PAMS 22Fall Smart Marketing With RRM 1 Introduction

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 54

Smart Marketing

with Random Rewards


SYSBS
2022-Fall
Zhou Ruikai

1
Zhou Ruikai

Research Interest
Marketing Modeling
Risk Perception
Entertainment, New Technology

Course Taught
Econometrics
Management Research Methods

2
Introduction

3
1. Random Rewards

2. Smart Decision Making

3. Course Overview

4
How do you buy an umbrella?

Color, pattern, foldable or not…explicit information is mostly enough

5
Search Goods

How to market an umbrella?

Wrap shows the pattern. To market search goods, marketers need to reduce the
cost for customers to search for related information

6
Search Goods

How to market an umbrella?

Good looking lady enhances the beauty of the pattern – even if the product
properties are given, marketers to change the way to present it.

7
How do you decide whether to watch a movie?

Genre, “IP”, stars

8
How do you decide whether to watch a movie?

Experience goods can only resolve the uncertainty after experience. Ratings,
especially from similar others, matter.

9
Experience Goods
How to market a movie?

10

Trailer, posters, casting

10
Experience Goods
How to market a movie?

11

To get an early ratings for people to refer to, free,early viewings are offered,
much like food where trial samples are offered

11
How do you buy healthcare supplements?

12

The essence of health supplements is that the uncertainty is often never fully
resolved, the effect to your body can hardly be actually felt or tested

12
Credence Goods

How to market healthcare supplements?

13

For such credence goods, marketers need extensive advertising to build brand
awareness and reputation

13
Credence Goods

How to market healthcare supplements?

14

Professional recommendation is also a way to build reputation and sell credence


good, remember how Will Smith sells X-ray machine

14
1. Uncertainty in Purchases

In many purchase occasions, consumers expect certainty from the products.


• Depending on how consumer resolves uncertainty of their purchase, products can be divided into
three types: search goods, experience goods, and credence goods.
• As uncertainty increases, the cost to build consumer trust increases
- for experience goods, giveaways, trials and a positive word-of-mouth are needed
- for credence goods, the hard work falls on building a brand reputation.

15

15
2. Selling at Random

On the other hand, there are companies directly exploiting uncertainty in their products, in the
name of “randomness”, “luck”, “probability”…

16

16
Probabilistic Selling

17

Pricebreakers is a Priceline service where random hotels are offered at a


discounted price.

17
Probabilistic Selling

18

Mystery boxes of clothes and books are also prevalent.

18
Gacha/Blind Box

19

Pop-Mart, a recent trendy blind-box seller, imitates the old way of Gacha selling.

19
Gambling and “Gambling”

20

Gambling, lottery, scratchers have been out there for long.

20
Randomness in Gaming

In some business, uncertainty is even a main source of revenues.


• The $2bn industry – electronic gaming – collects at least 20% revenues directly from “loot-boxes”
through microtransactions., and also indirectly since significant amount of rewards delivered by
watching ads (another main revenue source of e-gaming) are randomized.

21

21
Loot-boxes

22

22
3. Randomness in Marketing Campaign

A lot of marketing campaigns nowadays emphasize “gamification”.


• The essential of such “gamification” is randomness.
• Random prizes, roulettes, randomized pursuit of collection goal…

23

23
Lucky Elements in Customer Engagement

24

Mcdonald happy meal

24
Lucky Elements in Customer Engagement

25

Random elements are implemented in gaming process to increase the fun and
competitiveness.

25
4. Random Rewards

We will focus our discussion of random elements on the rewarding part.


• We use the term “random rewards”, borrowed from gaming industry.
• By being “rewards”, we mean that the products are always desirable if given for free.
• In other word, people are willing to pay.

Random losses, like decision to take risky jobs, can be reframed as “paying” to avoid risks.

26

26
What is Smart Marketing

A convenient breakdown of Smart Marketing includes:


S(pecific) – the goal of marketing is well-defined
M(easurable) – the achievement of the marketing goal is measurable
A(ttainable) – the path towards the goal is clear
R(ealistic) – the goal-pursuit process is well-grounded
T(imely) – the marketing scheme can quickly adapt to new trend

27

27
Decision Analysis

A standard procedure of decision analysis (as we will discuss later) solves the SMART
marketing problem in a systematic way:
• By defining objectives, our goal is specific
• By calculating objective function, our achievement is measurable
• By specifying action space, we ensure the attainability of our decision
• By modeling and estimating demand, we ground our decision in sound theories and facts (data)
• By adopting standard optimization procedure for decision making, we allow our decisions to be
automated and timely

28

28
1. Objective

We’ll start with a general framework of how we solve a decision making problem.
Our mission is to find the optimal business move 𝑎∗ , which maximizes an objective 𝝅(⋅).

