Professor Lijian Lu Fall 2021

ISOM2700 Operations Management

Practice Questions Set #6

(Solutions)

1. Inn at Penn has 200 rooms. For regular-fare customers, rooms are priced at $300 per night while the rooms are
priced at $700 per night for the high-paying customers who generally arrive at the last minute. The demand for
such high fare customers is distributed normally with mean 60 and standard deviation 50. Assume that there is
ample demand for regular-fare customers.

a. What should the protection level for the high fare be to maximize expected profit?

Underage Cost = Cu = Cost of holding too few rooms (lost sales)

= High fare room price - Regular fare room price
= 700 - 300
=400

Overage Cost = Co = Cost of holding too many rooms (leftover)

= Regular fare room price - Salvage value
= 300 - 0
= $300

Cu/(Cu+Co)=(700-300)/700=0.5714, z=0.18, Q=60+50*0.18=69.

b. Suppose that the Inn at Penn operates with the protection level of 80 rooms for high fare customers.
What is the probability that there are at least 5 rooms left unoccupied?

For 5 rooms to be left, demand for high-fare rooms should be at most 75.
z=(75-60)/50=0.3, from the Standard Normal Distribution Function, we get Φ(z)=61.79%.

2. JetRed Airways flies several daily flights from Philadelphia to Chicago. Based on historical data, the flight on
Wednesday evening before Thanksgiving is always sold-out. However, there are usually no-shows so the
airline decides to improve revenues by overbooking. The no-shows are Poisson-distributed with mean 8 and
the airline estimates that the cost of bumping a passenger (including full refund of purchase) is about 11 times
as much as the ticket price.
 ISOM2700 Practice Questions Set 6

Num. No Shows x Pr (Number of No show

<= x)
1 0.00
2 0.01
3 0.04
4 0.10
5 0.19
6 0.31
7 0.45
8 0.59
9 0.72
10 0.82
11 0.89
12 0.94
13 0.97
14 0.98
15 0.99
16 1.00

a. How many seats should JetRed overbook?

Suppose the ticket price is w. Then, Cu = w; Co = 11w -w = 10w

So, Cu/(Cu+Co)= w/(w+10w) =0.091

From the probability distribution table above, the cumulative probability of 4 no shows = 0.10 is just
higher than 0.091. Therefore, JetRed should overbook 4 seats.

b. JetRed management is dreading bad publicity around Thanksgiving so it decides instead that it does
not want to bump passengers more than 5% of the time. What is the maximum number of seats that
JetRed can overbook?

x = Number of overbooking; y = Number of no shows; T = Total number of seats

The question is asking the largest x to satisfy the inequality: Pr(# of no shows<x) <=5%

Because # of no shows and # of overbookings are integer, so Pr( y < x) = Pr( y <= x-1)

Pr(# of no shows<x) ≤ 0.05

=> Pr( y < x) ≤ 0.05

=> Pr(y ≤ x – 1) ≤ 0.05

When x=4, the company faces a probability of 5% that there are 3 or fewer no-shows.

2
 ISOM2700 Practice Questions Set 6

c. Suppose 6 seats are overbooked. How many seats can JetRed expect to have empty, on average?

This question is asking the expected number of empty seats if overbooked 6 seats.

From the distribution table of no-shows provided in this question, we can get the following table.

Number of
Num. Pr (Number
Pr(Number of Empty Seats if
No of No show
No show = x) overbooked 6
Shows x <= x)
seats
1 0 0 0
2 0.01 0.01 0
3 0.04 0.03 0
4 0.1 0.06 0
5 0.19 0.09 0
6 0.31 0.12 0
7 0.45 0.14 1
8 0.59 0.14 2
9 0.72 0.13 3
10 0.82 0.1 4
11 0.89 0.07 5
12 0.94 0.05 6
13 0.97 0.03 7
14 0.98 0.01 8
15 0.99 0.01 9
16 1 0.01 10

Expected number of empty seats = ∑16

x=1 Yx Pr(Number of No show = x )

Note: Yx is the number of empty seats if overbooked 6 seats (Y1 = 0, Y2 = 0, . . . , Y16 = 10), i.e., the
fourth column of above table.

