Door-to-Door Interviews

Interview people at their home

Popular before the 70's

Intimate, comfortable atmosphere

now, a lot of rejections and vert costly

Still used in underdeveloped countries

Executive Interviews

like door-to-door interviews, only at businesses

expensive

usually must qualify leads by phone first

Business people usually enjoy this

Business people are also too busy and cancel a lot

Standing in a mall and asking random shoppers

Sometimes permanent facility

cannot get a large metropolitan sample

More likely to get frequent shopper personalities

less expensive because respondents come to interviewer

until the 90's it was he most popular form 

inexpensive

less listed numbers and landlines

a lot of people screen calls now

conducted from a facility

controlled and monitored environment

CATI

Computer-Assisted Telephone Interviews

surveyor and computer

computer does all the work, surveyor simply reads and keys in the responses

more accurate results

no human or computer is involved

no one is present to explain things/probe response

no one is present to generate bias

ad hoc/one shot: selected sample of names

mail panels: pre-contacted and pre-screened

longitudinal study:questions the same respondents more then once

no interviewer present

real time reporting

reduced cost

personalization

high response rates

can contact hard-to-reach

simplified/enhanced panel management

Little understanding to the demographic 

Cannot reach all audiences

Does not work in all countries

Poor security

unrestricted Internet sample: self-selected respondents

Conducting Methods of

Online Surveys 

Web Survey System: software

Survey Design and Web Hosting Sites: online program. Do not need to download. Ready made templates.

Internet Samples of

Online Surveys

Screened Internet Samples: selecting group based on demographic characteristics

Recruited Internet Samples: Controlled. ensures target population

Recruited Panels: (popular) target consumer reached through ad database

Renting Internet Panels: paying a fee for panel

Random Website Intercept: ad's on other webpages

Data Capture of Visitors: internal database of visitors

IMRO

Internet Market Research Organization

development

dissemination

implementation

interactive research concepts

practices

information

high level of accuracy

vs

cost

validity: degree a measure reflects characteristic of interest

reliability: consistency of a measure to produce same results

survey structure

6-10 minutes

or

no more then 20 questions

the number of people that would qualify for a study

cost to find can exceed cost of surveying

a result of chance variation

the difference between the sample value and true value

cannot be avoided, but reduced with larger sample size

mistakes or problems in the research design

sample design - frame - population specification - selection - measurement - surrogate information - interviewer - measurement instrument - processing - nonresponse bias

a discrepancy between the information required to solve a problem and the information being sought.

general problems in the research design/properly define the problem

not everyone is going to respond to your survey (5% response rate)

Are those who did not respond different then those who did?

the greatest population all answer one question in the same way

deliberate falsification

or

un conscious misrepresentation

a set of questions

collects data

to accomplish an objective

provides decision making information

logical to the respondent

easily coded

1.objectives, resources, constraints
2. data-collection method (Internet, telephone)
3. types of questions
4. wording of questions
5. flow and layout
6. evaluate the effectiveness
7. get approval
8. pretest and make changes
9. prepare final copy
10. execute/implement

open ended

closed ended

dichotomous - 2 categories (ex. yes/no)

multiple choice

scaled response

Standard Questionnaire layout (9)

In order(6)

Design(3)

Screening questions

Interest generating

general/warm up

"work"/middle - gain commitment

prompters/strategic - encourages to continue

position sensitive/demographic (last)

space for open ended responses

instructions in CAPS

introduction and conclusion

support services:

formatting, screener writing, coordination and collection of data etc.

cost effective

increase productivity

do not have design and analytical capabilites

bidding war

questionnaire + incidence rate determine cost

Causal Research

Causation

a change in one variable CAUSES a predictable change in another

1. concomitant variation (correlation)

2. Appropriate time order of occurance

3. elimination of other causal factors

positive - advertising and sales

inverse - price and sales

By itself, does not prove causation.

