Marketing Letters 13:2, 75–89, 2002
# 2002 Kluwer Academic Publishers. Manufactured in The Netherlands.
Internal Marketing at Continental Airlines: ConvincingEmployees that Management Knows Best
ASHUTOSH PRASAD* AND ERIN STEFFES
School of Management, The University of Texas at Dallas, Richardson, TX 75083
*Corresponding author. Tel.: þ 1-972-883-2027; Fax: þ 1-972-883-2799; E-mail: [email protected]
Received October 25, 2001; Revised February 5 2002; Accepted February 13, 2002
Abstract
Maintaining a high level of customer service quality is necessary for success, particularly for firms in service
industries, but employees need to be motivated to provide it. Management can provide the motivation through the
use of compensation schemes and internal marketing. In this paper we examine profit sharing compensation
schemes that reward employees for achieving customer service objectives. We discuss how such schemes should
be implemented, and explain why management needs to convince employees through internal marketing programs
about the benefits of its scheme. We obtain the optimal amount of profit that the firm should share with
employees, and the optimal effort it should spend on internal marketing. Finally, we relate the analysis to a
successful scheme implemented by Continental Airlines.
Key words: internal marketing, incentives, employee compensation
1. Introduction
Customer service quality is important for all firms; in particular, firms in service industries,
such as hotels, airlines, and health clubs, cannot remain competitive if they provide poor
service (Cowell, 1984; Heskett et al., 1994; Schlesinger and Heskett, 1991). Maintaining
and improving service quality depends on employees. For instance, in the airline industry,
the service sold to customers is timely, reliable transportation for customers and their
baggage (Bethune and Huler 1998). A large number of employees, from pilots to baggage
handlers, must work together to deliver a satisfactory customer experience. Motivating
employees to provide better service is essential because the employee-customer interaction
is the primary component of the service firm’s product, and quality can be created or
destroyed each time the customer interacts with the firm’s employees (Lovelock, 1983).
To motivate employees, firms use a combination of internal marketing and incentive
schemes (Bethune and Huler, 1998; Cooper and Cronin, 2000). In this paper, we examine
the use of internal marketing and its relationship to incentive schemes that reward the
achievement of specific objectives by employees.
Internal marketing has been defined as the expense and effort of motivating employees
to provide better customer service (Cooper and Cronin, 2000). It may encompass some or
all of the following practices: Communicating with the internal audience of the organiza-
tion through seminars, memos, and regular meetings between management and employees;
increasing interaction between employees and customers; improving internal customer
service; using evangelists to explain new initiatives; and continuous training of employees
(Cooper and Cronin, 2000; George, 1990; Pitt and Foreman, 1999). As Kotler (2000)
notes, it is clear that internal marketing must precede external marketing since it makes
little business sense to promise the customer a level of service that the firm’s employees are
unwilling or unable to provide.
For our incentive scheme, we examine profit sharing, which is the primary method for
aligning the employee’s goals with the goals of the firm (Coughlan and Sen, 1989).
Employees are compensated with a share of the firm’s profits, and because this profit is a
consequence of their collective action, there is a motivation for each employee to work
harder. An alternative method is to set a quality performance goal for each employee, but
in practice the high cost of monitoring employee performance makes this an infeasible
proposition. Individual effort, therefore, is unobserved (e.g. Basu, Lal, Srinivasan and
Staelin, 1985).
An example of a successful profit-sharing scheme is provided by Continental Airlines in
1995 (Bethune and Huler, 1998; Brenneman, 1998). For every month that Continental
ranked above average in on time flight performance, Continental employees would each
receive a payment of $65. Continental management had determined that if their flights
departed and arrived on time, the airline would earn an additional $5 million in profits.
They then decided that half of this profit would be given to the employees, which funded
the potential monthly incentive payment of $65. A great deal of internal marketing from all
levels of management was required to convince employees that the program would be
successful (Bethune and Huler, 1998).
Strong anecdotal and empirical evidence supports the assertion that Continental’s
scheme worked. After the profit sharing program was launched in early 1995, Continental
improved its flight performance by landing 80% of flights on time, beating the industry
average of 79% (Bethune and Huler, 1998). The positive impact of the scheme was also
verified in an empirical study that compared Continental’s on time performance at airports
that implemented the scheme to airports that did not, and found a significant improvement
in the former (Knez and Simester, 2001).
We now discuss why an internal marketing campaign is required in addition to the
profit-sharing scheme to motivate employees to provide better service quality. For profit
sharing to work, employees must be convinced that management’s desired outcome is
actually linked to their increased effort with some probability. Internal marketing is used to
convince employees about management’s certainty that the benefits of the scheme will be
realized. Table 1 discusses the limitations of three alternative schemes that do not include
internal marketing.
The fundamental argument is that if the firm has information that would induce the
employees to work harder then it should communicate this information to the employees in a
credible manner through internal marketing. Thus, just as advertising is required to com-
municate information to customers, internal marketing is required to communicate informa-
tion to employees. The following research issues arise from the descriptions in Table 1:
76 PRASAD AND STEFFES
RI1: Continental shared half of the $5 million profit gain that resulted from improved
flight performance with employees. When is profit sharing appropriate and what
is the optimal proportion to share?
RI2: Continental spent a great deal of internal marketing effort on motivating and
promoting the scheme to employees. How much money or effort should manage-
ment spend on internal marketing?
