SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
S HA541
Transcripts
Transcript: Course Introduction
Welcome
to
Price
and
Inventory
Control.
I
am
Chris
Anderson,
I'm
the
author
of
this
course
and
a
professor
at
Cornell
University
School
of
Hotel
Administration.
My
teaching
and
research
focus
is
largely
on
revenue
management
and
pricing
with
an
app,
application
specifically
in
service
industries.
This
course
focuses
on
one
of
the
core
concepts
of
revenue
management.
That
being
marginal
analysis.
We're
gonna
look
at
how
firms
can
estimate
the
marginal
value
of
the
last
room
they
sell,
the
seat
on
the
plane.
Or
the
last
rental
car
in
the
parking
lot.
And
then
they
use
this
marginal
value
then
to
control
inventory,.
Or
to
set
prices
going
forward.
This
course
serves
as
a
solid
foundation
in
revenue
management
for
those
of
you
who
are
relatively
new
to
the
area
and
as
more
of
a
reinforcement
of
the
core
concepts
for
those
of
you
with
prior
RM
experience.
Thanks,
and
welcome
and
I
hope
you
find
the
course
impactful.
1
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Transcript: Price and Duration Controls
Successful
revenue
management
has
effective
control
of
price
and
duration
of
stay—the
two
strategic
revenue
management
levers.
Consider
this
matrix
that
plots
firms
along
the
dual
axes
of
duration
and
price,
where
duration
is
controlled
or
uncontrolled
and
price
is
relatively
fixed
or
variable.
This
chart
provides
an
introduction
to
the
revenue
management
perspective.
Firms
in
industries
traditionally
associated
with
revenue
management
(hotels,
airlines,
rental
car
firms,
and
casinos)
are
able
to
apply
variable
pricing
for
a
service
that
has
a
specified
or
predictable
duration.
These
firms
are
in
quadrant
2.
To
obtain
the
benefits
associated
with
revenue
management,
industries
should
attempt
to
move
to
quadrant
2
by
implementing
the
appropriate
strategic
levers.
Most
hospitality
firms
find
that
the
more
their
firm
operates
in
quadrant
2,
the
higher
their
revenue
per
available
time-‐based
unit.
Not
all
firms
within
quadrant
2
industries
practice
revenue
management
or
practice
revenue
management
well.
For
example,
a
luxury
hotel
that
is
not
implementing
strict
length-‐of-‐stay
controls
across
its
limited
set
of
prices
effectively
operates
in
quadrant
3
due
to
the
type
of
guests
to
whom
it
caters.
2
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Assume
you
want
to
move
your
firm
to
quadrant
2;
what
are
some
of
the
things
you
should
consider?
If
you
have
one
price,
you
can
institute
multiple
prices.
A
variable-‐price
approach
moves
the
firm
from
quadrant
3
(with
few
prices)
to
quadrant
4
(with
many
prices).
In
quadrant
4,
you
have
several
prices
but
uncontrolled
duration.
To
add
duration
controls,
you
may
use
advance
reservations
to
forecast
demand,
preferably
by
rate
class
and
by
length
of
stay.
Now
if
you
are
able
to
incorporate
length-‐of-‐stay
controls
as
well
as
multiple
prices,
the
firm
is
better
positioned
to
move
to
quadrant
2.
Transcript: Customer Segmentation and Demand Controls
Using
customer
segmentation
and
inventory
controls
to
manage
revenue
is
commonplace
in
many
industries.
But
this
wasn’t
always
the
case.
The
airline
industry
has
a
strong
influence
in
their
use.
Today
you
find
a
lot
of
volatility
in
airfares,
but
this
is
a
relatively
new
phenomenon.
In
the
early
years
of
air
travel
U.S.
airlines
were
subjected
to
government
regulations
that
consistently
kept
fares
high
and
made
air
travel
a
luxury
item.
But
eventually
the
demand
for
more
affordable
air
travel
led
to
the
passing
of
the
Airline
Deregulation
Act
in
1979.
The
result
of
this
act
was
complete
elimination
of
fare
restrictions,
leaving
the
airline
industry
in
a
free
market.
Almost
immediately,
a
number
of
new
airlines
arose
to
compete
with
the
existing
carriers
and
the
number
of
passengers
dramatically
increased.
A
new
way
of
pricing
was
introduced
as
existing
carriers
(serving
guests
willing
to
pay
higher
prices)
now
also
had
to
offer
lower
prices
to
compete
with
new
entrant
airlines.
So
how
did
they
price?
They
began
with
segmenting
customers.
If
we
oversimplify
we
could
assume
there
are
only
two
types
of
customers
seeking
to
travel—business
customers
travelling
for
work-‐related
issues
and
leisure
travelers.
The
typical
business
traveler
is
willing
to
pay
a
higher
price
in
exchange
for
flexibility
of
being
able
to
book
a
seat
at
the
last
minute
(or
cancel
his
ticket
if
his
plan
changes)
while
the
vacation
traveler
is
willing
to
give
up
some
flexibility
for
the
sake
of
a
more
inexpensive
seat.
The
demand
from
the
price-‐sensitive
customer
tends
to
come
before
the
demand
from
business
customer.
But
with
multiple
price
points
and
demand
for
more
expensive
seats
arriving
after
price-‐sensitive
demand
the
airlines
had
to
determine
how
many
seats
they
should
sell
to
the
early
price-‐sensitive
customers
and
how
many
they
should
protect
for
late,
full-‐fare
customers.
If
too
few
seats
are
protected,
the
airline
will
lose
the
full-‐fare
revenue.
If
too
many
are
protected,
flights
will
leave
with
empty
seats.