Thus, the expected number of empty seats = 0*0+0*0.01+0*0.03+…+8*0.01+9*0.01+10*0.01=2.34

3
 ISOM2700 Practice Questions Set 6

3. (Capacity-based RM) HK4 is a television station that has 25 thirty-second advertising slots during each evening.
It is early January and the station is selling advertising for Sunday, March 24. They could sell all of the slots
right now for $4,000 each, but, because on this particular Sunday the station is televising the Oscar ceremonies,
there will be an opportunity to sell slots during the week right before March 24 for a price of $10,000. For now,
assume that a slot not sold in advance and not sold during the last week is worthless to HK4. To help make this
decision, the salesforce has created the following probability distribution for last-minute sales:

Number of slots, Pr(Exactly x

x slots are sold)
8 0.00
9 0.05
10 0.10
11 0.15
12 0.20
13 0.10
14 0.10
15 0.10
16 0.10
17 0.05
18 0.05
19 0.00

a. How many slots should HK4 sell in advance?

Cu = Cost of reserving too few slots right before March 24 = 10000 – 4000 = $6000;
Co = Cost of reserving too many slots right before March 24 = 4000 – 0 = $4000
Prob(D ≤ Q) = Cu / (Co + Cu) = 0.6
From the table, the optimal protection quantity is 13. Therefore, HK4 should sell 25 – 13 = 12 slots in
advance.

b. In practice, there are companies willing to place standby advertising messages: if there is an empty slot
available (i.e., this slot was not sold either in advance or during the last week), the standby message is placed
into this slot. Since there is no guarantee that such a slot will be available, standby messages can be placed
at a much lower cost. Now suppose that if a slot is not sold in advance and not sold during the last week, it
will be used for a standby promotional message that costs advertisers $2,500. Now how many slots should
HK4 sell in advance?
Cu = Cost of reserving too few slots right before March 24 = 10000 – 4000 = $6000
Co = Cost of reserving too many slots right before March 24 = 4000 – 2500 = $1500
Prob(D ≤ Q) = Cu / (Co + Cu) = 0.8

From the table, the optimal protection quantity is 15. Therefore, HK4 should sell 25 – 15 = 10 slots in
advance.
4
 ISOM2700 Practice Questions Set 6

c. Suppose HK4 chooses a booking limit of 10 slots on advanced sales. In this case, what is the probability
there will be slots left over for stand-by messages?

If the booking limit is 10, there are 15 slots for last-minute sales. There will be standby messages if there
are 14 or fewer last-minute sales, which has probability 0.7.

d. One problem with booking for March 24 in early January is that advertisers will withdraw their commitment
to place the ad (typically this is a result of changes in promotional strategies; for example, a product may be
found to be inferior or an ad may tum out to be ineffective). Because of such opportunistic behavior by
advertisers, media companies often overbook advertising slots. HK4 estimates that in the past the number of
withdrawn ads has a distribution as follows.

S Pr(The number of
withdraws ≤ S)
0 0.00012
1 0.00123
2 0.00623
3 0.02123
4 0.05496
5 0.11569
6 0.20678
7 0.32390
8 0.45565
9 0.58741
10 0.70599

Assume each withdrawn ad slot can still be sold at a standby price of $2,500 although the company misses
an opportunity to sell these slots at $4,000 a piece. Any ad that was accepted by HK4 but cannot be
accommodated because there isn't a free slot costs the company $10,000 in penalties ($10,000 here is NET penality).
How many slots (at most) should be sold?

Cu = Cost of selling too few slots = 4000 – 2500 = $1500;

Co = Cost of selling too many slots = $10000
Prob(D ≤ Q) = Cu / (Co + Cu) = 0.1304
From the probability distribution above,
Prob(D ≤ 5) = F(5) = 0.11569, Prob(D ≤ 6) = F(6) = 0.20678, so the optimal overbooking quantity is 6,
that is, sell up to 25 + 6 = 31 slots.

4. (Price-based RM)

HKUST Center for the Arts is planning a concert by Hong Kong Philharmonic Orchestra. A recent survey from staffs
and students of HKUST reveals the willingness-to-pay for the concert as follows:

WTP (HKD) Demand if

price = WTP
150 1
125 10
100 32
75 72
50 97
25 113

5
 ISOM2700 Practice Questions Set 6

a. What is the revenue-maximizing price for the concert?

