Professional Documents
Culture Documents
Rahul Sahai
Rahul Sahai
About me: Rahul Sahai
• “Pricing and Modern Analytics” for the modern render consumer, you
can expect to pay whatever you want for the tablet.This is reflected in the bag
prices of the tablets too. They have an average $140 price tag, which is more
than $250 less than the $370 iPad mini only.
MC
MB
“
Pricing Strategies can be categorized in 3 Groups:
Main Strength Data readily available Data can be made Takes customer
available via Basic perspective into account
Market Research
Main Weaknesses Does not take Does not take Data is harder to collect
competition & customers into account and analyze- Needs
customers into account (WTP) support from Quantitative
(WTP etc) Models
Overall Weakest Approach Sub Optimal for setting Overall Best Approach
Evaluation prices but appropriate Direct link to customer
for commodities and centricity but the value
only if products/ service needs to be
cannot be differentiated communicated via a robust
communication strategy
Few modern pricing techniques
• Price your
product to the
closest available
alternate
method of
purchase for the
customer
• Import parity
pricing used in
Blue- Your locations Oil & Gas
Black- Customer Locations
downstream
Bubble- minimum time taken to travel from closest location
Real Options Pricing: Factoring uncertainities
Valuing a real option looks at the premium between inflows and outlays for a particular project.
Inputs to the value of a real option (time, discount rates, volatility, cash inflows and outflows) are
each affected by the terms of business, and external environmental factors that a project exists in.
Risk Coverage Pricing- Leverage Recursive
Partitioning (PARTY Algorithm)
Pricing schemes, discounts, loyalty programs and
promotions- Leveraging advanced analytics (MLP)
Load Dataset (7044 Rows of Data)- Telecom Data
set for evaluating customer churn
customerID gender SeniorCitizen
Partner Dependents
tenure PhoneService
MultipleLines
InternetService
OnlineSecurity
OnlineBackup
DeviceProtection
TechSupportStreamingTVStreamingMovies
Contract PaperlessBilling
PaymentMethod
MonthlyCharges
TotalCharges
Churn
7590-VHVEG Female 0 Yes No 1 No No phone service
DSL No Yes No No No No Month-to-month
Yes Electronic check
29.85 29.85 No
5575-GNVDE Male 0 No No 34 Yes No DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No
3668-QPYBK Male 0 No No 2 Yes No DSL Yes Yes No No No No Month-to-month
Yes Mailed check 53.85 108.15 Yes
7795-CFOCW Male 0 No No 45 No No phone service
DSL Yes No Yes Yes No No One year No Bank transfer (automatic)
42.3 1840.75 No
9237-HQITU Female 0 No No 2 Yes No Fiber opticNo No No No No No Month-to-month
Yes Electronic check70.7 151.65 Yes
9305-CDSKC Female 0 No No 8 Yes Yes Fiber opticNo No Yes No Yes Yes Month-to-month
Yes Electronic check
99.65 820.5 Yes
1452-KIOVK Male 0 No Yes 22 Yes Yes Fiber opticNo Yes No No Yes No Month-to-month
Yes Credit card (automatic)
89.1 1949.4 No
6713-OKOMC Female 0 No No 10 No No phone service
DSL Yes No No No No No Month-to-month
No Mailed check 29.75 301.9 No
7892-POOKP Female 0 Yes No 28 Yes Yes Fiber opticNo No Yes Yes Yes Yes Month-to-month
Yes Electronic check
104.8 3046.05 Yes
6388-TABGU Male 0 No Yes 62 Yes No DSL Yes Yes No No No No One year No Bank transfer (automatic)
56.15 3487.95 No
9763-GRSKD Male 0 Yes Yes 13 Yes No DSL Yes No No No No No Month-to-month
Yes Mailed check 49.95 587.45 No
7469-LKBCI Male 0 No No 16 Yes No No No internet
Noservice
internet
Noservice
internet
Noservice
internet
Noservice
internet
Noservice
internet
Two
service
year No Credit card (automatic)