𝑎∗ = 𝑎𝑟𝑔𝑚𝑎𝑥 ∈𝑨 𝜋 𝑎
(in other word, 𝜋 𝑎∗ = max 𝜋(𝑎))
∈𝑨

29

29
Different Objectives

30

Even for something as simple as buying drinks from vending machine, we can
easily have different objectives under different scenarios.

30
Different Objectives

Four kinds of gamers, which are divided through two dimensions:


competitiveness and social participation, can also be a metaphor of the four
kinds considerations of business objectives.

31
Business Objectives

32

While profit can easily become the objective in our decision analysis, different
considerations of profit need to be included – Is it stable across different possible
future? Does it show a steady growth or fluctuation? Will it affect the social
image? Is it worth to sacrifice profit can gain a larger market share?

32
2. Action Space

𝑎∗ = 𝑎𝑟𝑔𝑚𝑎𝑥 ∈𝑨 𝜋 𝑎
The maximizing mission poses us another
immediate task:
• Define our action space 𝑨, which includes all our
decision alternatives

33

The first step to specify our action space is to think of a framework – just like 4P’s
can be a good framework to analyze marketing mix decisions.

33
Random Reward Mechanism

Eligibility Random
Rewards
Condition Procedure

34

For random rewards, gaming industry break it down into three parts.

34
Eligibility

35

Typical eligibility choice for a random reward include money barrier


(microtransactions), skill barrier (achievement prizes) and habit barrier (login
prizes)

35
Random Procedure

36

Random procedure of random rewards include two parts: effect (visuals, sounds)
and probability. For the probability part, other than specifying the odds for single
random reward, bundle odds and pity-timers are also popular choices.

36
Reward

37

Avatars are the most simple form of random rewards. Based on this, various
advanced considerations can be taken into account:
Does the reward give the player competitive advantage? (equipment, armors)
Is there a collection? (set pieces)
Is the acquirement channel exclusive? (region-exclusive, time-limited events)
Is the reward exchangeable with others? (in-game currencies)

37
3. Optimization

𝑎∗ = 𝑎𝑟𝑔𝑚𝑎𝑥 ∈𝑨 𝜋 𝑎
The key of solving decision making problem is calculating 𝜋(𝑎):
• AKA. how different decisions can impact our objective?
• For example: -How different prices (𝑎) influence revenues(𝜋)?
-How different flavors (𝑎) influence enjoyment(𝜋)?

Once 𝜋(𝑎) is calculated, decision making is straightforward – we find the maximal point
of the function. This process is called “optimization”.
• When action space is complicated, we need computers to solve the problem.

38

38
3. Optimization

𝑎∗ = 𝑎𝑟𝑔𝑚𝑎𝑥 ∈𝑨 𝜋 𝑎
Sometimes, the optimal solution is not what we are looking for.
We may have constraints, which is a special form of action space limitation:
• budget, legal considerations, ethics
We may also deal with inconsistent objectives, which requires transforming some objectives
to constraints, or combining objectives together:
• market share vs revenue, individual vs group

We will talk about the computing procedures for optimization problem in Session 5.

39

39
4. Demand Estimation

𝑎∗ = 𝑎𝑟𝑔𝑚𝑎𝑥 ∈𝑨 𝜋 𝑎
In business scenario, however, calculating 𝜋 𝑎 may not be an easy task.
Unlike a vending machine problem, business decision involves prediction
• Prediction of consumer behavior

Marketing Consumer Business


Decision Behavior Objective

40

40
4. Demand Estimation

Prediction is the phase where our precious data kicks in.


• We would use data to understand our consumers and how they would react to our business decisions,
and therefore impact our business.

41

The commonly-used STP process, where companies segment their customers,


target their demands, and position the products, is a rough way to conducting
prediction. In STP, companies base their marketing decisions on a categorical
prediction outcome – segmentation.

41
4. Demand Estimation

𝑎∗ = 𝑎𝑟𝑔𝑚𝑎𝑥 ∈𝑨 𝜋 𝑦 𝑎, 𝑥 , 𝑎
𝑠. 𝑡. 𝑦 (⋅) = 𝑎𝑟𝑔𝑚𝑖𝑛 (⋅) ||𝑦 𝑎 , 𝑥 − 𝑦 ||
The above equations give us an idea how we predict consumer behavior, through a process
called demand estimation:
In the first equation, the mechanism of how our business decisions impact our profit (𝜋(𝑎)), is
further broken down into two parts:
• The profit is jointly dependent on our business decisions and the resulting purchase behavior
(“demand”, 𝑦 ) – this part is rooted in our business nature and should be defined ourselves
• The demand is a function of business decisions (𝑎) and consumer characteristics (𝑥) – the process of
finding such an equation is called “estimation”

42

42
Decision and Demand Function

43

Without an idea of demand function, business optimization cannot be conducted.