Littlewood
(working
for
British
Airways)
proposed
a
way
to
make
this
determination.
He
proposed
that
discount-‐fare
bookings
should
be
accepted
as
long
as
their
value
exceeds
that
of
anticipated
full-‐fare
bookings,
assuming
that
customers
can
be
segmented
according
to
when
they
purchase
their
tickets.
This
simple,
inventory
control
system
was
the
beginning
of
what
eventually
lead
to
revenue
management.
3
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Using
Littlewood’s
approach
we
can
designate
two
fare
classes
as
having
fares
of
R2
and
R1,
where
R2
is
greater
than
R1.
The
demand
for
class
R1—the
lower
fare—comes
before
demand
for
class
R2.
The
question
now
is
how
much
demand
for
class
R1
should
be
accepted
so
that
the
optimal
mix
of
passengers
is
achieved
and
the
highest
revenue
is
obtained?
Littlewood
suggested
closing
down
class
R1
when
the
certain
revenue
from
selling
another
low
fare
seat
is
less
than
the
expected
revenue
of
selling
that
same
seat
at
the
higher
fare.
In
other
words,
as
long
as
the
probability
of
selling
all
remaining
seats
at
the
higher
price
is
greater
than
the
ratio
of
the
lower
price
over
the
higher
price,
we
are
better
off
not
selling
at
the
low
price
and
keeping
it
for
the
high
price.
This
is
our
Target
Probability.
Let’s
look
at
an
example.
Grand
Sky
Airlines
sells
tickets
on
one
of
its
85-‐passenger
planes
for
€150
(the
discounted
fare)
and
€250
(the
full-‐fare).
In
general,
their
customers
are
aware
of
the
pricing
and
those
seeking
discounts
tend
to
book
early.
Sean,
one
of
the
managers
at
Grand
Sky,
knows
that
he
can
fill
his
entire
plane
at
€50
per
seat
if
he
so
desires,
but
at
some
point
it
is
best
to
stop
selling
discounted
seats
and
reserve
some
inventory
for
later
arriving
higher
yielding
(€250)
passengers.
How
does
Sean
calculate
this
target
or
the
point
at
which
to
stop
selling
€150
seats
and
reserve
the
remaining
seats
for
the
€250
customers?
Using
Littlewood’s
rule
(R1
divided
by
R2)
we
can
calculate
Sean’s
target
probability.
In
this
case
it
is
.6
or
60%.
As
long
as
the
probability
of
selling
all
remaining
seats
(“n”
seats)
at
€250
is
equal
to
or
greater
than
60%
then
Grand
Sky
is
better
off
selling
seats
at
€250.
Now
we
need
to
calculate
the
probability
of
selling
“n”
or
more
seats.
We
use
historical
data
to
help
calculate
the
probability
of
future
events.
The
graph
shows
the
number
of
€250
seats
Grand
Sky
Airlines
sold
each
day
for
the
last
100
days.
4
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
For
example,
24
€250
seats
were
sold
on
one
of
the
days
and
33
€250
seats
were
sold
on
12
of
the
days.
Now
we
can
calculate
the
probability
of
selling
at
least
a
certain
number
of
seats
at
€250
and
compare
that
number
to
our
target
of
60%.
To
calculate
this
probability,
divide
the
number
of
days
we
sold
“n”
or
more
seats
by
the
total
observations.
We
could
start
the
calculations
at
any
point
(selling
24
or
more
seats,
selling
25
or
more
seats,
etc.).
But
we
can
use
our
graph
to
select
a
reasonable
starting
point.
On
the
graph
we
see
that
the
mid-‐point
is
around
34
seats.
This
will
make
a
good
starting
point
for
our
calculations.
It
may
be
easier
to
calculate
these
probabilities
if
we
look
at
the
data
in
a
table
format.
This
table
displays
the
same
data
that
we
just
saw
in
the
form
of
a
graph.
We
want
to
calculate
the
probability
that
there
will
be
future
demand
for
34
or
more
seats.
Start
by
finding
the
number
of
days
34
or
more
seats
were
sold
in
the
past.
To
do
this,
add
the
frequencies
when
demand
was
34
or
more
seats.
We
add
the
frequency
of
demand
at
34,
35,
etc.
up
to
41
together
to
arrive
at
56
days
when
demand
was
34
or
more
seats.
Now
divide
56
by
the
total
observations
(100).
This
gives
us
the
probability
that
demand
will
be
greater
or
equal
to
34
seats
at
€250
as
.56
or
56%.
If
we
do
the
same
calculation
for
the
sale
of
33
or
more
seats
we
arrive
at
a
probability
of
68%.
Now
we
can
compare
these
probabilities
to
our
target
probability.
Remember,
as
long
as
the
probability
of
selling
all
remaining
seats
at
€250
is
≥
60%
then
Grand
Sky
is
better
off
selling
seats
at
€250
rather
than
€150.
The
probability
of
selling
the
33rd
seat
at
€250
is
68%
thus
greater
than
60%
and
the
probability
of
selling
the
34th
seat
at
€250
is
56%,
less
than
60%.
5
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
We
can
also
look
at
this
from
a
monetary
viewpoint.
The
probability
of
selling
34
or
more
seats
at
€250
is
56%.
To
find
the
expected
revenue
of
the
last
seat
we
sell
multiple
.56
by
€250
which
equals
€140.
Given
the
expected
revenue
of
the
34th
seat
is
€140,
less
than
the
€150
we
obtain
for
certain
if
sold
as
a
discounted
seat,
34
seats
is
not
our
threshold.
Let’s
look
at
selling
33
or
more
seats.
This
probability
is
68%,
giving
us
an
expected
revenue
from
selling
the
33rd
seat
at
€250
of
€170.
