Scribd
Upload
22 views3 pages

Wharton - Business Analytics - Week 3 - Summary Transcripts

The document discusses predictive analytics techniques for making predictions about future outcomes using historical data. It covers regression analysis for single period predictions and probability models like Buy Till You Die for multi-period predictions. Key concepts explained include regression terminology, applications, and using it to determine optimal prices.

Uploaded by

soravit137
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
22 views3 pages

Wharton - Business Analytics - Week 3 - Summary Transcripts

The document discusses predictive analytics techniques for making predictions about future outcomes using historical data. It covers regression analysis for single period predictions and probability models like Buy Till You Die for multi-period predictions. Key concepts explained include regression terminology, applications, and using it to determine optimal prices.

Uploaded by

soravit137
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Week 3 – Predictive Analytics: Making Predictions Using Data

Summary Transcripts

Introduction to Predictive Analytics


Data analysis helps you examine the past and predict the future. You can make predictions about a
fixed period in the future.

Applications of Predictive Analysis:


Predictions we may want to make from the data are:

• Which customer is going to buy next?


• How many purchases are they going to make?
• What is the customer churn?

Classification of Predictive Analytics


Predictive analytics is divided into two—forecasting one period ahead and forecasting more than one
period ahead.

For forecasting one period ahead, regression analysis is used. For more than one period ahead, Buy
Till You Die (BTYD) model and probability are used.

One Period Ahead: Regression Analysis


Regression analysis quantifies the relationship between two or more variables. As price increases,
sales decrease. Regression helps quantify this relationship.

Regression Analysis: Terminology


Regression explains a dependent variable (say, demand) as a function of independent variables (say,
price). For example, “What would be the demand at different prices?”. Linear additive model is used
for such predictions.

Regression Models
Regression for demand analysis is given by:
𝑆𝑎𝑙𝑒𝑠𝑡 = 𝑎 + 𝑏1 𝑃𝑟𝑖𝑐𝑒𝑡 + 𝑒𝑡
Where,
Salest = Dependent variable
Pricet= Independent variable
b1 = Price Sensitivity

Simple Regression
Simple regression is given by:
𝑌𝑡 = 𝑎 + 𝑏1 𝑋𝑙𝑡 + 𝑒

To understand relationship between advertising and sales, substitute:


Y = Sales (dependent variable)

Business Analytics: From Data to Insights Page 1 of 3


X = Advertising (independent variable)

Applications of Regression Models


Key performance indicators (KPIs) use people's past behavior to predict future behavior. Regression
models are helpful in this area. Regression models are limited at projecting way into the future
because they need a lot of input data.

Using regression models, data from the first part could be used to predict the second part. To make
predictions for period three, period two data can be used as the independent variable. Regression
models are helpful to understand if a customer is going to churn in the next period. But regression
models cannot be used to determine when a customer is going to churn.

Regression Analysis: Making Predictions One Period Ahead


Regression is an ideal tool for understanding the drivers of demand and for demand prediction. It can
be used for determining optimal prices. Here are the steps for going from data to decisions using
regression:
• Start with the overall data
• Quantify relationship between variables
• Make demand predictions
• Determine the optimal prices

Optimal Pricing
Optimal price is the that maximizes overall profit. Steps for determining the optimal pricing are:
• Make prediction at different prices
• Look for the revenue and profit
• Understand what price (optimal price) should be charged

Regression with Multiple Independent Variables


When multiple independent variables are involved, regression is expressed as:
𝑌𝑖 = 𝑎 + 𝑏1 𝑋1𝑖 + 𝑏2 𝑋2𝑖 + ⋯ + 𝑏𝑘 𝑋𝑘𝑖 + 𝑒𝑖

Where,
Yi = Dependent variable
Xi = Independent variable

Recency-Frequency-Monetary Value (RFM)


Period two prediction for customers' worth can be made using data from:
• Recency: Valuable activity
What did the customers buy recently?
• Frequency: Beneficial activity
How many purchases did they make?
• Monetary value: Average monetary gain
What is the monetary value of each purchase?

Business Analytics: From Data to Insights Page 2 of 3


Recency vs. Frequency
Recency and frequency are indicators of unobservable propensities of customer behavior, which will
help make long-term predictions. Keeping the monetary value (M) aside, Recency (R) trumps
Frequency (F).
More than One Period Ahead: Probabilities
Probabilities for different coin tosses are:
• When coins are very headsy, people are inclined to donate.
• When coins are very tailsy, people are disinclined to donate.
• For a U-shaped curve, some people are into it and some people who aren't.

More Than One Period Ahead: BTYD Model


Buy Till You Die (BTYD) model is a powerful tool to make long-range projections for both “buying” and
“dying”. We employ a “Buy Till You Die” model to predict future donation behaviors.

The model uses three inputs: Recency (R), frequency (F), and the number of people for each
combination of R/F. This requires a small amount of data and provides an easier structure to work
with (i.e., data are aggregated from individual-level to R/F groups). By assuming certain probability
distributions for donors’ propensities, we can construct a robust model that is easy to implement on
Excel.

The “BTYD” modeling approach has a long track record of success in a variety of different domains.
This model does a good job of predicting the actual number of purchases by frequency group and
helps us map out a good trajectory for the next five years and beyond.

Probability of Being “Alive”


Using a larger dataset from a different non-profit firm, we create a “heat map” that shows which
combinations of RF will likely yield the most valuable donors. When probabilities are zero, the donors
are considered ‘dead’.

Business Analytics: From Data to Insights Page 3 of 3

You might also like