Download SV Regression - Vision Critical Intranet

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Value proposition wikipedia , lookup

Predictive analytics wikipedia , lookup

Sensitivity analysis wikipedia , lookup

Price discrimination wikipedia , lookup

Transcript
Advanced Analytics Training: Sydney
Part I: Basic Analytical Methods
Jane Tang
December 14, 2010
Analytics
• Small group:
– Jane Tang (Calgary)
– Rosanna Mau (Vancouver)
– Tanveer Husnani (Toronto)
• Handle all of analytics within VC: anything beyond
data tables:
• Analytics
• Reporting Tools
• Statistical Support for ….
How we do it?
- Up front quote on cost based on available information
- Your project is charged for our time only
- You get the mark up for analytics
- Internal cost: $150/hr, external cost: $250/hr
- Consultative process
- We work with you and with your client to design the sample
and questionnaire, and to find the statistical methods to
best answer your research question
- We are up front about our methods. There’s no black box.
We are happy to provide details on any of the statistical
methods we utilize short of giving you the actual code.
- We work with you to find the best way to present the
results.
3
How we can help
- Statistical Models:
Correspondence Map
TURF/Shapley Value
Paired Comparison
Choice Based Conjoint
(CBC)
van Westendorp Pricing
Analysis
Q-Sort
MaxDiff Experiments
Discrete Choice Model
(DCM)
Segmentation
Typing/Assignment
Similarity Sort
Convergent Cluster
Analysis
Consumer Decision Tree
Thurstone Case V
Shapley Value
Regression
SEM/Path Models
Finite Mixture/Latent
Class Segmentations
Key Driver analysis
-
Quick & Easy
Complex
3-5 hrs
50+ hrs
Intranet: search for Analytics
-
http://intranet/display/research/Analytics
4
Van Westendorp
Correspondence Analysis & Map
Thurstone
TURF & SV
Driver analysis & SV regression
Case Study: Brand+
Van Westendorp
Van Westendorp - Overview
• Price Sensitivity Meter.
– Gauges thresholds and provides a range of acceptable prices.
– There is a relationship between price and quality and that
consumers are willing to pay more for a higher quality product.
– Consumers have ideas about what reasonable price they are
willing to pay for a product and that consumers are willing to pay
more for a high quality product.
– Easy to use. Good for early stage pricing model.
– Not in a competitive context.
– Recommend sample size: n=150-200. Minimum: n=100
Van Westendorp - Questionnaire
• Price Sensitivity Meter: 4 questions to the questionnaire.
• By aggregating the proportion of people (at each of these prices),
we can develop a series of curves that will indicate the range of
acceptable prices in the market place.
– At what price would you consider the product to be getting expensive,
but you would still consider buying it? (EXPENSIVE)
– At what price would you consider the product too expensive and you
would not consider buying it? (TOO EXPENSIVE)
– At what price would you consider the product to be getting inexpensive,
and you would consider it to be a bargain? (BARGAIN)
– At what price would you consider the product to be so inexpensive that
you would doubt its quality and would not consider buying it? (TOO
CHEAP)
Van Westendorp - Questionnaire
•
•
•
Newton/Miller/Smith Extension.
With the addition of two purchase probability questions (at the BARGAIN
and EXPENSIVE price points), it is possible to plot trial and revenue curves.
These curves will indicate the price that will stimulate maximum trial and the
price that should produce maximum revenue for the company.
– At the (expensive price) how likely are you to purchase this product?
– At the (bargain price) how likely are you to purchase this product?
van Westendorp Output - PSM
Product Price - Van Westendorp Price Curves - (n=100)
100%
MGP=$9.06
MDP=$11.61
Percent of Respondents
75%
Too Cheap
Cheap
50%
Expensive
Too Expensive
IDP = Indifference Price Point
OPS = Optimal Price Point
MGP = Marginal Point of Cheapness
MDP = Marginal Point of Expensiveness
IDP=$11.18
25%
OPS=$9.21
0%
$0.00
$5.00
$10.00
$15.00
$20.00
Price ($)
$25.00
$30.00
$35.00
$40.00
van Westendorp Output – Trial &
Revenue
Price - Trial and Revenue Curves - (n=100)
18%
$25,000
16%
MaxTrial=$12.00
$20,000
MaxRevenue=$12.00
14%
$15,000
10%
8%
$10,000
6%
4%
$5,000
2%
0%
$0.00
$5.00
$10.00
$15.00
$20.00
Price ($)
$25.00
$30.00
$35.00
$0
$40.00
Revenue Per Hundred People
Likelihood to Buy (%)
12%
Trial
Revenue
Correspondence Maps
Correspondence Maps Background
• An approach to portray categorical data in multiple
dimensions (maps).
• Appropriate for frequency data
• Works with checklist or “top box” data
• Brand Association Grid
• Can combine data gathered in different
questionnaires
• Shows both brands and attributes in the same plot
• Visually displays relationships in the table
• Part of the Brand+ Deliverable