We
can
go
back
to
our
original
question:
What
is
the
point
at
which
to
stop
selling
€150
seats
and
reserve
the
remaining
seats
for
the
€250
customers?
Assume
that
on
our
85
seat
plane
the
85th
seat
is
sold
first
and
the
1st
seat
is
sold
last.
Thus
the
airline
is
better
off
selling
up
to
52
seats
(85
total
minus
33)
at
€150
and
reserving
the
remaining
33
seats
for
the
€250
paying
customers.
In
essence
we
are
calculating
the
expected
marginal
revenue
of
keeping
a
seat
(or
room)
for
later
arriving
higher
yielding
guests.
We
should
continue
to
sell
at
lower
discounted
rates
as
long
as
these
rates
exceed
the
expected
marginal
revenue
of
selling
at
higher
rates.
Littlewood,
K.
(1972).
Forecasting
and
control
of
passenger
bookings.
Proceedings
from
the
Twelfth
Annual
AGIFORS
Symposium,
Nathanya,
Israel.
Transcript: Class Protection
We’re
going
to
continue
our
discussion
on
using
Littlewood’s
rule.
Now
our
focus
is
going
to
change
from
controlling
rates
to
actually
controlling
segments.
We’ll
have
a
quick
recap
of
Littlewood’s
rule
and
then
we
will
move
to
how
we
use
that
technique
to
control
segments.
Remember,
Littlewood’s
rule
is
about
allocating
inventory
to
certain
price
classes.
Now
we’re
going
to
think
about
allocating
inventory
to
certain
types
of
business,
whether
that’s
a
transient
late-‐arriving
customer
or
a
group
traveler
who's
making
that
request
one
or
two
years
in
advance
of
check-‐in.
As
a
quick
sort
of
review,
suppose
we
have
two
prices,
€200
and
€250.
And
just
for
argument's
sake
let’s
say
today
is
Wednesday
the
20th
and
we’re
looking
at
controlling
inventory
for
next
Wednesday.
As
of
today
we
have
15
rooms
available
for
next
Wednesday.
The
decisions
we
need
to
make
today
are
about
those
15
rooms:
Should
we
continue
to
sell
some
of
those
at
€200
or
should
we
keep
them
all
for
€250-‐paying
customers?
Euro
paying
customers.
Right,
so
what
stage
is
a
function,
of
how
many
rooms
we
have
left
as
well
as
days
before
arrival.
When
do
we
want
to
stop
selling
at
200?
6
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
We’ve
collected
some
data
to
help
us
assess
those
potential
outcomes.
Basically,
this
data
would
be
the
number
of
requests
for
€250-‐paying
customers
in
this
last
week
prior
to
arrival.
We’re
focused
on
demand
for
the
higher-‐yielding
class
versus
demand
for
the
lower-‐yielding
class.
Let’s
assume
we’ve
collected
data
for
over
say
the
last
100
Wednesdays
and
during
those
100
Wednesdays,
in
that
last
week
before
arrival,
the
average
demand
is
15.
That
demand
has
a
standard
deviation
of
5
to
represent
its
uncertainty.
Remembering
back
to
Littlewood’s
rule,
we
want
to
keep
selling
at
€200
as
long
as
that
200
exceeds
the
potential
revenue
from
selling
at
€250.
And
the
potential
revenue
from
selling
at
€250
is
the
probability
that
we
would
sell
all
those
remaining
rooms
at
250
times
250.
So
given
that
demand
has
a
mean
of
15
and
a
standard
deviation
of
5,
let’s
assume
some
distributional
form
for
that
demand.
For
ease,
let’s
assume
that
it
follows
a
normal
distribution,
so
a
nice
sort
of
symmetric-‐about-‐the-‐mean,
sort
of
bell-‐shaped
distribution.
We
can
use
some
built-‐in
functionality
in
Excel
to
help
us
estimate
how
many
rooms
to
keep
for
the
€250-‐paying
customers.
Excel
always
calculates
probabilities
from
the
left
hand
side.
Basically
the
probability
that
demand
is
less
than
or
equal
to
some
critical
level,
we
want
probability
the
demand
is
greater
than
or
equal
to
our
critical
level
being
€200
over
€250.
7
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Excel
has
a
function
called
NormInv
and
what
NormInv
does,
or
Norm
Inverse
for
the
full
version
of
that
formula-‐-‐we
provide
a
probability
and
some
description
of
that
demand,
in
our
case
the
mean
of
15
and
the
standard
deviation
of
5,
and
it
tells
us
the
number
that
corresponds
to
that
probability.
So
in
Excel
we
would
simply
use
Norm
Inverse
of
1
minus
200
over
250,
15,
and
5,
and
that
would
return
to
us
10.8.
That
basically
means
that
the
probability
of
us
selling
10.8
or
more
rooms
is
200
over
250.
So
if
we
kept
exactly
10.8
rooms
for
the
€250-‐paying
customers,
we
would
be
indifferent
between
that
10.8th
room
as
a
€250
room
versus
it
as
a
€200
room.
8
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Given
we
can’t
sell
partial
rooms,
and
then
we’re
going
to
keep
10
rooms.
So
basically
the
probability
of
selling
10
rooms
is
a
little
more
than
200
over
250,
but
keep
in
mind
that
if
we
were
to
keep
11,
the
probability
would
be
less
than
200
over
250.
So
our
logic
here
is
to
keep
10
rooms
and
allow
us
to
keep
selling
up
to
5
more
rooms
at
€200.
Now
that
we
can
have
a
solid
idea
of
how
we
might
use
Littlewood’s
rule
to
calculate
how
many
rooms
to
keep,
we
can
extend
that
now
to
segments.
