests did not translate to preferences in buying behavior Test marketing: study conducted in a field setting Usual connotation: setting is an actual marketplace Main uses of test markets: Test sales potential for a new product or service Test variations in the marketing mix for a product or service Types of Test Markets Standard: firm tests the product or other marketing- mix variable through an existing distribution channel Controlled: outside research firm distributes the product through pre-specified types and numbers of distributors  new handlebars Standard Standard HD could use or a distribution channel that it Controlled? already uses  a new cake decorating kit Standard Controlled HD would need or a distribution channel that it Controlled? doesn’t already have in place  Other Types of Test Markets Electronic test markets consumers in a panel carry identification cards that they present when buying goods and services Simulated test markets (STMs) carry out test marketing via a computer program  Selecting Test-Market Cities Three main criteria: 1. Representativeness 2. Degree of isolation 3. Ability to control distribution and promotion  Selecting Test-Market Cities Canada’s test-market “capital”: First mini Ikea store in N. America First Big Mac, Chicken McNuggets in Canada (McDonalds) First ATM in Canada (Canada Trust) First Tim Horton’s Dark Roast Coffee, Iced Cappuccino First Satellite TV (Bell Canada)