Transcript Pricing and Yield Management

Welcome
B2C eCommerce
Trends in Pricing
Jonathan Wareham
[email protected]
Price Levels
Assumption that electronic markets have less friction
than comparable markets.
• Search costs lower
• Competition increases
• Average prices should fall, converging on market
level
Study of prices of books and CDs and software sold on
internet:
Higher prices & greater variance in
electronic channel !!!!!
u
u
u
u
Possible Causes
1. Superior disc. pricing techniques: lower
registration and menu costs
2. Heterogeneity: wine in store or restaurant
• Versioning
3. Temporal preference: consumer behavior
and types
4. Imperfect information: bait and switch
5. Neural real estate: 5% sites/75% traffic
6. Market immaturity: eMarkets too young
Fixed Prices
P
$1.00
1 Coke
Q
Fixed Prices
P
P Consumers Surplus
Dead Weight Loss
Q
MC
Q
Get a little more revenue
P
P1
P2
P3
Q1
Q2
Q3
Q
2nd Degree Price Discrimination
 “product line pricing”, “market
segmentation”, “versioning”
 Gold Club, Platinum Club, Titanium Club,
Synthetic Polymer Club
 First Class, Business Class, World
Traveler Class
 Professional Version, Home Office
3rd Degree Price Discrimination
 The practice of charging
different groups of consumers
different prices for the same
product
 Examples include student
discounts, senior citizen’s
discounts, regional &
international pricing, coupons
Maximize the Revenue !
Perfect (1st degree) Price Disc.
P
Q
Perfect Price Discrimination
Price $
Profits:
.5(4-0)(10 - 2)
= $16
10
8
6
4
Total Cost
2
MC
D
1
2
3
4
5
Quantity
Prefect Price Discrimination
 Practice of charging each
consumer the maximum amount
he or she will pay for each
incremental unit
 Permits a firm to extract all
surplus from consumers
 Difficult: airlines, professionals
and car dealers come closest
Caveats:
 In practice, transactions costs and
information constraints make this is difficult
to implement perfectly (but car dealers and
some professionals come close).
 Price discrimination won’t work if you
cannot control three things:
 Preference profiles
 Personalized billing; (anonymous
transactions lesson seller’s discriminatory
power over consumers)
 Consumer arbitrage
What is different about this site?
Conclusions
1. Internet double edged sword:
•
Consumers enjoy lower search costs, but…
•
eMarketers have superior tools to register
your consumption patterns and price
sensitivity
2. The end of fixed pricing???
•
Fixed pricing as an institution only 100 years
old!!
•
Developed in response to large scale
economies/production models….with
standard products !!!!
Horizontal Differentiation
 The game of location (proximity to
customer’s tastes)
1/2
Bob
Bob
Alice
Alice
Vertical Differentiation
Price
High
Low
Quality
How???
1. Versions
2. Timing and delays
3. Ease of use
4. Pathways into site
5. Segregation of markets and users
6. Analysis of click stream and previous purchasing
history
Making Self-Selection Work
 May need to cut price of high end
 May need to cut quality at low end
 Value-subtracted versions
 May cost more to produce the lowquality version.
 In design, make sure you can turn
features off!
How Many Versions?
 One is too few
 Ten is (probably) too many
 Two things to do
 Analyze market
 Analyze product
Analyze Your Market
 Does it naturally subdivide into
different categories? AND
 Are their behaviors sufficiently
different?
 Example: Airlines
 Tourists v. Business travelers
 “This created visible differentiation in
customer service. It was essential for
our customers to see the perks that
the others were getting.”
Analyze Your Product
 Dimensions to version
 High and low end for each dimension
 Design for high end, reduce quality
for low end
 Low end advertises for high end in
service industries – Cheap rates
 High end – Flagship products advertises for low end in many
products.
Goldilocks Pricing
 Mass market software (word,
spreadsheets)