Keep
in
mind
that
group
requests
typically
are
made
one,
two,
even
three
years
prior
to
check-‐in.
These
are
large
conferences
looking
for
large
blocks
of
rooms,
typically
at
very
big
discounts.
And
so
one
of
the
questions
that
you
have
to
face
is
what
part
of
my
hotel
or
what
segment
of
my
rooms
do
I
want
to
keep
for
these
low-‐yielding,
early-‐arriving
customers.
Obviously,
they’re
very
valuable
customers,
but
you
don’t
want
to
sell
all
your
rooms
to
these
customers
because
you
have
later-‐arriving,
higher-‐yielding
customers.
Right,
so
we
could
use
the
same
Littlewood
logic
to
look
at
this.
I'm
going
to
sell
X
rooms
to
groups,
given
the
probability
of
selling
capacity
minus
X
to
higher-‐yielding
people
as
the
same
rate
as
that
group
class.
Obviously,
we’d
like
to
keep
all
our
rooms
for
these
people
if
we
could
stock
out.
We
estimate
that
probability
of
stock
out
using
our
Norm
Inverse
function,
given
the
ADR
for
the
transients
9
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
versus
the
ADR
for
the
corporate
and
government
versus
the
ADR
for
the
group.
This
way
we’re
calculating
those
allocations
across
those
segments,
and
in
essence
determining
what
our
mix
is
for
our
property
Transcript: Ideal Car Rental—Types of Cars
Peter
Carter,
at
Ideal
Car,
has
100
days
of
data
showing
daily
rentals.
We
can
use
this
information
to
help
Peter
determine
how
many
cars
he
should
stock,
striking
a
reasonable
balance
between
utilization
rates
and
the
possibility
of
running
out
of
cars
with
the
subsequent
loss
of
revenue.
The
chart
shows
the
monthly
revenue
and
costs
for
the
three
types
of
cars
in
Ideal’s
fleet.
The
first
step
in
solving
our
problem
is
to
find
how
many
times
a
car
must
rent
to
cover
its
fixed
costs.
We’ll
demonstrate
with
the
economy
class
car.
The
economy
cars
have
monthly
fixed
costs
of
€336
(€256
in
lease
plus
€80
in
insurance).
Given
that
each
economy
car
nets
€24
per
rental
(€26
rental
rate
minus
€2
in
variable
cleaning
costs)
that
means
a
car
needs
to
rent
at
least
14
times
per
month
to
cover
its
fixed
costs
(i.e.
needs
to
rent
14
times
to
break
even)
as
€336
divided
by
€24
equals
14.
We’ll
assume
that
each
month
has
30
days.
That
means
on
any
given
day
for
a
car
to
be
profitable
it
needs
to
have
a
probability
of
renting
of
14/30
or
46%.
Now
instead
of
30
days,
let’s
look
at
100
days
of
data
for
economy
cars.
The
chart
shows
the
frequency
of
the
number
of
cars
rented
during
the
last
100
days.
On
2
of
the
100
days
only
10
economy
cars
were
rented.
There
is
also
two
days
when
11
cars
were
rented.
On
7
days
12
cars
were
rented
and
so
on.
If
Ideal
had
stocked
10
cars
then
they
would
have
rented
all
10
cars
on
10
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
all
100
days
(because
demand
is
never
less
than
10).
Each
of
these
10
cars
is
very
profitable.
If
they
stocked
11
cars
then
the
11th
car
would
have
rented
on
all
days
except
the
two
days
where
demand
was
only
10.
The
probability
of
demand
for
the
11th
car
is
98/100
or
98%
Remember,
for
a
car
to
be
profitable
it
needs
to
have
a
probability
of
renting
of
≥
46%.
After
calculating
all
the
probabilities
we
see
that
19th
car
rents
enough
to
cover
fixed
costs
and
generates
a
little
profit.
Whereas
if
we
stocked
20
cars
the
20th
car
would
not
rent
often
enough
to
cover
its
fixed
costs.
In
the
100-‐day
period
we
rent
20
or
more
cars
only
43
times,
thus
the
probability
of
renting
the
20th
car
is
43%,
less
than
the
46%
we
need.
Another
way
to
look
at
this
is
by
the
number
of
times
the
car
rents.
Remember,
in
any
given
month
a
car
must
rent
at
least
14
times
to
generate
enough
contribution
to
cover
fixed
costs.
Because
we
are
now
considering
a
little
more
than
3
months
worth
of
data
a
car
must
rent
100/30
X
14
or
46
times
to
be
profitable—more
than
the
20th
car
but
less
the
19th
car.
11
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Transcript: Ask The Expert: Upgrading/Upselling Opportunities
When
are
upgrades
and
upsells
useful?
Upgrades
are
especially
useful
when
there
is
a
mismatch
between
supply
and
demand.
There
are
several
reasons
why
capacity
mismatches
may
occur
in
practice,
including
forecast
errors
and
strategic
supply
limits
that
aim
to
skim
revenues
from
customers
with
high
willingness
to
pay.
Upgrades
become
a
key
managerial
lever
in
the
case
of
travel
and
service
industries
in
general
when
capacity
is
relatively
fixed
and
difficult
to
change
in
the
short
run
as
demand
fluctuates
over
time.
How
are
upgrades
useful?
Upgrades
help
balance
demand
and
supply
by
shifting
excess
capacity
of
high-‐grade
products
to
low-‐grade
products
with
excess
demand.
Upgrading
allows
firms
to
get
consumers
to
commit
to
purchases
at
lower
prices
and
then
extract
additional
revenues
with
the
upgrade/upsell.