18.95 326.8 No
8091-TTVAX Male 0 Yes No 58 Yes Yes Fiber opticNo No Yes No Yes Yes One year No Credit card (automatic)
100.35 5681.1 No
0280-XJGEX Male 0 No No 49 Yes Yes Fiber opticNo Yes Yes No Yes Yes Month-to-month
Yes Bank transfer (automatic)
103.7 5036.3 Yes
5129-JLPIS Male 0 No No 25 Yes No Fiber opticYes No Yes Yes Yes Yes Month-to-month
Yes Electronic check
105.5 2686.05 No
3655-SNQYZ Female 0 Yes Yes 69 Yes Yes Fiber opticYes Yes Yes Yes Yes Yes Two year No Credit card (automatic)
113.25 7895.15 No
8191-XWSZG Female 0 No No 52 Yes No No No internet
Noservice
internet
Noservice
internet
Noservice
internet
Noservice
internet
Noservice
internet
One
service
year No Mailed check 20.65 1022.95 No
9959-WOFKT Male 0 No Yes 71 Yes Yes Fiber opticYes No Yes No Yes Yes Two year No Bank transfer (automatic)
106.7 7382.25 No
4190-MFLUW Female 0 Yes Yes 10 Yes No DSL No No Yes Yes No No Month-to-month
No Credit card (automatic)
55.2 528.35 Yes
4183-MYFRB Female 0 No No 21 Yes No Fiber opticNo Yes Yes No No Yes Month-to-month
Yes Electronic check
90.05 1862.9 No
8779-QRDMV Male 1 No No 1 No No phone service
DSL No No Yes No No Yes Month-to-month
Yes Electronic check
39.65 39.65 Yes
Split into Train Data Set and Test/ Validation
Data Set
We have a new package, rsample, which is very useful for sampling methods. It has
the initial_split() function for splitting data sets into training and testing sets. The return
is a special rsplit object.
Feature Engineering “Discretize & Transform”
Numeric features like age, years worked, length of time in a position can generalize a group (or cohort).
We can split into six cohorts that divide up the user base by tenure in roughly one year (12 month) increments. This should help the ML
algorithm detect if a group is more/less susceptible to customer churn.
What we don’t like to see is when a lot of observations are bunched within a small part of the range. We can use a log transformation to even
out the data into more of a normal distribution. It’s not perfect, but it’s quick and easy to get our data spread out a bit more.
Model Customer Churn with Keras
• Deep Learning has been available in R for some
time, but the primary packages used in the wild
have not (this includes Keras, Tensor Flow,
Theano, etc, which are all Python libraries).
• We’re going to build a special class of ANN
called a Multi-Layer Perceptron (MLP). They are
both highly accurate and serve as a jumping-
off point for more complex algorithms.
• MLPs are quite versatile as they can be used for
regression, binary and multi classification (and
are typically quite good at classification
problems).
Predicting, Confusion Table, Accuracy
# A tibble: 1,406 x 3
truth estimate
class_prob <fctr> Truth
<fctr> <dbl> 1 yes no
0.328355074 2 yes yes no yes
0.633630514 3 no no
0.004589651 4 no no no 950 161
Prediction
0.007402068 5 no no
yes 99 196
0.049968336 6 no no
0.116824441 7 no yes
0.775479317 8 no no
0.492996633 9 no no
0.011550998 10 no no
0.004276015 # ... If you know which customer is going to churn for what reason,
with 1,396 more rows you will be able to offer schemes to retain them as well
Real Estate Pricing: Our dataset has 21613 observations
and 20 columns- The variable we want to predict is price
Decision trees cannot only be used as a powerful tool for predictive models but also for
exploratory data analysis. The learned tree relies on several variables in order to distinguish
between cheaper and pricier houses. The features we split along are grade, sqft_living, but also
some features related to the area (longitude and latitude).
