Market Research test 1

Exploratory research

flexible, preliminary investigation into an under-researched topic, gathering initial information

Objectives of exploratory research

• discover initial ideas and insights

• generate possible explanations for phenomena

• establish priorities

• develop hypotheses

Is exploratory research mainly quantitative or qualitative? What kinds of methods are used?

Qualitative, methods such as:

• lit review

• case analysis

• projective techniques (sentence completion, word association, etc)

• focus group interviews

• in depth interviews

• observational techniques

Focus groups

a small number of individuals are brought together in a room to sit and talk about a topic of interest to the focus group sponsor

• 1 moderator

• dynamic, conversational vibe

• normally 8-12 participants or 6-8 for deeper discussions

What should the makeup of a focus group be like?

• diverse, but not too diverse that everyone feels lonely and seperate

• strangers, with some things in common

• not dominated by 1-2 people

Possible objectives of focus groups

• generating ideas

• understanding consumer vocabulary

• reveal consumer needs, motives, perceptions, etc

• develop questionnaires for larger scale research

Common mistakes of focus group moderators

• only moving in the realm of rational declarations of the participants

• collecting opinions but not considering root cause

• questionnaire type interview (pretty closed)

• inability to control the group

• too dominant over the group

  • becoming a participant, not a moderator

In depth interviews

one on one interviews aiming to obtain detailed insights on consumption

• can take in home, business, point of consumption, etc

When are in depth interviews particularly useful?

when the population is unknown, and we need to gain preliminary insights to nail down the target pop.

Projective techniques

understanding respondent’s deepest feelings by having them project those feelings onto something unstructured (sentence completion, story completion, word association, etc)

When are observational methods of research useful?

if focus groups & interviews seem too obtrusive, or you feel like respondents may not be truthful

Path tracking

form of observational research that tracks the path that customers take when walking around the store

Facial encoding

a form of observational research where marketers can measure human emotion through facial expressions (algorithms trained to recognize expressions/emotions)

Eye tracking

a form of observational research that uses invisible near-infrared light and high-def cameras to project light onto the eye and record the direction it’s reflected. Can be screen based, wearable, etc

FMRI

a form of observational research where magnets track changes in blood flow across the brain (brain activity can be predictive of future action)

EEG

a form of observational research that reads brain cell activity using censors placed on the scalp

Objectives of descriptive research

• describe characteristics of groups of consumers

• estimate the proportion of people who behave in a certain way

• make predictions

• etc

Cross-sectional study

the most frequently used form of descriptive research in marketing, analyzing data from a population at a specific point in time

Longitudinal study

repeatedly observing/collecting data from the same individuals over a long period of time

Panel

a sample of brands, consumers, stores, firms, etc

Panel data

repeated measures of certain variables of panel entities over time

Big data

originally referred to a set of data management technologies, first employed by info tech companies and social media firms including Google, Facebook, and Yahoo to enable the processing of massive volumes of data in a timely fashion

The three Vs of big data

Volume: walmart collects millions of gigabytes of customer data per hour

Velocity: data streams at unprecedented speeds and must be dealt with in (near) real time

Variety: 80% of data is unstructured (messages, GPS, readings from scanners, etc)

Causality

change in one variable (x) produces change in another variable (y)

x cause, y effect

3 conditions for causality

1. correlation between x and y

2. x should proceed y

   1. x is the only factor that affects y (elimination of alternative explanations)

Counterfactual condition

the unseen condition, what would have happened under a different scenario (ex: if Johnny was shown commercial A instead of B, but if we show him one we can’t know how the other would have affected him)

Why randomization is important

it ensures that groups are on average similar, so we can compare groups and not individuals H

Hypothesis

unproven propositions about some phenomenon of interest (can be determined by statistical procedure)

Null Hypothesis

H₀, the hypothesis that a proposed result is not true for the population (statement of no difference/no effect)

Alternative hypothesis

H_a, the hypothesis that a proposed result is true for the population (opposite of the null hypothesis)

not accepted w/o convincing evidence, it’s what the client wants to prove

Hypothesis testing

statistical procedure that determines which of the two hypotheses is statistically true

select a sample, calculate a relevant sample statistic, and derive its probability distribution

p-value

the probability of obtaining the observed result under the null hypothesis by coincidence

result is statistically significant if the p value is less than the chosen threshold (usually 0.05)

A/B testing

an example of a randomized controlled experiment: participants are randomly assigned to 2 groups, A and B, and receive different treatments

proves or disproves causality between marketing activities and market outcomes

Hypotheses for a T-test

H₀: the two population means are not different

H_a: the two population means are different

Some limitations of A/B testing

• You can’t correct a decision once it’s made

• It can’t adapt to changes in a dynamic environment

• If the test period is too short it might be inaccurate, while too long is a waste of money

  • Heterogenity; different user segments may react differently to changes, but these differences would be masked with one single A/B test

Quasi-experimental design

the use of methods and procedures that appear similarly structured as a randomized control trial, but the study lacks random assignment, includes preexisting factors, or does not have a control group

Matching

lining up twins in two groups to make them comparable and control for possible confounding variables

Stratification variable

dividing a population into distinct, non-overlapping subgroups for analysis

Natural experiments

observational study that leverages naturally occurring situations or events rather than manipulated interventions to investigate causal effects of an independent variable on a dependent variable

ex: sudden social media outage, other unexpected events

Internal validity

the extent to which the observed results are due to experimental manipulationH

History

a threat to internal validity: specific events that are external to the experiment but occur at the same time, may affect dependent variable

ex: testing a new curriculum, but a widespread computer crash impacts studying

solution: two group before/after design. that way if history effect occurs, it affects both groups and evens out

Stable Unit Treatment Value Assumption (SUTVA)

foundational, implicit assumption in research with 2 main components:

1. No spillover/interference between units: treatment of one unit does not affect the outcome of another

   1. No hidden variations of treatment: there’s only one version of each treatment level, so the potential outcome doesn’t depend on how treatment was administered

Statistical regression effect

if a variable is extreme on 1st measurement, it will probably be closer to the mean on the second measurement, regardless of treatment (sports illustrated curse)