In Field

Laboratory

Internal - measures what it intends to measure (competing explanations ruled out)

External - can be generalized to the public, set & times

(is it inter-changable to represents others)

X - exposure to experimental treatment

(what is being measured and compared)

O - observation (study of the test groups)

Horizontal line - timing

Vertical line - simultaneous

History*

Maturation*

Instrument Variation

Selection Bias

Mortality*

Testing Effects

Regression to the mean

Randomization

Physical Control

Design Control

Statistical Control

Treatment - manipulated variable

Subjects - participators

Dependent - measured variable

Plan/Procedure

Test group

&

Control group

test - exposed to manipulated variable

control - no change of variable, for comparison

Pre-Experimental

One-Shot Case Study

True Experimental

Quasi-Experiments

offer little or no control over extraneous factors

cheap

lead to new hypotheses 

exposing test units to treatment variable

then observing

one-group Pre-test-post-test design

observing - exposing - observing

randomly assigned treatments to randomly selected test units (R)

more valid results

before and after with control group design

random assignment, control group & pre+post measurement of both groups

after only with control group design

random assignment, control group & post measurement of both groups

lack of control over scheduling treatments

OR

must assign treatments in non-random

Interrupted Time-Series Designs

repeated measurement (before&after) "interrupts previous data patterns

Multiple Time-Series Designs

interrupted time-series designs + control group

estimate market share volume

cannibalization rate (effects on current products)

characteristics of consumers/refine marketing

behaviour of competitors

Standard (traditional)

Scanner (electronic)

Controlled

STM (simulated)

Decision to conduct test marketing

4 factors

cost and risk of failure vs probability of success

likelihood +speed competitors can steal

investment needed to produce product

damage unsuccessful product could do to the firm

1. Define Objective

2. Select a basic approach

3. develop detailed test procedures

4. select test markets

5. execute plan

6. analyze results

2+ test markets (existing) --- 3+ test (new)

geographically dispersed

demographically representative

6-12 month test period

variety of media outlets

dominant newspaper

reflective of population

Purchase data

Awareness data

Competitive data

Source of sale

1 region

follows pretest

collect data within a matter of days

then launch in additional regions

try product in foreign market before rolling out globally

High cost

Security

Implementation problems

obtaining information from a subset of a larger group

sample - the subset of all the members 

census - the data obtained 

Steps to developing a sample plan (7)

1. Define the population

2.choose data-collection method

3.identify sample frame

4. select sampling method

5. determine sample size

6. develop operational procedures for elements

7. Execute

Geographic Area

Demographics

Usage of product

Awareness of product

Probability: a known non-zero likelihood of being selected

Non-Probablility: specific elements of the population have been selected 

Systematic

Stratified

Cluster

Simple Random

Convenience

Snowball

Judgement

Quota

sampling error - does not represent population

(types: administrative & random)

non-sampling error - any other error

(also known as measurement error)

population is split into two or more mutually exclusive subsets. (like male and female)

THEN number is assigned and they are chosen at random

This sample has greater statistical efficiency 

smaller sampling error

1. identify demographic/classification factors

2. determine what population falls into what subgroup

3. select simple random samples from each stratum

sampling units are selected from a number of small geographic areas to reduce data-collection costs

1. population divided into mutually exclusive and exhaustive subsets

2. random sample of the subset is selected

one-stage

two-stage

multi-stage (national)

The Data Analysis Procedure

(5 Steps)

1. Validation and Editing (quality Control)

2. Coding

3. Data Entry

4. Machine Cleaning of Data

5. Tabulation of Statistical Analysis

Validation of Data

(not to be confused with validity, that what was being measured actually was measured)

Interviews were conducted as specified

to detect interviewer fraud or failure to follow instructions

contacting a percentage of participants to confirm validation

1. List responses

2. Consolidate Responses (congregate answers that mean the same thing)

3. Set codes

4.Enter codes

One-Way Frequency Table - a table showing the number of respondents choosing each answer

Cross Tabulations - Examination of response to one question relative to the response of another question

Mathematical Differences

Statistical Significance

Managerially important differences

1. Stating the hypothesis

2. Choosing the appropriate statistics

3. Developing a decision rule/ standard

4. Calculating the Value of test statistics

5. Stating the conclusion