Strategic Market Research by Dr. Anne E. Beall Prepared by Matthew A. Gilbert, MBA Chapter 7: Analyzing Numeric Data to Determine What Drives Markets
Nov 01, 2014
Strategic Market Research
by Dr. Anne E. BeallPrepared by Matthew A. Gilbert, MBA
Chapter 7: Analyzing Numeric
Data to Determine What
Drives Markets
Hypothesis-Driven Analyses
Testing hypotheses works better than just expecting data to magically provide an answer.
Process Begin with hypotheses.
Create a set of cross-tabulations that test the assertions.
Divide the data by the number of categories.
Review data for each of your groups.
See if hypothesis are supported.
Predicting Perceptions & Behavior
Only way to accurately identify if something has an effect is to use experimental design. Experimental Design: Takes an independent variable and
manipulates it in some way and then measures the effect of this manipulation on a dependent variable.
When the only thing that changes is the independent variable, any effect on a dependent variable is due to that independent variable.
Predicting Perceptions & Behavior
Example: You can measure the effect of color on the attractiveness of a specific car by manipulating only the color of a car and measuring perceptions of the attractiveness to see if the color has an effect. You might learn that color has a large effect on perceptions
of a VW Bug and that the car is more attractive when it is shown in unusual colors.
The “Smile” Experiment
Same photos of people created with “regular” and “beautiful” smiles. Respondents saw only one.
Respondents rated attractiveness, intelligence, happiness, career success, friendliness, kindness, wealth, popularity with opposite sex, sensitivity.
Results: When people have a beautiful smile with white, straight teeth, people perceive them as more attractive, intelligent, happy, successful, friendly, interesting, kind, wealthy, popular and sensitive.
The “Smile” Experiment
Correlation Analysis
In cases where you cannot do an experiment, you can learn which variables are related by using a correlation analysis.
You can determine if gender, age, or income are highly correlated with purchasing a new TV by correlating these variables with actual purchases or respondents’ stated intention to purchase the TV.
Correlation Analysis
Low correlations indicate that the two things are not related to one another and that one variable does not vary with the other.
High correlations indicate the opposite.
Important: Correlations do not indicate causation. Just because two things are highly correlated does not
mean that one causes the other.
Having a high income does not cause a person to be interested in buying a new television.
Determining Major Segments
Another major question organizations want to answer is what the major segments in a given market are. Organizations want to understand which specific groups
are most likely to purchase their products and why.
They want to learn how large these groups are, what they are like demographically, and how to communicate with them effectively.
Segmentations are a way of describing a market as well as a way of providing direction for an organization’s marketing efforts.
Segmentation
Segmentations should be customized for a specific company’s product or service.
Unless a segmentation is based on data collected for that specific category, it won’t be very useful.
Example: If you want to know about the segments in the dog food market, conduct a segmentation study specifically about dog food and the major brands. Don’t buy secondary segmentation that was designed for
all businesses and expect to figure out the segments of dog-food buyers.
Segmentation
Example: Cell Phone Segmentation Client wants to identify people who use a lot of cellular
service and who are most likely to purchase their brand.
Wants to offer certain packages to these individuals.
Conduct an analysis that determines if age is highly related to cell phone usage and that younger consumers use cell phone service the most.
Might determine that what predicts brand usage among young users is the cell phone service their parents use.
Predictor of brand usage among older users might be the brands their friends are using.
Segmentation
Example: Cell Phone Segmentation (Continued) Might wind up with a segmentation as follows:
1: Under 25 whose parents use Carrier A
2: Under 25 whose parents use other carriers
3: 30 to 45 whose friends use Carrier A
4: 30 to 45 whose friends use other carriers
5: 40 to 55
6: 51 to 64
65+
Segmentation
Example: Cell Phone Segmentation (Continued) Then profile the groups to determine if their behavior and
attitudes are what we predicted
Assume that predictions are borne out in data and that Segment 1 had the highest usage and greatest loyalty toward Carrier A.
Segment 1 is followed by Segment 3.
Segmentation
Example: Cell Phone Segmentation (Continued) Company now has clear direction on which segments are
most valuable to them and which ones they should target.
Additional profiling might reveal greater details to help design products and services targeted segment wants.
Segmentation
Regional Bell Operating Companies (RBOC) Local clients who wanted to enter long distance market.
Deregulation was in infancy and there were special rules about data usage: local phone companies not allowed to use their own data to target customers for long distance.
Segmentation
Regional Bell Operating Companies (RBOC) Problem: Any segmentation had to use data that could be
purchased from external data vendors. Clients purchased large databases of information for a geographic area and assign every household in the database to a segment.
Solution: Created an algorithm to assign everyone in the database to a specific segment. After all households were assigned to a segment, clients would contact people with an offer for long distance service. Thinking through how segments would be used and what data constraints there were for each was very important.
Segmentation
Overlapping Segmentation When a segment needs to serve different purposes for two
different groups in an organization.
Present segments in a way to helps to visualize where and how they overlap. (See Figure 2 on Page 61).
Determining the Best Configuration
Question: How can organizations identify what the best configuration is for their product or service?
Answer: Use conjoint or discriminant analysis to identify the best configurations for a certain group of people or for a market overall. Benefit: Enable us to figure out what the potential demand
is for a large number of product configurations.
Discriminant: Enables you to control some configurations.
Conjoint: Doesn’t enable you to control configurations.
Determining the Best Configuration
LCD Display (Attribute) Large (Level 1)
Small (Level 2)
Alarm Music
Music or Loud Sound
Music, Loud Sound or Bright Light
Determining the Best Configuration
CD Player Present
Not Present
Price $35
$30
$25
Determining the Best Configuration
Then show different configurations to respondents and ask them how likely they are to buy each one.
Create a “Simulator” which represents results in a tabular format (See Table 2 on Page 63)
Other Analyses
How related are two variables?
Which set of things is most related to a variable?
Are there significant differences between groups?
Do specific groups differ in a set of attitudes or group of behaviors?
Are preferences for products or services similar or different than what would be expected by chance?
What is the underlying structure of a set of attitudes?
What are the natural groups in the market?
What is the optimal configuration for a product or service?
Other Analyses
How related are two variables? Correlation
Which set of things is most related to a variable? Regression Analysis
Are there significant differences between groups? t-Test, Analysis of Variance (ANOVA, f-test)
Other Analyses
Do specific groups differ in a set of attitudes or group of behaviors? Mutlivarate Analysis of Variance
Are preferences for products or services similar or different than what would be expected by chance? Chi-Squared Analysis
What is the underlying structure of a set of attitudes? Factor Analysis
Other Analyses
What are the natural groups in the market? Cluster Analysis
What is the optimal configuration for a product or service? Conjoint of Discriminant Analysis
Other Analyses