43
4. Demand Estimation

𝑎∗ = 𝑎𝑟𝑔𝑚𝑎𝑥 ∈𝑨 𝜋 𝑦 𝑎, 𝑥 , 𝑎
𝑠. 𝑡. 𝑦 (⋅) = 𝑎𝑟𝑔𝑚𝑖𝑛 (⋅) ||𝑦 𝑎 , 𝑥 − 𝑦 ||
The second equation illustrates how the estimation process in general works
• With historical data on consumer characteristics(𝑥 ) and business decisions(𝑎 ), we try to find a
function, 𝑦 (𝑎, 𝑥), that matches historical purchases (𝑦 ).
• The logic is that, if 𝑦 matches historical data, it should be a good prediction for future consumer
purchase behavior, given the information of consumers and our potential decisions.
• Generally, we cannot get a perfect match between prediction 𝑦 and data 𝑦 , so we must decide a
criteria to select among different imperfect 𝑦 (⋅). Such a criteria is called estimation method.
We will talk more about demand estimation in Session 4.

44

44
Using Past to Predict the Future

45

The estimation procedure is essentially using past to predict the future – just like
we use past behavior of players to classify them, and customize game content
based on the prediction of such classification.

45
Limitation of Doing Things Fully Data-based

46

Suppose we want to predict what shape of snowflake is more liked, simply using
data won’t give us answers, because “no two snowflakes are alike”.
To use the past to predict future, we need certain criteria to determine whether
two flakes are comparable or not – this process is called modeling.

46
5. Modeling

In most occasions, our data is not enough for us to distinguish the infinite possibilities of
demand function.
• “Big data” won’t help if we allow the function complications to grow exponentially with factors.
We can, however, use certain rationale to rule out some forms of the demand function.
• Such rationale can come from intuition, experience, theory, and approximation considerations etc.
Normally, such rationale is area-specific.
• The process of narrowing down our estimation task is called “modeling”.

In Session 2 and 3, we will talk about some prevalent theories regarding consumer behavior
facing random rewards.

47

47
6. A Summary of Smart Decision Making Procedures

Step Description Math Notation Course Session

Defining and Calculating What is the business goal 𝜋 y , 𝑎 =? 1


Objective Function and how it relates to
consumers
Specifying Action Space What is the key decision 𝑨 =? 1
and Constraints to make
Modeling What theories to refine 𝑦 ⋅ ∈ {… ? } 2 and 3
the functional form of
demand
Demand Estimation What method for 𝑦 ⋅ =? 4
estimation
Optimization What method for 𝑎𝑟𝑔𝑚𝑎𝑥 ∈𝑨 𝜋 𝑎 =? 5
optimization
48

48
1. Course Structure and Objectives Session 2:
Consumer Psychology and
Management Considerations
in Using One-time Random
Sub-Objective 1: Rewards
Understanding the Rationale
behind Randomness in
Marketing Session 3:
Consumer Psychology and
Management Considerations
in Using Repeated Random
Objective: Rewards
Understanding How to Design
Randomness in Marketing
Session 4:
Demand Estimation Method
for Extracting Decision Factors
Sub-Objective 2: from Business Data
Practicing Data-based
Methods to Optimize
Randomness in Marketing Session 5:
Optimization Method for
Smart Decision Making
49

49
2. Homework and Class Participation

After each class, a topic will be assigned, on which you will need to write a short essay by
searching information or using your own experience.
Students are encouraged to participate in each of the classes:
• Assigned students should turn on their camera and actively join discussions
• Selected students may be asked to briefly share their essay
• Other students needs to participate by asking questions or sharing comments.

50

50
3. Term Essay

Students will need to pick a real-life business case where randomness is used in marketing,
and conduct thematic analyses after each class.
• The theme and details will be provided at the end of each class.
• At the end of the course, students are required to combine five thematic analyses to form a
comprehensive essay regarding how to optimize the randomness usage in the selected case.

51

51
4. Software

The later half of the course (Session 4 and 5) involves computing tasks that require
software tools.
• This course will focus on explaining the intuition/logic of these tasks and their solutions.
• Brief instructions (on or off the class) will be given about using R as the tool to solve these tasks.
• R usage (or other computing tools) is encouraged, but not required, in completing homework and
final essay.

52

52
Short Essay

Provide a rough analysis of the ROI (return of investment) of both:


1. An insurance plan (pick as you wish, please attach product details in the file)
2. A lottery program (pick as you wish, please attach product details in the file)
And discuss your purchase intention

Thematic Analysis (1st Part of the Final Essay)

Find a random reward case and analyze its eligibility, random procedure and reward
• The case needs to have its odd disclosed
• Better come from your own life experience (games, marketing campaigns etc)
53

53
To be continued...

54

You might also like