Network effects
User confusion
 Default choice: 3 versions
 Extremeness aversion
 Small/large v. small/large/jumbo
Extremes Aversion
 Bargain basement at $109,
midrange at $179
 Midrange chosen 45% of time
 High-end at $199 added
 Mid-range chosen 60% of time
 Wines
 Second-lowest price
 “Framing effects”-example
Cross-Subsidies
 Prices charged for one product are
subsidized by the sale of another product
 May be profitable when there are significant
demand complementarities effects
 Examples
 Browser and server software
 Drinks and meals at restaurants
 Long distance and local access
 Auto spare parts
 Razor & Blades
 Burger, fries, drinks
 Auto financing
Lessons
 Version your product
 Delay, interface, resolution, speed,
etc.
 Add value to online information
 Use natural segments
 Otherwise use 3
 Control the browser, access,
comparisons, etc.
 Bundling & cross subsidies may
reduce dispersion
Down & Dirty
 First degree (perfect) price
discrimination
 “market of one”
 Second degree price discrimination
 “product line pricing”, “market
segmentation”, “versioning”
 Third degree price discrimination
 “different prices to different groups”
 Other definitions in literature…
...Decisions Are Not Always
“Rational”
Tickets; $7.95
Tickets; $6.95
$1.00 Discount
for Children &
Seniors
$1.00 Extra
for Middle Aged
People
Price Perception Issues are
Complex...
More Acceptable
Pricing
Product-Based
Open
Discretionary
Discounts and
Promotions
Rewards
Less Acceptable
Pricing
Customer-Based
Hidden
Imposed
Surcharges
Penalties
RM coming of age
1978:

Airline deregulation in the U.S.
1985:
1992:

People Express vs. American Airlines


Edelman Award: RM for AA $1.4 billion in 3 years
virtually every airline has implemented RM
National Car Rental (vs. GM)

Edelman Award: RM for SNCF

AA: $1 billion incremental revenues from RM
Marriott Int’l RM: 4.7% increase in room revenue

1997:

1999:
2000-01:
2003:

Deregulation Europe: telecom, media, energy …
e-distribution supports dynamic pricing & profiling

Dell, Amazon & Coca Cola experiment dynamic pricing

RM spans wide range of industries …

RM Evolution
HealthCare/
Hospitals
Telco/ISP
Insurance/
banking
Sports
Parks
Cruise lines
Entertainment
Car rental
Airlines
1980
Rail
Transp.
Hotels
1985
1990
Freight,
Cargo
Energy
Tour
Operators
Media
1995
Manufact.
2000
Retailers
YM: Where and When?
1) Perishable: impossible to store excess
resources
2) Choose now: future demand is uncertain
(how many rooms to sell at low price)
3) Customer segmentation with different
demand curves
4) Same unit of capacity can be used to
deliver different services
5) Producers are profit driven and price
changes are accepted socially
Major Types







Revenue Management (EMSR)
Peak-Load Pricing
Markdown Management
Customized Pricing
Promotions Pricing
Dynamic List Pricing
Auctions
Revenue Management
 Set of techniques use to manage
 Constrained, perishable inventory (time)
 When customer willingness to pay increases
towards departure
 Applications:
 Airlines, Hotels, Car Rentals, News Vendors
 Main techniques: Open and close certain
rate categories (rate fences) based on
historical probabilities and forecasts of
future demand
The RM Challenge
Arrivals of
high paying
customers…
Closer to
departure!
Arrivals of
low paying
customers
…Earlier!
Peak-Load Pricing
 Tactic of varying the price of constrained
and perishable capacity to reflect
imbalances between supply and demand
 Based on changing prices only, not
availability like RM. No perishable inventory
 Simple= when demand increases, raise
prices
 Industries= utilities (electricity, telephones)
theme parks, toll bridges, theatres
(afternoon showings)
Markdown Management
 Techniques used to clear excess,
perishable inventory over time
 Customer demand decreases over
time (opposed to RM)
 Used in retailing of fashion apparel
and consumer electronics where there
is a high obsolescence
Customized Pricing
 Occurs when the seller has the
opportunity to offer a unique price to
a buyer
 Equivalent to first degree price
discrimination
 Used by car dealers, professional
services, industrial sales, made to
order manufacturing, person to
person negotiation of nonstandardized products
Promotions Pricing
 Similar to markdown management
 Portfolio of tools to address different
customer segments.
 Example Automobile Sales
 Low income like cheap financing and low
down payment
 High income like cash back, additional
add-ons, services warranties/agreements
Dynamic List Pricing
 Dynamically move prices up and
down according to perceived changes
in demand.
 Products not constrained, can reorder
more.
 Not traditionally used because of high
menu costs
 Now used in Internet and traditional
retailing due to new technologies.
Auctions
 Variable pricing mechanisms
 Often used for instances when prices
are not easily determined
 English
 First price sealed bid
 Vickrey
 Dutch
The RM Challenge
Arrivals of
high paying
customers…
Closer to
departure!
Arrivals of
low paying
customers
…Earlier!
Expected Marginal Seat Revenue
 “ESMR” Kernel in many YM systems
 Peter Belobabba, MIT
 Belobaba, P. “Application of a
Probabilistic Decision Model to Airline
Seat Inventory Control,” Operations
Research, vol 37(2) 1989.
EMSR a simple example