What
are
some
of
the
main
concerns
with
upgrading?
In
addition
to
potentially
not
having
enough
high-‐valued
inventories
available,
upgrading
can
create
strategic
consumer
issues
especially
for
those
receiving
free
upgrades.
Consumers
tend
to
expect
upgrades
and
may
become
dissatisfied
if
usual
upgrades
become
unavailable.
What
type
of
data
do
you
need
to
determine
if
and
when
you
should
upgrade?
The
key
to
proper
management
of
upgrades
is
a
solid
understanding
of
total
demand
for
higher-‐valued
inventory
and
when
this
demand
materializes.
Essentially
you
must
be
able
to
estimate
the
likelihood
that
you
won't
need
that
high-‐valued
room
once
you've
upgraded
it
and
made
it
available
to
a
lower-‐valued
customer.
Transcript: Simultaneous Decision Making
Up
until
this
stage,
we’ve
been
focusing
on
a
single
constraining
resource.
Right,
how
many
rooms
should
I
allocate
to
which
different
prices,
how
should
I
manage
the
seats
on
my
particular
flight?
Going
forward,
going
to
add
some
complexity
basically,
so
we’re
going
to
focus
on
not
just
one
resource
but
multiple
resources,
so
you
can
think
of
this
as
guests
staying
multiple
nights
at
your
property
or
individuals
flying
with
your
airline
but
stopping
at
interconnected
cities
and
moving
on
to
subsequent
cities.
Under
this
context
of
guests
staying
12
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
more
than
one
night
or
flyers
having
interconnecting
traffic,
it
results
in
consumers
purchasing
different
products
using
the
same
resource.
So
I
might
have
a
guest
who
checks
in
today
for
a
one-‐night
stay.
I
might
have
another
guest
who
checks
in
today
for
a
two-‐night
stay.
Both
of
those
guests
are
staying
tonight,
but
they
each
bought
a
different
product.
One
bought
a
one-‐night
stay,
one
bought
a
two-‐night
stay,
but
they’re
both
using
rooms
tonight.
So
when
I
decide
how
many
rooms
to
allocate
to
each
of
those
two
different
product
classes,
I
need
to
realize
that
they’re
both
using
inventory
tonight.
So
going
forward,
we’re
going
to
think
about
how
to
incorporate
that
complexity
in
my
allocation
decisions.
This
might
be
clarified
with
a
simple
example.
So
let’s
look
at
our
property
for
the
upcoming
week.
We
have
a
very
simple
structure
here.
We
have
two
prices,
150
and
200
Euros,
and
guests
stay
one
or
two
nights.
13
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
In
this
particular
example,
we
forecasted
demand
for
the
upcoming
week,
and
it
turns
out
that
only
Wednesday
is
booming
with
us
and
on
Wednesday
we
have
demand
in
excess
of
capacity.
So
managing
inventory
on
Wednesday
is
relatively
straightforward.
We
want
to
make
sure
we
accept
requests
that
bring
in
as
much
revenue
as
possible
and
reject
those
requests
that
bring
in
less
revenue.
So
in
this
context
we’d
accept
two-‐night
stays
at
200,
we’d
accept
two-‐night
stays
at
150,
but
we
may
reject
some
one-‐night
stays
at
150
given
their
lower
revenue.
Now
we
extend
this
example
to
not
just
Wednesday
having
demand
in
excess
of
capacity
but
also
Thursday.
So
now
we
have
two
constraining
resources
and
while
it
seems
still
straightforward,
if
we
were
to
maximize
revenue
on
Wednesday
by
accepting
two-‐night
stays
and
rejecting
some
one-‐night
stays,
and
then
move
on
to
Thursday
and
maximize
revenue
on
Thursday
by
accepting
some
two-‐night
stays
and
rejecting
some
one-‐night
stays,
we
quickly
realize
that
the
decisions
I
made
on
Wednesday,
i.e.
the
two-‐night
stays
I
accepted
on
Wednesday,
well,
those
people
are
now
staying
on
my
property
on
Thursday,
and
those
14
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
decisions
I
made
on
Wednesday
impact
the
decisions
I
can
make
on
Thursday.
So
I
can’t
just
make
my
Wednesday
decisions
first
and
then
my
Thursday
decisions,
I
really
need
to
make
my
Wednesday
and
Thursday
decision
simultaneously
and
not
just
Wednesday
and
Thursday,
but
because
the
guests
who
stayed
two
nights
on
Tuesday
are
going
to
be
on
my
property
on
Wednesday
and
impact
my
Wednesday
decisions
and
then
impact
my
Thursday
decisions,
I
really
need
to
make
my
decisions
for
that
whole
week
simultaneously
versus
one
at
a
time.
This
moves
us
to
this
framework
of
simultaneous
decision-‐making.
One
of
the
common
aspects
of
this
simultaneous
decision-‐making
framework
is
that
most
of
these
settings
have
some
sort
of
limited
resource.
This
limited
resource
might
be,
as
in
our
earlier
example,
how
many
rooms
we
have
available
on
Wednesday
and
Thursday
or
it
might
15
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
be
how
much
cash
you
have
on
hand
to
purchase
items
or
it
might
be
how
many
items
you
have
in
stock
which
you
can
use
to
manufacture
goods
and
sell
to
consumers.
The
thing
that’s
common
across
this
framework
is
this
constrained
or
limited
resource.
Our
goal
is
to
mathematically
model
these
and
as
a
function
of
that
we
need
to
be
able
to
evaluate
performance.
And
the
easiest
way
to
evaluate
performance
is
to
have
some
sort
of
single
unifying
objective.