Real Estate Pricing: The association between the
features can also lead to interesting insights
Real Estate Pricing: Predicting prices based on the
most important variables driving pricing (Zipcode)
• Availability • Warranty
2 3
1 4
• Price • Credit
Combinations:
Terms
• Price X 4 Options
• Availability X 4 Options
• Warranty X 5 Options
• Credit Terms X 2 Options
• Hence total number of Choices possible = 4X4X5X2= 160 Choices
Conjoint Questionnaire- VOC
Rank on a scale of 0 to 10 the Propositions sent in the template
Public
Simulation: Collection of Ratings
Preference Ratings for Each of the 160 Choices (Propositions) collected from Customer; The same is grouped and
averaged as per the sub segment you are collecting inputs for
Avg Preference
Choices Price Availability Warranty Credit Terms Person A Person B Person C Person D Rating
1 80% of New Phone Off the Shelf Zero Standard- 37 Days 0 0 0 0 0.0
2 80% of New Phone Off the Shelf Zero Extended 67 Days 0 0 0 0 0.0
3 80% of New Phone Off the Shelf 6 months Standard- 37 Days 1 0 0 0 0.3
4 80% of New Phone Off the Shelf 6 months Extended 67 Days 1 0 0 0 0.3
5 80% of New Phone Off the Shelf 1 Year Standard- 37 Days 2 0 0 0 0.5
6 80% of New Phone Off the Shelf 1 Year Extended 67 Days 2 0 0 0 0.5
7 80% of New Phone Off the Shelf 2 Years Standard- 37 Days 3 0 0 0 0.8
8 80% of New Phone Off the Shelf 2 Years Extended 67 Days 3 0 0 0 0.8
9 80% of New Phone Off the Shelf 3 Years Standard- 37 Days 4 0 6 0 2.5
10 80% of New Phone Off the Shelf 3 Years Extended 67 Days 4 0 8 0 3.0
Public
Simulation: Prepare the Data for Regression
Converting the Choices to Binary for Conducting Regression
Avg Pref
Choices 80% of NP 75% of NP 60% of NP 50% of NP Off the Shelf 2 Weeks 4 Weeks 6 Weeks 0W 6m W 1Y W 2Y W 3Y W Std 37 Ext 67 Rating
1 1 0 0 0 1 0 0 0 1 0 0 0 0 1 0 0.0
2 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0.0
3 1 0 0 0 1 0 0 0 0 1 0 0 0 1 0 0.3
4 1 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0.3
5 1 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0.5
6 1 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0.5
7 1 0 0 0 1 0 0 0 0 0 0 1 0 1 0 0.8
8 1 0 0 0 1 0 0 0 0 0 0 1 0 0 1 0.8
9 1 0 0 0 1 0 0 0 0 0 0 0 1 1 0 2.5
10 1 0 0 0 1 0 0 0 0 0 0 0 1 0 1 3.0
Public
Simulation: Address Multicollinearity Errors
Removing 1 Level of Each Attribute to Address Multicollinearity Errors
Choices 75% of NP 60% of NP 50% of NP Off the Shelf 4 Weeks 6 Weeks 6m W 1Y W 2Y W 3Y W Ext 67 Avg Ranking
1 0 0 0 1 0 0 0 0 0 0 0 0.0
2 0 0 0 1 0 0 0 0 0 0 1 0.0
3 0 0 0 1 0 0 1 0 0 0 0 0.3
4 0 0 0 1 0 0 1 0 0 0 1 0.3
5 0 0 0 1 0 0 0 1 0 0 0 0.5
6 0 0 0 1 0 0 0 1 0 0 1 0.5
7 0 0 0 1 0 0 0 0 1 0 0 0.8
8 0 0 0 1 0 0 0 0 1 0 1 0.8
9 0 0 0 1 0 0 0 0 0 1 0 2.5
10 0 0 0 1 0 0 0 0 0 1 1 3.0
In statistics, multicollinearity (also collinearity) is a phenomenon in which one predictor variable in a multiple regression model can
be linearly predicted from the others with a substantial degree of accuracy. In this situation the coefficient estimates of the multiple
regression may change erratically in response to small changes in the model or the data. Multicollinearity does not reduce the
predictive power or reliability of the model as a whole, at least within the sample data set; it only affects calculations regarding
individual predictors. That is, a multivariate regression model with collinear predictors can indicate how well the entire bundle of
predictors predicts the outcome variable, but it may not give valid results about any individual predictor, or about which predictors
are redundant with respect to others.