Hotel; 210 rooms
Business Customers = 159$ night
Leisure Customers = 105$ night
We are now in February, the hotel has 210
rooms available for March 29.
Leisure Customers book earlier
Business Customers book later
How many rooms to sell at low price now?
How many to save to try and sell a high
price later?
What if we don not sell them all at 159$ then we lost 105$ per room!!!!
Terms
 Booking limit: Maximum number of
rooms to be sold at low price
 Protection level: Number of rooms to
be saved for the business customers
who arrive later
 Booking limit = 210 – protection level
Depiction: What should Q be?
210 rooms
Q+1 rooms protected
(protection level)
Q
210- (Q-1) rooms sold at discount
(booking limit)
Decision Tree
Revenue
Yes – sell (Q+1) room now
Lower protection
level from Q+1 to
Q?
No – protect
(Q+1) rooms
Sold at full price
later
Not sold by March
29
105 $
159 $
0$
Historical Demand
Demand for
# days
rooms at full
with
price
demand
0-70
12
71
3
72
3
73
2
74
0
75
4
76
4
77
5
78
2
79
7
80
4
81
10
82
13
83
12
84
4
85
9
86
10
above 86
19
TOTAL
123
Probability
9,8%
2,4%
2,4%
1,6%
0,0%
3,3%
3,3%
4,1%
1,6%
5,7%
3,3%
8,1%
10,6%
9,8%
3,3%
7,3%
8,1%
15,4%
100,0%
Cumulative
probability
9,8%
12,2%
14,6%
16,3%
16,3%
19,5%
22,8%
26,8%
28,5%
34,1%
37,4%
45,5%
56,1%
65,9%
69,1%
76,4%
84,6%
100,0%
100,0%
Decision Tree
Revenue
Yes – sell (Q+1) room now
Lower protection
level from Q+1 to
Q?
No – protect
(Q+1) rooms
1-F(Q)
F(Q)
105 $
159 $
0$
Calculation
(1-F(Q))($159) + F(Q)($0) =
(1-F(Q))*($159)
Therefore we should lower booking limit
to Q as long as
(1-F(Q))*($159)<=$105
Or
F(Q)>=($159-$105)/$159 = 0.339
Rational
 Find smallest Q with a cumulative
value greater than or equal to 0.339.
 Optimal protection is Q=79 with a
cumulative value of .341
 Booking limit: 210 -79 =131
 Save 79 rooms for business travlers
 Sell 131 rooms for tourist travlers
Demand for
# days
rooms at full
with
price
demand
0-70
12
71
3
72
3
73
2
74
0
75
4
76
4
77
5
78
2
79
7
80
4
81
10
82
13
83
12
84
4
85
9
86
10
above 86
19
TOTAL
123
Probability
9,8%
2,4%
2,4%
1,6%
0,0%
3,3%
3,3%
4,1%
1,6%
5,7%
3,3%
8,1%
10,6%
9,8%
3,3%
7,3%
8,1%
15,4%
100,0%
Cumulative
probability
9,8%
12,2%
14,6%
16,3%
16,3%
19,5%
22,8%
26,8%
28,5%
34,1%
37,4%
45,5%
56,1%
65,9%
69,1%
76,4%
84,6%
100,0%
100,0%
Overbooking
 Lost revenue due to seats
 Penalties and financial compensation
to bumped customers
 X = # of no-shows with distribution
of F(x)
 Y = number of seats overbooked
 Airplane has S# of seats
 We will sell S+Y tickets
Overbooking Calculation
 C = penalties and bad will caused by
bumping customers
 B represents the opportunity cost of
flying with an empty seat (or the
price of the ticket)
 The optimal number of overbooked
seats
 F(Y) >= B/B+C
Overbooking Example
 # of customers who book but fail to
show up are normally distributed
mean=20 std.=10
 It costs $300 to bump a customer
 Hotel looses $105 if it does not sell
room at $105
 Overbooking b/b+c
$105/($105+$300) = .2592
Overbooking Example
From normal distribution we get
Φ(-.65)= 0.2578 & Φ(-.64) = 0.2611
Take z*=-0.645
Overbook Y=20-(0.645*10)=13.5
Excel =Norminv(.2592, 20, 10) gives
13.5
 Round up to 14 means 210+14=224