– Can be combined with Driver analysis results.
Emotional Response to Federal Party Leaders—National
Here is how the public’s perceptions define our current market
Size of type indicate
importance of attribute. Larger is more important
This is the “white space”. These
are the key drivers of voting,
and there is opportunity for a
new party leader to claim this
space as May and Layton don’t
“own” these attributes
Sadness
Submission
Dion
Shame
Surprise
Wo nder
Anticipation
Displeasure
What we asked:
Canadians were
Asked to “Please
click up to 4 words
that describe
feelings you have
about [NAME]
leader of [PARTY].”
These official
portraits were
shown on the page
with the list of
emotions.
May
Acceptance
Contempt
Fear
Harper
Optimism
Layton
Disgust
Duceppe
Joy
Anger
Aggressiveness
Love
About this map:
The map was
created using
correspondence
analysis. Words
that are near each
other, and near the
pictures, are related
mathematically—
based on an
analysis of
Canadians’ answers.
Interpreting Correspondence Maps
• In general, the distance between two points
describes the strength of the association
• Brands are displayed as points in the space
• Attributes are displayed as points in the space
• Correspondence Analysis (CA) maps can be
used to show how brands are positioned
• They can demonstrate opportunities for brand
re-positioning
Thurstone Case V Analysis
Thurstone Case V Background
•
•
•
A Thurstone chart is based on the construction of a uni-dimensional
interval scale.
Using Thurstone Case V analysis, it is possible to convert rank order
preferences into interval scale data.
– Thurstone Case V analysis goes beyond a simple order of
preference.
– It shows how much more each product is preferred, relative to the
most preferred and least preferred product/attribute.
The Thurstone scale is developed from preference judgments provided
by the respondents.
– For example, a person may prefer a Lexus to Ferrari and a Ferrari
to a Volvo. These simple preferences can be converted to interval
scale (That is, we are able to determine, how much greater is the
preference for a Lexus over a Ferrari than a Ferrari over Volvo?).
– These preferences are then aggregated (proportion of respondents
that prefer a Lexus to a Pinto etc.). This gives the reader an idea
of how these preferences are reflected in the market.
Thurstone Methodology
• Thurstone values are relative. Preference is measured
to least preferred products.
– Gives no strategic/diagnostic information. E.g. Why is a Lexus
preferred over a Pinto?
• Please rank the products … in the order of your preference
– Does not require the full ranking
• Select the top 5 products that you prefer …. (record order)
– Please indicate your preference for each of the products
Thurstone Output
•
•
•
In this example, Lexus is the most
preferred car brand while Pinto is the
least preferred. Relative to these
brands, Ferrari (2nd most preferred)
falls just behind Lexus (.85 relative
ranking).
It can also be said that Lexus is
preferred almost twice as much as
Volvo.
Of all cars tested, Pintos are the
least preferred car.
1 Lexus
0.85
Ferrari
0.55
Volvo
0 Pinto
TURF/Shapley Value
TURF Background
• A TURF Analysis, (Total Unduplicated Reach & Frequency) is a
combinatorial technique, originally developed for media-mix
models to find the best combination of magazines to place ads
to achieve the maximum audience reach and frequency of
exposure.
• Typically not presented in a competitive context
• While answering product manager’s specific question, it does
little to provide strategic information about the product/flavor –
how much is this flavor worth?
TURF Background
• TURF Analysis is frequently employed for the building or
extending of product & flavor lines, finding the optimal
combination of products/flavors for a fixed number of products in
the line.
– What combination of 2 (3,4 or 5) flavors would maximize reach?
– Gives good indication of the optimal number of products/flavors
that should be included in the line.
•
Typically reach can be maximized by offering all of the potential flavors
or options. However, production costs, cannibalism, shelf space, and
capacity often prevent this strategy. Decisions need to be made to
determine which options will reach or appeal to the largest amount of
customers.
• Another application of Turf is in determining the best message
mix in marketing communications.
– The combination that messages that “reaches” the largest
proportion of the target group.
TURF Example
• Ice-cream flavor example
– Reach is defined by the Top2boxes ratings on a 5-pt purchase intent
scale.
• Vanilla -70%.
• Mango - 60%.
• Pistachio - 25%
• If the client company only has the resources to launch 2 flavors,
what recommendations would you make to your client?
– The two flavors with the highest top2boxes ratings?
