Product Cannibalization A Prototypical Marketing Science Problem - - PowerPoint PPT Presentation

▶
product cannibalization
SMART_READER_LITE
LIVE PREVIEW

Product Cannibalization A Prototypical Marketing Science Problem - - PowerPoint PPT Presentation

Apr 03, 2023 305 likes •836 views

The webinar will start at: 13:00:00 The current time is: 13:00:49 Central Daylight Time UTC-5 Product Cannibalization A Prototypical Marketing Science Problem Introduction Your Hosts Today Stefan Conrady stefan.conrady@bayesia.us

slide-1
SLIDE 1

A Prototypical Marketing Science Problem

Product Cannibalization

13:00:49

The current time is:

13:00:00

The webinar will start at: Central Daylight Time UTC-5

slide-2
SLIDE 2

2 BayesiaLab.com

Your Hosts Today

  • Stefan Conrady

stefan.conrady@bayesia.us

  • Stacey Blodgett

stacey.blodgett@bayesia.us

Introduction

slide-3
SLIDE 3

3

Motivation & Background

  • Definitions
  • Introductory Example

Representation

  • Conceptual Framework: Bayesian Networks
  • Probabilistic Reasoning

Learning, Estimation, and Inference

  • Causal Reasoning?
  • Unsupervised Learning
  • Disjunctive Cause Criterion
  • Assign Utilities
  • Evaluate Policies

Today’s Program

stefan.conrady@bayesia.us All Fictional Numbers

slide-4
SLIDE 4

4

Webinar Slides & Recording Available

stefan.conrady@bayesia.us

slide-5
SLIDE 5

5 BayesiaLab.com

Definitions

  • Typically, a new product adversely affects the sales of existing products:
  • If it affects your competitor’s products, it’s
  • If it affects your own products, it’s

Motivation & Background

Conquest

slide-6
SLIDE 6

6 www.BayesiaLab.com 6 2 3

  • M

a r

  • 1

8

slide-7
SLIDE 7

7 BayesiaLab.com

Introductory Example: 2000 BMW X5

  • First SUV in the BMW product portfolio.

Motivation & Background

X5

slide-8
SLIDE 8

8 BayesiaLab.com

Introductory Example: New BMW X3 vs. Existing BMW X5

  • New, smaller X3 launched in 2004

Motivation & Background

X3

Product B

X5

Product A

Cannibalization?

slide-9
SLIDE 9

Bayesian Network Representation

slide-10
SLIDE 10

10 BayesiaLab.com

Conceptual Network

Bayesian Network Representation

+ –

Product B causes lower sales of Product A

P(SalesB) P(SalesA|SalesB)

“Cannibalization”

slide-11
SLIDE 11

11 BayesiaLab.com

Inference

  • Computing the cannibalization effect C of Product B on Product A:
  • C(B

A) = -0.3 (unit effect)

Bayesian Network Representation

Existing Product A Mean: 1.200 Dev: 0.748 Value: 1.200 20.00% 40.00% 1 40.00% 2 New Product B Mean: 0.000 Dev: 0.000 Value: 0.000 100.00% 0.00% 1 0.00% 2 Existing Product A Mean: 0.900 Dev: 0.831 Value: 0.900 (-0.300) 40.00% 30.00% 1 30.00% 2 New Product B Mean: 1.000 Dev: 0.000 Value: 1.000 (+1.000) 0.00% 100.00% 1 0.00% 2

Obvious, as we encoded that as our domain knowledge into the network.

slide-12
SLIDE 12

12 BayesiaLab.com

Can’t we do this in Excel?

Bayesian Network Representation

slide-13
SLIDE 13

13 BayesiaLab.com

Example: BMW Portfolio of “Utility-Type” Vehicles in 2018

Motivation & Background

All products are cannibalizing each other!

slide-14
SLIDE 14

14 BayesiaLab.com

A Fully Connected Network?

Bayesian Network Representation

?

Can we specify it? No. Can we machine-learn it? Perhaps.

slide-15
SLIDE 15

Learning & Estimating Cannibalization

slide-16
SLIDE 16

16 BayesiaLab.com

Couldn’t we just ask auto buyers?

Learning & Estimating Cannibalization

slide-17
SLIDE 17

17 BayesiaLab.com

Understanding Cannibalization by Other Means?

  • Trade-Ins
  • New and old product not comparable
  • Auto Buyer Surveys (2nd Choice)
  • Respondents tend to exaggerate their

counterfactual choice (“I would have bought the convertible, but we need the third row.”)

  • Choice Experiments
  • Hypothetical choices are noncommittal
  • Expensive to conduct

Learning & Estimating Cannibalization

slide-18
SLIDE 18

18 stefan.conrady@bayesia.us

slide-19
SLIDE 19

19

Optimization Attribution Simulation Explanation Prediction Description

Model Purpose Model Source

Association/Correlation Causation Theory Data

Theory

Map of Analytic Modeling & Reasoning

BayesiaLab.com

slide-20
SLIDE 20

20

Optimization Attribution Simulation Explanation Prediction Description

Model Purpose Model Source

Association/Correlation Causation Theory Data

Map of Analytic Modeling & Reasoning

BayesiaLab.com

slide-21
SLIDE 21

A Fictional Case Study

Learning & Estimating Cannibalization

slide-22
SLIDE 22

22 BayesiaLab.com

Case Study Question:

  • What is the cannibalization effect of B on A, C, and D?