WTP (HKD) Demand if Revenue

price = WTP
150 1 150
125 10 1,250
100 32 3,200
75 72 5,400
50 97 4,850
25 113 2,825

The revenue-maximizing price is 75 HKD.

b. Estimate the parameters (a, b) of a linear demand model, D(p) = a – bp, using the WTP information. What
is the revenue-maximizing price based on the linear demand model?
∗
∑ni=1 xi yi − nx̅y̅
b = n 2
∑i=1 xi − nx̅ 2
a∗ = y̅ − b∗ x̅

a* = 140.27

b* = 0.984
Optimal price = a/2b = 140.27 / (2*0.984) = 71.275 HKD

c. HKUST Center for the Arts wishes to differentiate the price for staffs and students. The following table
presents the detailed information about willingness-to-pay for staffs and students, respectively. What is the
revenue-maximizing prices for staffs and students based on the WTP information? Is it worthwhile
differentiating prices for staffs and students?

WTP (HKD) Demand if Staff demand Students demand

price = WTP if price = WTP if price = WTP
150 1 1 0
125 10 9 1
100 32 22 10
75 72 30 42
50 97 39 58
25 113 52 61

WTP (HKD) Staff demand if Revenue from Students demand Revenue from
price = WTP Staff if price = WTP Students

150 1 150 0 0

125 9 1,125 1 125

100 22 2,200 10 1,000

75 30 2,250 42 3,150

50 39 1,950 58 2,900

25 52 1,300 61 1,525

6
 ISOM2700 Practice Questions Set 6

For both staffs and students, the profit-maximizing price is 75 HKD.

As the profit-maximizing price is the same as that without price differentiation, it is not worth-while to
differentiate prices between staffs and students.

5. (Price-based RM)

An airline offers two fare classes for coach seats on a particular flight: full-fare (H) class at $440/ticket and discount
fare (I) class at $218/ticket. There are 230 coach seats on the aircraft. Discount fare tickets must be purchased at least
three weeks in advance, and these tickets are expected to sell out. The demand for tickets at full fare is estimated to
have the following discrete distribution:

Demand for
tickets at full Probability Cum.Prob.
price
40 0.0200 0.0200
41 0.0600 0.0800
42 0.0400 0.1200
43 0.0100 0.1300
44 0.0600 0.1900
45 0.0700 0.2600
46 0.0200 0.2800
47 0.0300 0.3100
48 0.0300 0.3400
49 0.0500 0.3900
50 0.0300 0.4200
51 0.0500 0.4700
52 0.0400 0.5100
53 0.0600 0.5700
54 0.0900 0.6600
55 0.1100 0.7700

a. What’s the optimal protection level for full fare tickets?

Underage Cost = Cost of protecting too few seats = $440-$218 = $222

Overage Cost = Cost of protecting too many seats = $218

Pr(Demand for tickets <= Q) = critical fractile = 222/(222+218) = 0.505

Protection level = 52

b. What’s the optimal booking limit for low-fare seats?

Protection level = 52
Booking limit = 230 - 52 = 178

7
 ISOM2700 Practice Questions Set 6

6. (Capacity-based RM) Doubletree in Austin has 150 standard rooms. Doubletree generally sells those rooms
through two channels, one through their own website, call center and front desk usually at a high rate and the
other through agencies like Priceline at a low rate. Suppose Doubletree charges a high rate at $120 per room per
night through their own channel (they never mark down the price through their own explicit channel to avoid
any bad gambling image which hurts reputation) while “implicitly” sells some rooms to the agencies at a low
rate of $50 per room per night. Bargain customers who seek low rate usually will buy far in advance of the
premium customers through the agency channel.

To make it simple, suppose the customers always stay for one night, there is ample demand from the bargain
customers for the low rate, and the number of premium customers is however uncertain, which is distributed
according to the following table:

Number of high fare customers Probability

90 0.10
100 0.15
110 0.15
120 0.20
130 0.25
140 0.15

a. How many rooms shall Doubletree sell to the agencies like Priceline in advance?

Decision variable Q = how many rooms to reserve for the high fare customer
Cost of underage Cu = Cost of reserving too few rooms for high fare customers= 120 - 50 = $70.

Cost of overage Co = Cost of reserving too many rooms for high fare customers = 50 - 0 = $50.

The critical ratio=70/(70+50)=0.583. Therefore, we want to find optimal Q such that

Pr(number of high fare customers <= Q) = 0.583.