So
from
our
context
we’re
going
to
maximize
revenue
or
maximize
profit,
but
in
other
contexts
you
might
want
to
minimize
cost,
right?
We
have
to
be
able
to
map
this
performance
metric,
revenue,
to
our
decisions,
how
many
rooms
to
accept
across
each
of
the
rate
classes
and
lengths
of
stays.
In
addition
to
having
both
this
sort
of
unifying
objective,
which
is
a
function
of
these
decision
variables,
we’re
also
going
to
have
these
constraints,
right?
I
only
have
100
rooms
available
on
Wednesday
and
100
rooms
available
on
Thursday.
We’re
going
to
add
some
other
sort
of
logical
constraints;
those
logical
constraints
are
things
like,
I
can’t
accept
negative
reservations,
right?
So
that’s
pretty
easy
for
you
to
think
logically,
but
again
we’re
going
to
do
this
computationally
so
we
have
to
define
those
things
as
well.
Transcript: Optimization at Snap Électrique
Excel
Solver
is
a
tool
that
we
can
use
in
our
simultaneous
decision
making
process.
To
use
Solver,
we
must
build
a
model
that
specifies:
•
The
decisions
we
need
to
make;
we
refer
to
these
as
decision
variables
16
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
•
•
The
measure
to
optimize,
called
the
objective—for
example,
maximize
profit
or
minimize
costs
Limitations
on
how
we
make
decisions,
called
constraints—for
example,
limited
resources
Solver
will
find
values
for
the
decision
variables
that
satisfy
the
constraints
while
optimizing
(maximizing
or
minimizing)
the
objective.
We
will
use
Snap
Électrique
to
help
describe
how
to
use
Solver.
We
begin
with
the
decision
variables.
They
usually
measure
the
amounts
of
resources
to
be
allocated
to
some
purpose,
or
the
level
of
some
activity,
such
as
the
number
of
products
to
be
manufactured.
For
Snap
Électrique,
we
need
to
decide
how
many
of
each
of
the
four
products
to
make.
Once
we
define
the
decision
variables,
the
next
step
is
to
define
the
objective,
which
is
normally
some
function
that
depends
on
the
decision
variables.
For
Snap,
the
objective
is
to
maximize
profit.
We
know
that
each
LCD
touch
screen
yields
a
profit
of
€29,
each
integrated
audio
system
€32,
each
voice
and
audio
processor
€72,
and
each
custom
kiosk
€54.
Then
our
objective
function
might
be:
(€29
times
the
number
of
LCDs)
+
(€32
times
the
number
of
integrated
audio
systems)
+
(€72
times
the
number
of
voice
and
audio
processors)
+
(€54
times
the
number
of
custom
kiosks).
We’d
be
finished
at
this
point,
if
the
model
did
not
require
any
constraints.
In
most
models
constraints
play
a
key
role
in
determining
what
values
can
be
assumed
by
the
decision
variables
and
what
sort
of
objective
value
can
be
attained.
It
is
the
constraints
that
require
us
to
use
optimization
models
like
Solver.
17
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Constraints
reflect
real-‐world
limits
on
production
capacity,
market
demand,
available
funds,
and
so
on.
To
define
a
constraint,
you
first
compute
a
value
based
on
the
decision
variables.
Then
you
place
a
limit
(≤,
=,
or
≥)
on
this
computed
value.
Many
constraints
are
determined
by
the
physical
nature
of
the
problem.
For
example,
if
our
decision
variables
measure
the
number
of
products
of
different
types
that
we
plan
to
manufacture,
producing
a
negative
number
of
products
would
make
no
sense.
This
type
of
non-‐
negativity
constraint
is
very
common.
Often
we
have
constraints
that
require
decision
variables
to
assume
only
integer
(whole
number)
values
at
the
solution.
Integer
constraints
normally
can
be
applied
only
to
decision
variables,
not
to
the
quantities
calculated
from
them.
For
Snap,
we
cannot
allocate
more
resources
to
production
than
we
have
in
inventory.
Also
we
cannot
produce
negative
or
partial
products.
Now
let’s
look
at
the
model
we
created
for
Snap
Électrique.
In
the
worksheet,
we
have
reserved
cells
B4
though
E4
to
represent
our
decision
variables—the
optimal
mix
of
products
to
produce.
Solver
will
determine
the
optimal
values
for
these
cells.
The
profits
for
each
product
(€29,
€32,
€72,
and
€54)
are
entered
in
cells
B5,
C5,
D5,
and
E5.
This
allows
us
to
compute
the
objective
in
cell
F5.
Remember,
our
objective
is
the
sum
of
the
number
of
products
made
times
the
profit
margin.
In
cells
B8:E10,
we've
entered
the
amount
of
resources
needed
to
produce
each
type
of
product.
With
these
values,
we
can
enter
a
formula
in
cells
F8
to
F11
that
computes
the
total
amount
of
resource
used
for
any
number
of
products
produced.
Now
open
Solver.
This
may
be
under
the
Tools
menu
or
the
Data
menu
depending
on
what
version
of
Excel
you
are
using.
18
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
We
must
let
Solver
know
which
cells
on
the
worksheet
represent
the
decision
variables,
constraints,
and
objective
function.
In
the
Set
Objective
box,
type
or
click
on
cell
F5,
the
objective
function.
In
the
By
Changing
Variable
Cells
edit
box,
type
B4:E4
or
select
these
cells
with
the
mouse.
To
add
the
constraints,
click
the
Add
button
and
select
cells
F8:F10
in
the
Cell
Reference
edit
box
(this
will
show
the
number
of
units
or
hours
needed),
and
select
cells
G8:G10
in
the
Constraint
edit
box
(the
number
of
units
or
hours
available).