Learning & Estimating Cannibalization

A B C D

slide-23
SLIDE 23

23 BayesiaLab.com

Daily Sales Data

Learning & Estimating Cannibalization

Objective: To machine-learn a Bayesian network model from the sales data.

slide-24
SLIDE 24

24 BayesiaLab.com A desktop software for:

  • encoding
  • learning
  • editing
  • performing inference
  • analyzing
  • simulating
  • ptimizing

with Bayesian networks.

slide-25
SLIDE 25

25 BayesiaLab.com

Data Import Wizard

slide-26
SLIDE 26

26 BayesiaLab.com

Variable Type Definition

slide-27
SLIDE 27

27 BayesiaLab.com

Discretization

slide-28
SLIDE 28

28 BayesiaLab.com

Unconnected Network

slide-29
SLIDE 29

29 BayesiaLab.com

Unsupervised Learning Using the EQ Algorithm

slide-30
SLIDE 30

30 BayesiaLab.com

Final Network

How can we use this network to calculate the causal effect of B on A, C, and D? Counterintuitive arc directions!

slide-31
SLIDE 31

31 BayesiaLab.com

Disjunctive Cause Criterion

slide-32
SLIDE 32

32 BayesiaLab.com

VanderWeele and Shpitser (2011)

  • “We propose that control be made for any [pre-treatment]

covariate that is either a cause of treatment or of the outcome

  • r both.”

Disjunctive Cause Criterion

Implementation in BayesiaLab: Likelihood Matching on Confounders in Direct Effects Analysis  Causal Effect, i.e., the Cannibalization Rate

IMPORTANT ASSUMPTION: NO UNOBSERVED CONFOUNDERS

Cannibalizing Product Cannibalized Product Confounder

slide-33
SLIDE 33

33

Optimization Attribution Simulation Explanation Prediction Description

Model Purpose Model Source

Association/Correlation Causation Theory Data

Confounders

Map of Analytic Modeling & Reasoning

BayesiaLab.com

slide-34
SLIDE 34

34 BayesiaLab.com

Final Network

We need to define confounders and non-confounders. By default, all nodes are confounders.

slide-35
SLIDE 35

35 BayesiaLab.com

Computing the Direct Effect of B on A

slide-36
SLIDE 36

36 BayesiaLab.com

Direct Effect of B on A

slide-37
SLIDE 37

37 BayesiaLab.com

Direct Effect of B on C

slide-38
SLIDE 38

38 BayesiaLab.com

Direct Effect of B on D

slide-39
SLIDE 39

39 BayesiaLab.com

Adding a Decision Node

slide-40
SLIDE 40

40 BayesiaLab.com

Adding Utility Nodes

slide-41
SLIDE 41

41 BayesiaLab.com

Comparing Policies “B” vs. “No B”

Policy “B”: Utilities=90.285

slide-42
SLIDE 42

42 BayesiaLab.com

Comparing Policies “B” vs. “No B”

Policy “No B”: Utilities=98.321

slide-43
SLIDE 43

43

VR

In Conclusion…

slide-44
SLIDE 44

44 stefan.conrady@bayesia.us

Upcoming Webinars:

  • March 30

Good Friday — No Webinar

  • April 6

t.b.d.

  • April 13

t.b.d.

Register here: bayesia.com/events

Webinar Series: Friday at 1 p.m. (Central)

slide-45
SLIDE 45

45 BayesiaLab.com

slide-46
SLIDE 46

46 BayesiaLab.com

User Forum: bayesia.com/community

slide-47
SLIDE 47

47

Try BayesiaLab Today!

  • Download Demo Version:

www.bayesialab.com/trial-download

  • Apply for Unrestricted Evaluation Version:

www.bayesialab.com/evaluation

BayesiaLab Trial

BayesiaLab.com

slide-48
SLIDE 48

48

  • April 11–13

Sydney, Australia

  • May 16–18

Seattle, WA

  • June 26–28

Boston, MA

  • July 23–25

San Francisco, CA

  • August 29–31

London, UK

  • September 26–28

New Delhi, India

  • October 29–31

Chicago, IL

  • December 4–6

New York, NY

BayesiaLab Courses Around the World in 2018

Learn More & Register: bayesia.com/events stefan.conrady@bayesia.us

slide-49
SLIDE 49

49 BayesiaLab.com

San Francisco

Introductory BayesiaLab Course in San Francisco, California July 23–25, 2018

slide-50
SLIDE 50

50 BayesiaLab.com

Chicago

6th Annual BayesiaLab Conference in Chicago November 1–2, 2018

slide-51
SLIDE 51

51

Thank You!

BayesiaLab.com

stefan.conrady@bayesia.us linkedin.com/in/stefanconrady facebook.com/bayesia BayesianNetwork