From the above distribution table, we get the cumulative probabilities below:

Number of high fare

Probability Cum.Prob.
customers
90 0.1 0.1
100 0.15 0.25
110 0.15 0.4
120 0.2 0.6
130 0.25 0.85
140 0.15 1

The cumulative probability of 120 high fare customers = 0.6, which is just higher than the critical fractile.
The smallest such Q is 120. Hence, the optimal protection level is 120 rooms and Double tree shall sell
150-120=30 rooms in advance to the agencies.

b. What is expected total revenue of Doubletree from both channels (using your solution from part a)?
The revenue from the agencies = 30*$50=$1,500;
The expected revenue from the premium customers is:
$120*(90*0.1+100*0.15+110*0.15+120*(0.2+0.25+0.15)) = $13,500
Thus the expected total revenue they have is 1,500 + 13,500 = $15,000.
8
 ISOM2700 Practice Questions Set 6

c. The number of no-shows at Doubletree has a distribution as that in the following table. Doubletree estimates
the cost of bumping a high fare customer is $320(including full refund of purchase). What is the optimal
maximum number of reservations to accept per day (suppose overbooked rooms are only sold to the agency
channel at $50)?
Number of no-shows Probability
0 0.05
1 0.05
2 0.10
3 0.15
4 0.15
5 0.20
6 0.20
7 0.10

Decision variable Q = number of overbooking

Cost of underage Cu =50.

Cost of overage Co =320-120=200.

The critical ratio=50/(50+200)=0.2

From the above distribution table, we get the cumulative probabilities below:

Number of no-shows Probability Cum.Prob.

0 0.05 0.05
1 0.05 0.1
2 0.1 0.2
3 0.15 0.35
4 0.15 0.5
5 0.2 0.7
6 0.2 0.9
7 0.1 1

The smallest number Q such that Pr (the number of no-shows <= Q) = 0.2 is 2.

Therefore, they should overbook 2 rooms and book a maximum of 150 + 2 = 152 rooms.

*Boundary issue: change the number in the future to avoid such

boundary confusions. (For example, change the third probability 0.1 to
0.09, and other probabilities also be changed accordingly.

9
 ISOM2700 Practice Questions Set 6

7. Le Meridien in San Francisco has 160 rooms. The hotel has an ample low fare demand at the room rate of $200
per night, but the demand from the high fare class which pays $450 per night on average, is uncertain. The high
fare demand is normally distributed with mean 60 and standard deviation 42.

a. How many rooms should Le Meridien protect for high fare customers to maximize expected revenue?
(Leave your answer in decimal form, i.e., no need to round to an integer value.)

Cost of underage = Cost of protecting too few rooms = $450-$200 = $250

Cost of overage = Cost of protecting too many rooms = 200 - 0 = $200
The critical ratio = Cu / (Cu + Co) = 250/450 =0.5556.

From z-table, Φ(0.13) =0.5517, Φ(0.14)=0.5557. The ratio 0.5556 falls between Φ(0.13) and Φ(0.14), we
*choose the entry with the higher z, i.e., z = 0.14. Thus, Q = µ + z* σ = 60 + 0.14 *42 = 65.88
We can also use Excel to get z value. NORMSINV (0.5556)=0.1398

b. Suppose Le Meridien sets a booking limit of 100 for low fare customers. How many high fare customers
does the hotel expect to turn away due to a lack of rooms? (This is not in the scope of this course)

The expected number of high fare customers that would be turned away is the expected loss sales, which is
L(z)* σ.

Setting the booking limit at 100 rooms implies a protection level of 60 rooms for the high fare customers.

z= (60-60)/42=0

From the Standard Loss Function Table (L(z)-table), we know L(0)=0.3989

L(z)* σ=L(0)*42= 0.3989* 42 = 16.75

*Notes:

• L(z) is the standard loss function, i.e., the expected number of lost sales as a fraction of the standard
deviation. The lost sales = 𝐿(𝑧) ∗ 𝜎𝑑𝑒𝑚𝑎𝑛𝑑
• Standard Loss Function Table (L(z)-table):
https://www2.isye.gatech.edu/~mgoetsch/cali/logistics_systems_design/appendices/standard_distributions.p
df (page3)
• We can also use Excel to get the value of L(z). L(z)=NORMDIST(z,0,1, FALSE)-z*NORMSDIST(-z)

10