We
can
only
use
equal
to
or
less
than
the
amount
of
units
or
hours
we
have
in
stock.
The
constraint
is
set
to
≤.
Define
the
non-‐negativity
constraint
on
the
decision
variables.
Depending
on
the
version
of
Excel,
we
do
this
by
making
sure
the
Make
Unconstrained
Variables
Non-‐Negative
box
is
checked
or
click
Options
then
Assume
Non-‐Negative.
19
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Also
click
on
Assume
Linear
Model
or
use
Simplex
LP,
depending
on
your
version
of
Solver.
To
find
the
optimal
solution
click
Solve.
When
the
next
window
appears
click
OK.
After
a
moment,
the
Solver
returns
the
optimal
solution
in
cells
B4
through
E4
and
a
new
window
appears.
Here
are
the
results.
This
shows
that
we
should
build
zero
LCD
Touch
Screens.
Transcript: Marginal Value of Last Room Sold
What
happens
when
I
accept
a
reservation?
When
I
accept
a
reservation,
assuming
the
guest
shows,
I
receive
the
revenue
from
that
individual.
I
also
decrease
my
capacity
to
sell
to
subsequent
consumers
by
that
reservation.
The
act
of
accepting
a
reservation
really
decreases
the
available
capacity
to
your
hotel.
Thinking
about
this
from
a
marginal
value
standpoint,
I
don’t
want
to
accept
a
reservation
unless
it's
at
least
as
high
as
the
marginal
value
of
that
room.
Here
we
determine
how
many
reservations
to
accept
across
our
two
rate
classes
of
€350
and
€250
where
guests
can
stay
one,
two,
or
three
nights.
20
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
We
decided
which
reservations
to
accept
with
the
goal
of
maximizing
revenue.
In
that
context
we
made
almost
€515,000
(€514,850).
Now
let’s
step
back
and
say
Well,
what
if
I
had
one
more
room
on
the
14th?
On
the
very
first
day,
if
I
had
179
rooms,
what
if
I
had
180
rooms?
We’ll
just
simply
change
the
rooms
that
we
have
available
from
179
to
180
and
rerun
our
optimization
mode.
If
we
rerun
our
optimization
model,
it
turns
out
that
our
revenue
now
is
€515,000.
So
by
having
one
more
room
we
can
make
€515,000
versus
earlier
we
were
making
€514,850.
So,
in
essence,
that
one
incremental
room
generated
€150
incremental
Euros.
Our
goal
now
is
to
take
that
idea
from
179
to
180
and
sort
of
automate
that.
It
turns
out,
given
we’re
doing
things
computationally,
that’s
relatively
easy.
When
our
results
come
back,
we
simply
click
on
the
sensitivity
report
on
the
right
side,
and
by
doing
that
we
generate
what
is
referred
to
as
the
shadow
prices.
What
we
see
here
in
the
very
first
row
of
that
shadow
price
table
is
€150.
So,
corresponding
to
cell
H4,
was
the
rooms
that
were
available
on
December
14th,
we
have
a
shadow
price
of
€150,
which
before
we
calculated
manually.
21
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
So
if
you
think
about
things
on
the
14th,
if
I
had
180
rooms
versus
179
rooms
I
would
have
made
€150
more.
If
I
had
178
rooms
versus
179
rooms,
I
would
lose
€150.
If
we
go
back
and
compare
the
solutions
to
having
179
versus
180
rooms,
we
see
some
very
interesting
results.
When
we
had
179
rooms
available
on
the
14th,
we
accepted
44
reservations
for
three-‐night
stays
at
€250.
When
we
had
180
rooms
on
the
14th,
we
actually
accepted
45
of
those
three-‐night-‐stay
€250
requests.
Because
those
are
three-‐night
requests,
those
individuals
are
now
going
to
stay
into
the
15th
and
into
the
16th.
Because
of
that,
if
I
accept
that
three-‐
night
stay
on
the
14th,
I
have
to
accept
less
stays
on
subsequent
days.
It
turns
out
on
the
15th
I
accept
one
less
€350
one-‐night
stay.
If
you
look
at
the
16th,
on
the
16th
before
I
accepted
one
three-‐night
stay
at
€250,
now
I
accept
no
three-‐night
stays
at
€250.On
the
17th
before
I
accepted
19
two-‐night
stays
at
€250;
now
on
the
17th
I’ve
accepted
20
of
those
two-‐night
stays
at
€250.
22
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
You
see,
the
simple
act
of
having
one
more
room
available
on
the
14th
has
this
massive
chain
reaction
on
subsequent
stay
days.
It’s
this
chain
reaction
that
creates
this
odd
marginal
value.
We
see,
then,
that
we
can
generate
those
shadow
prices
for
all
subsequent
days.
It
turns
out
for
the
subsequent
days
the
shadow
prices
are
much
more
straightforward,
either
€350
or
€250.
The
next
part
is
how
we
use
these
shadow
prices.
Just
like
having
one
more
room
on
the
14th
we
generate
€150.
Having
one
less
room
would
decrement
us
by
€150.
The
same
thing
on
the
15th-‐If
I
had
one
more
room
I
could
increase
my
revenue
by
€350.
If
I
had
one
less
room
I
would
lose
€350.
My
focus
is
now
is
back
to
our
marginal
analysis,
thinking
about
accepting
our
reservation.
If
I
accept
a
reservation
for
the
14th
that
means
instead
of
having
179
rooms
I
have
178
rooms.
If
I
only
have
178
rooms,
I’m
going
to
lose
€150.
Logically,
I
would
not
accept
any
reservation
unless
it
brought
in
at
least
€150.