Public
Simulation: Run the Regression
Running Regression to Evaluate Coefficients of Explanatory Variables (Part Utility measured in Utils)
SUMMARY OUTPUT
Regression Statistics
Multiple R 0.931986636
R Square 0.868599089
Adjusted R Square 0.858832806
Standard Error 1.108546155
Observations 160
ANOVA
df SS MS F Significance F
Regression 11 1202.2375 109.2943182 88.93854602 1.8563E-59
Residual 148 181.8734375 1.228874578
Total 159 1384.110938
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%
Intercept -1.15 0.303587868 -3.788030165 0.000220281 -1.749926814 -0.550073186 -1.749926814 -0.550073186
75% of NP 0.2125 0.247878456 0.857274987 0.392678974 -0.277338192 0.702338192 -0.277338192 0.702338192
60% of NP 3.35625 0.247878456 13.539902 1.1027E-27 2.866411808 3.846088192 2.866411808 3.846088192
50% of NP 4.69375 0.247878456 18.93569162 2.22115E-41 4.203911808 5.183588192 4.203911808 5.183588192
Off the Shelf 0.4125 0.247878456 1.664122033 0.098203911 -0.077338192 0.902338192 -0.077338192 0.902338192
4 Weeks -0.55625 0.247878456 -2.244043348 0.026313493 -1.046088192 -0.066411808 -1.046088192 -0.066411808
6 Weeks -1.11875 0.247878456 -4.513300666 1.29283E-05 -1.608588192 -0.628911808 -1.608588192 -0.628911808
6m W 0.6484375 0.277136539 2.339776281 0.020631502 0.100781752 1.196093248 0.100781752 1.196093248
1Y W 1.5390625 0.277136539 5.553444908 1.26373E-07 0.991406752 2.086718248 0.991406752 2.086718248
2Y W 3.1328125 0.277136539 11.30422035 9.48248E-22 2.585156752 3.680468248 2.585156752 3.680468248
3Y W 4.8671875 0.277136539 17.56241715 5.06854E-38 4.319531752 5.414843248 4.319531752 5.414843248
Ext 67 0.14375 0.175276537 0.820132588 0.413459421 -0.202617907 0.490117907 -0.202617907 0.490117907
Public
Simulation: Calculating Part Utilities
What are the Part Utilities for Each of the Levels?
Zero 0.0000
6 months 0.6484
Warranty 1 Year 1.5391
2 Years 3.1328
3 Years 4.8672
Public
Simulation: Calculating Value of Utils
What are the Part Utilities for Each of iPhone Models (11, 11 Max and 11 Pro Max) ?
New Phone Price 50% of New Phone Customer Utility (Utils) 80% of New Phone Customer Utility (Utils) Value of Util in 000
Public
Simulation: Willingness to Pay Calculation
How much will the Customer Pay for Each of the Levels?
Attribute Levels Customer Utility (Utils) Value- 11 Value- 11 Max Value- 11 Pro Max
Public
Simulation: Putting it all Together
Which are the Choices where Customer Value gets maximized?
Choices 80% of NP 75% of NP 60% of NP 50% of NP Off the Shelf 2 Weeks 4 Weeks 6 Weeks 0W 6m W 1Y W 2Y W 3Y W Std 37 Ext 67 Utility WTP K19 WTP K38 WTP K50
Highest Utility & WTP
Public
Q&A