Overbooking metrics
 Service level based:
 P(denial) =0.05
 E[#denials]=2
 Etc.
 Cost based: assign a cost to each and
optimize
Overbooking cost (airlines):




Direct compensation cost
Provision cost of hotel/meal
Reaccom cost (another flight/airline)
Ill-will cost (~ “lifetime customer value”)
Industries
Overbooking
 Airlines
 Hotels
 Car rentals
 Education
 Manufacturing
 Media
No Overbooking
 Restos
 Movies, shows
 Events
 Resort hotels
 Cruise lines
CRM
DPRM
 “Attract & retain
customers”
 maximize profit from
each customer
 Segment by customer
LTV
 Price/availability= fct. of
forecasted customer
LTV to the organization
 Ignores capacity issues
and opportunity costs
(displacement)
 Wealth of data
 “generate revenue”
 maximize profit from
available assets
 Segment by customer
WTP
 Price/availability = fct.
of forecasted demand &
available supply
 Ignores customer value
issues and long term
revenues
 Quantifiable value
Maximize long-term profits
CRM & RM
Variables to track







Actual win or loss
Number of days played
Credit history
Length of stay at hotel
Individual spending preferences
Demographics
Psychographic profiles
Theoretical Revenue
 Theoretical =
(total amount wagered) X
(house advantage)
100$ hand x 10 hours x 100 Hands/hour
x .01 (house adv. 49/51) = $1,000
Can you track every single person???
 Not always
 Difficult in table games
 Theoretical =
(total amount wagered) X
(house advantage)
Where..
Total amount wagered = estimated
average bet x estimated time played
Future estimates…
 ADT = Average Daily Theoretical
Revenue
 Assumes that this level is constant
 Multiply by estimated # of days of
future trip to gain value
 Combined with CRM data on
consumption of food and beverage,
entertainment, pshychographics, etc
Rooms, a scarce resource
 Heads in beds: make money on
gaming
 Comp. Rooms: traditionally a fixed
number of rooms given to big
gamblers
 Used averages to cost out, did not
dynamically look at “opportunity cost”
ReInvestment amount
% of the ADT
ADT $1,000
Reinvestment amount = 30%
= $300
Total value of the room, F&B,
Entertainment, etc. must be less than the
 Room 200, F&B 100, Ent. 80..more than
ADT x reinvest.
 Ergo…try and sell room..
 Sophisticated applications use dynamic
pricing to asses opportunity costs..





Requirements
 RM – Yield management like the
airlines..
 Player tracking systems..Use cards
like Harras, to register all activity and
psychographic profiles
 POS resturants, theaters, spas, retail
stores, entertainment, etc…
 CRM integrates all of the above!!
 Statistical analysis and optimization
applications.