23
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
That’s
fairly
straightforward
for
us—it
means
we’re
going
to
accept
reservations
at
€250
and
€350.
If
you
look
at
the
15th,
now
it
says
if
we
had
one
less
room
on
the
15th,
our
revenue
would
go
down
by
€350.
That
tells
us
is
we’re
going
to
close
the
15th
to
€250
reservations,
specific
to
€250
one-‐night
stays.
The
marginal
value
on
the
15th
is
€350,
whereas
a
one-‐night
request
for
a
€250
only
brings
in
€250.
If
we
go
back
to
the
14th,
if
a
guest
was
going
to
stay
two
nights
on
the
14th,
so
he
is
going
to
consume
one
room
on
the
14th
and
one
room
on
the
15th,
so
logically
he
has
to
bring
in
revenue
in
access
of
those
two
marginal
values—that
is,
the
€150
plus
the
€350,
which
is
a
total
of
€500.
That
means
while
everything
is
open
on
the
14th
and
I
have
closed
the
€250
one-‐night
stays
on
the
15th,
I’m
actually
going
to
sell
some
multi-‐night
€250s
on
the
14th,
i.e.
I
would
allow
reservations
to
be
made
to
the
€250
rate
for
a
two-‐night
stay
on
the
14th
because
that
would
bring
in
€500
worth
of
revenue.
Keep
in
mind
the
marginal
value
here
is
the
€150
plus
the
€350
for
a
total
of
€500.
Transcript: Using Rate And Availability Controls
In
this
lesson
we
examine
rate
and
availability
controls
at
the
Hotel
Ithaca.
You
will
have
an
opportunity
to
practice
in
the
following
lesson.
We’ll
use
the
Hotel
Ithaca
spreadsheet
and
Solver
to
do
allocations,
generate
shadow
prices,
and
use
the
shadow
prices
to
determine
availability.
The
information
on
this
tab
is
divided
into
five
parts.
Part
1
stores
the
decision
variables.
Part
2
stores
the
total
number
of
rooms
sold
on
each
day.
This
includes
both
arrivals
on
that
stay
date
as
well
as
stay-‐overs—that
is,
guests
who
checked
in
yesterday
or
the
day
before
and
stayed
two
or
three
nights.
Part
3
contains
the
rooms
available—that
is,
hotel
capacity
minus
any
reservations
(and
stay-‐
overs)
already
accepted
for
those
dates.
Part
4
stores
the
total
revenue
for
all
rooms
and
days
listed.
Part
5
displays
the
forecasted
demand
for
the
stay
dates
in
question.
We
have
already
built
the
model.
Now
we
need
to
run
Solver.
24
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.
SHA541:
Price
and
Inventory
Controls
School
of
Hotel
Administration,
Cornell
University
Our
objective
is
to
maximize
revenue
by
determining
the
number
of
reservations
we
will
accept
for
each
day,
rate
class,
and
length
of
stay.
We
have
two
constraints.
The
first
constraint
is
that
the
number
of
reservations
we
are
able
to
accept
must
be
less
than
or
equal
to
our
forecasted
demand.
The
second
constraint
is
that
the
number
of
reservations
we
accept
must
be
less
than
or
equal
to
the
rooms
available.
We
also
need
to
require
that
the
result
be
non-‐negative
and
use
“Simplex
LP”
for
a
solution
method.
Click
Solve.
Once
the
solution
comes
back,
click
Sensitivity.
Click
OK.
In
Excel,
we
navigate
to
the
Sensitivity
report
tab
and
copy
the
shadow
prices.
Navigate
to
the
Restrictions
tab.
We
have
already
created
a
table
to
use
in
calculating
our
demand.
Paste
the
copied
shadow
prices
into
the
shadow
price
column.
Now
we
can
determine
our
minimum
available
rates
by
using
the
shadow
prices.
The
bid
prices,
the
average
of
the
appropriate
shadow
prices,
become
our
minimum
daily
rates.
For
a
one-‐night
stay,
the
bid
price
is
equal
to
the
minimum
daily
rate.
The
bid
price
for
a
two-‐night
stay
is
the
average
of
the
first
two
nights’
shadow
prices.
The
bid
price
for
a
three-‐night
stay
is
the
average
of
the
shadow
prices
for
all
three
nights
of
the
stay.
Copy
these
three
formulas
down
to
fill
columns
C,
D,
and
E.
Now
we
want
to
check
to
see
if
our
rate
(€195,
€250,
or
€350)
is
greater
than
these
bid
prices.
If
our
rate
is
greater
than
the
bid
price,
then
the
rate
is
available.
If
the
rate
is
less
than
the
bid
price
our
rate
is
not
available.
In
F3,
G3,
and
H3
we
enter
these
formulas.
The
formula
will
place
an
X
in
the
cell
if
the
rate
is
closed
and
leave
the
cell
blank
if
the
rate
is
open.
Copy
the
columns
and
paste
into
the
remaining
cells
for
the
€250
and
€350
rates.
We
can
take
this
one
step
further
and
use
Excel’s
conditional
formatting
to
color-‐code
the
cells
and
make
it
easier
to
read.
The
green
cells
indicate
the
rate
is
available.
The
red
cells
indicate
the
rate
is
closed.
For
example,
on
Oct.
20th
we
will
accept
a
one-‐
or
two-‐night
reservation
at
€195,
but
a
three-‐night
reservation
is
closed
at
the
€195
rate.
The
lowest
rate
we
can
offer
for
a
three-‐night
reservation
is
€250
as
the
bid
price
is
€233.
25
© 2015 eCornell. All rights reserved. All other copyrights, trademarks, trade names, and logos are the sole property of their respective owners.