• What the top2boxes rating does not tell you is the
duplicated/unduplicated reach of the flavors!!!!
TURF Example
Vanilla
Mango
Pistachio
20%
50%
10%
25%
• If the marketing objective is to maximize REACH
– Vanilla and Mango = 20% + 50% + 10% = 80%
– Mango and Pistachio = 50% +10% +25% = 85%
– Vanilla and Pistachio =20% + 50% + 25% = 95%
TURF Methodology
•
To obtain REACH, respondents are asked about their purchase intent
on all the potential products/flavors that make up the product line,
typically using a 5- or 10- point purchase intent scale.
– The TURF Analysis uses Top Box (or Top2Box) purchase intent as the basis
for the building of combinations of products/flavors.
•
Alternatively, a MaxDiff exercise can be used to solicit preference for
each products/flavors.
•
To obtain FREQUENCY,
– How often would you purchase each of the following?
• Once per week, twice per month, once per month ...
– How many of Flavor X, would you purchase in your next shopping trip?
– For your next 5 trips, how many of each … would you purchase
TURF Issues
• When using the frequency aspect of TURF it will determine which
factor solution will result in the most units sold, not the most
customers sold to.
• When considering frequency as a factor in the analysis, it is
important first to understand the nature of the product. We will have
to determine whether the products are complementary or
substitutable. This will affect how the analysis is conducted.
– Substitutable: Higher purchase of one variety leads to a lower
purchasing of another variety. For example: Ice cream flavors.
– Complementary: Varieties do not compete with each other. Higher
purchase of one variety does not lead to lower purchasing of another
variety. For example: Shampoo and Conditioner.
TUR Output
Driver Analysis/Shapley Value Regression
Driver Analysis
• Questions the client might ask
– What drives product purchase, satisfaction or loyalty?
– What drives perceptions of value?
– How are satisfaction, value and loyalty inter-related?
• Stated vs. Derived Importance:
– Stated Importance:
•
•
•
•
Scale based Questions: everything is important.
Direct questioning: why it doesn’t work?
Tradeoff Methods: MaxDiff
Pro Social Bias
– Derived Importance: uncovering inter-relationships
SV Regression
• The Shapley Value Principle:
– it represents the worth of each player over all possible
combinations of players.
• In Shapely Value Regression, we extend this to the
problem of comparative usefulness of possible
drivers.
– SV regression assigns a value for each potential
drivers calculated over all possible combinations of
predictors in regressions.
– OLS regression for all possible combinations of explanatory
variables
– Contribution measured by R-square
SV Regression
• The correlated nature of customer satisfaction data
does not present a problem with SV.
– It is inherently stable and can be used as a tracking
tool.
• If two variables are perfectly correlated, only one would load
in OLS regression.
• The SV regression analysis would compute the same score
for each of these variables.
• Generally conducted over the entire scale,
– Doesn’t distinguish between drivers of satisfaction
and drivers of dissatisfaction.
SV Regression Background
Example Output:
Overall R-Square
53%
SV
sdSV
Is best for symptoms caused by either seasonal allergens or irritants
7% like cold1%
air or smoke
Is steroid-free
2%
0%
Goes immediately to the site of your nasal symptoms
4%
0%
Works fast because it is a spray
1%
0%
Is best to treat not only allergies, but also symptoms caused by environmental
8%
1%irritants
Is best for preventing nasal congestion and nasal allergy symptoms
8%before they
1% begin
Works best for both indoor and outdoor allergies
9%
1%
The number one prescribed allergy medicine
3%
0%
Provides the most powerful relief of nasal congestion
7%
1%
Is different because it treats your symptoms by blocking leukotrienes
4%
0%
Higher Order Driver Analysis
Easy to Use
Durable
Product
Use Again
Functional
CSAT
Loyalty
Recommend
Responsive
SOW
Knowledge
Service
Helpful
Format
Analysis
Billing
Value
Delivery Method
WWPF
Price Compare
Price Evaluate
Brand +
Brand+
• An unique approach to Brand Research:
– Typically Usage & Attitudes Studies
– Brand Advancement Measures (BAM) are
normative and category specific.
• BAM scores are analyses derived from brand
association data. As such they are typically created
from your existing banks of market-specific brand
attributes. The use of additional generic attribute ratings
is therefore not required. Also, because BAM scores can
be created from set of attributes, is it possible to derive
BAM scores for previous waves of brand tracking
research. Derived importance is used to identify which
attributes are driving market share.
http://intranet/display/research/Analytics
[email protected]