Preference variability; not all customers expect the same amount of capacity to be dedicated to them.
Demand management options:
Demand partitioning/segmentation: tie separate types of demand to separate capacities:
Bank: caching a cheque (teller) vs. apply for mortgage (mortgage dept.).
Regular service vs. information requests.
Appointments; e.g., hair dresser, doctor's office, care maintenance, hospital example
p. 159/160.
Price incentives.
Promoting off-peak demand:
Winter special on cleaning the lawn mower.
Off peak pricing of electricity, cell phone call evening/night time rates, hotel and airfare rates, etc.
Reservations -- preselling:
Reservations provide information on (reduce uncertainty of) future demand.
Reservations (appointments) allow allocation of demand to existing capacity.
However, no-shows, even when paid for imply unsold capacity.
--> overbooking: "hedge against no-show" (or extra revenue if no-show is already paid for).
Particularly important in service factories.
However, lose money when expected no-shows do show!
Reason: cost of one too many bookings > cost of a lost sale through a no-show.
"good overbooking strategy minimizes the expected opportunity cost and the expected cost
of turning away reservations." (p. 162)
Some intuitive notions to set the stage:
Problem: Suppose that the cost of a no-show is $x (x > 0) but
the cost of an OVERbooking; i.e., the
cost of you making one too many promises, is $0. What is your overbooking policy?
Problem: Suppose that the cost of a no-show is $0 but the cost
of an overbooking is $x (x > 0). What is your overbooking policy?
Problem: Suppose that the cost of a no-show is equal to the cost of an
overbooking. What is your overbooking policy?
Problem: Suppose that the cost of a no-show (x) is different
from the cost of an overbooking (y) and x,y > 0.
What is your overbooking policy?
Problem: Suppose that from experience you know the probabilities of specific numbers of no-shows.
Could you use that information in
determining an overbooking policy?
Surfside hotel example (p. 162-164):
No-shows (d)
Probability P(d)
Reservations overbooked x
Cumulative probability P(d<x)
0
.07
0
0
1
.19
1
.07
2
.22
2
.26
3
.16
3
.48
4
.12
4
.64
5
.10
5
.76
6
.07
6
.86
7
.04
7
.93
8
.02
8
.97
9
.01
9
.99
Opportunity cost of a room not occupied because of a no-show: $40.
Estimated opportunity cost: unit opportunity cost × expected number of no-shows.
Expected number of no-shows = Σ{(n no-shows) × P(n no-shows)}.
0(.07) + 1(.19) + 2(.22) + ... + 9(.01) = 3.04 (See Table 11.5 for P(n no-shows)).
Unit 'walking' cost = cost of 'walking' a customer ((over)booked customer
(customer whose demand cannot be served by your existing capacity): $100.
Overbooking loss table (p. 163):
Define d: no-shows; x: overbooked
Expected loss(x overbooked) = (Estimated walking cost for x > d) + (Estimated opportunity cost for x < d).
Estimated walking cost for x > d = &Sigma{ ( (x - d) × unit walking cost ) x P(d) }.
Estimated opportunity cost for x < d = &Sigma{ ( (d - x) × opportunity cost ) x P(d) }.
See 'Expected loss' ($) in Table 11.6 (figure above).
Average expected loss is least at x (overbooking) = 2.
Alternative solution: marginal analysis; critical fractile criterion:
Cu: cost of under-estimating a no-show ($40).
Co: cost of over-estimating a no-show (= 'walking cost' = cost of one overbooking too many) ($100).
Critical fractile = Cu / (Cu + Co) = .286
P(d < x): probability of having booked too many (Table 11.5).
P(d < x) must be lower or equal to Cu / (Cu + Co).
Table 11.5: this is at x = 2.
Excursion for those who are interested: derivation of the critical fractile criterion:
Principle of marginal return: keep overbooking as long as the expected cost of the next overbooking is less than or equal to
the expected cost of its associated no-shows.
Cost of a single lost sale (no-show): Cu.
Cost of one overbooking too many: Co.
Estimated cost of the next overbooking: Co [ P(d < x) ]
Estimated cost of associated no-shows: Cu [ P(d ≥ x) ]
Yield management: Allocation of a fixed (capacity), perishable resource to match expected and potential demand
so as to maximize revenue and profit.
More or less fixed capacity; e.g., airlines, hotels.
Segmentation of customers in classes; e.g., frequent flyers, business class, cattle car.
Perishable inventory.
Product sold in advance.
Variable demand.
Low marginal sales cost; high marginal capacity cost
Problem: what sort of businesses does this apply to?
Combinations of pricing, segmentation and timing become possible:
e.g., Allocate 60% of the flight to economy class with a Saturday
night stay-over at 70% of standard price.
Hold expensive seats until shortly (??) before the flight.
Notice how these types of operations need comprehensive and fast information systems.
Test Questions:
How are the 'Expected Loss($)' amounts at the bottom of Table 11.6 computed? What do they mean?
Why is the probability P(d < x) (Table 11.5) a so-called 'cumulative probability?'
Name the three basic forms of capacity management; provide examples of each.
How does the principle of marginal return apply to overbooking?
Do the principles of capacity management also apply to a news-stand operator who must decide how many papers to order for the next day?
Or the bookstore owner who must decide how many of the new Harry Potter books to order, or the restaurant owner who
must decide how many steakes and how many salmon to order to have ready for this evening's customers?
What do we mean with the term 'yield management?' How is yield management different from other forms of
capacity management?
In a service business, even if we know the number of customers that will arive in the next time interval,
there might still be variation in their demand on capacity. What are some of the causes of this variation?
Do you see a relationship between capacity planning and queue time?