On examining the marketing literature related to these issues, we observe that theoretical
work on incentives has received wide application. Examples can be found in the areas of
salesforce compensation, employee compensation and franchising and channel manage-
ment (e.g. Basu et al., 1985; Chu and Desai, 1995; Hauser, Simester and Wernerfelt, 1997;
for a review see Coughlan and Sen, 1989). The literature on group incentives, in contrast,
is rather limited in marketing (e.g. Knez and Simester, 2001) and in the economics
literature (e.g. Holmstrom, 1982; Hermalin, 1998). In addition, we do not find theoretical
work on the use of internal marketing. This paper, therefore, makes an original contribu-
tion in filling this gap in the literature.
In the rest of the paper we develop a model with asymmetric information between the
firm and employees to answer these questions. Sections 2 through 4 develop and analyze
the model. Section 5 discusses results and Section 6 concludes.
2. The Model
Consider a firm with N identical employees. We assume that the particular task, such as
improving flight timeliness, can be treated as separate from regular work since it represents
an extra effort for which incentive must be provided (Hauser, Simester and Wernerfelt,
1997). Since the incentive is typically small, employees are assumed to be risk-neutral.
Employees expend effort, e, and have a disutility from effort given by e2=2. The firm may
also exert effort, denoted E at a cost E2=2.
Table 1. Alternative Schemes
Possible Scheme Limitation
1. Give $65 directly to employees. Employees will have no incentive to work
harder since they are assured of the money
regardless.
2. Offer a share of the profit realized due to
customer service improvement.
This works only if the employees and the firm
are equally informed. Otherwise, employees
don’t know what realized profit to expect;
hence less incentive.
3. Offer a share of the profit realized due to
customer service improvement, and inform
employees what the profit is expected to be.
Employees should rationally disbelieve the firm’s
reported expectation since it has an incentive to
lie by claiming a very high value (see Section 4
below).
INTERNAL MARKETING AT CONTINENTAL AIRLINES 77
The firm’s effort E consists of both a productive component and a component not
directly productive in achieving the objective even in normal times when no internal
marketing is required. When we consider the internal marketing scenario, we will dis-
tinguish it as the additional effort that the firm applies at that time, denoted DE. Internal
marketing programs such as communicating the incentive program to the employees,
training and educating the employees on how to achieve the goal, and monitoring and
planning the firm’s progress towards the goal, also have direct and indirect effects on
productivity, and therefore, are not distinguished from normal effort on this dimension. It
is best seen as enhanced effort during the period when management is trying to convince
employees about its scheme.
Let the function gðP
e;EÞ denote the total effort directed towards the goal by the firm
and employees, whereP
e is the sum of the efforts of all employees. The success of the
effort in terms of how well it translates into profit is uncertain. A stochastic probability-of-
success parameter, denoted y, is used, where the parameter value is realized after effort has
been expended. The probability of success depends on exogenous factors that are not
influenced by the firm such as how the market will respond to its efforts, macroeconomic
factors affecting revenues, and demand and competitive scenarios. It is assumed that the
firm, due to market research, managerial analysis, or greater experience, can accurately
forecast the probability of success. However, except in a full information scenario, the
employees do not know the correct value precisely although they know the range y 2 ½y; �yy�in which y lies.
The firm promises a share x of the expected profit to employees as the incentive. Each
employee, thus, expects to receive xyVgðP
e; EÞ=N ; which is the profit shared with
employees divided by the total number of employees. Here, V is an exogenous productivity
factor that links effort to gross profit. One can interpret V as productivity in the best case
scenario (y¼ 1), for example V¼ $40 per worker hour. If y¼ 0.5 the same effort translates
into only half as much profit. Since they are confounded, one can consider yV together as a
random variable, however, as we consider y to be a probability measure, it is necessary to
include V separately as a scaling parameter. The expected profit for the firm is written as:
ð1 � xÞyVgX
e;E� �
�1
2E2; ð1Þ
i.e., the profit retained by the firm less the disutility of effort for the firm.1 To continue, we
require a functional specification of the total effort gðP
e;EÞ: We assume that this is a
linear sum of firm and employee efforts (e.g. Hermalin, 1998):
gX
e;E� �
¼X
e þ lE ð2Þ
The parameter l is the firm’s relative efficiency of effort, which will be referred to as
‘relative efficiency’. It determines whether a unit of effort by management counts more,
less or the same as a unit of employee effort in affecting the outcome. At the extreme when
l¼ 0, managerial activity has no direct effect on the outcome. However, as we shall see,
78 PRASAD AND STEFFES
even when l¼ 0 management may engage in some effort to provide a signal to employees
and thereby indirectly affect the outcome.
3. Full Information Benchmark
We first examine a full information model where both the firm and employees are cor-
rectly informed about the probability of success. The firm announces a sharing rule x
and the employees choose their actions after learning x. Each employee’s problem is to
maximize his or her utility with respect by spending the optimal amount of effort. For
employee i:
eFI ;i ¼ arg max xyVXN
j¼1
ej þ lE
" #,N �
1
2e2
i ð3Þ
From which is obtained the optimal level of effort:
eFI ;i ¼ xyV=N ð4Þ
where subscript FI denotes the full-information scenario. Since employees are considered
to be identical, the subscript i will be omitted in the remainder of the paper. Examining (4)
we see that employee effort is positive and increasing in x, y and V. In each case an
increase in the parameter implies higher rewards for effort. Effort is decreasing in N since
the total incentive is shared with more coworkers resulting in a lower reward for effort.
Anticipating employees’ choice of effort, the firm makes its decisions on x and E by
maximizing expected profit with respect to these variables:
xFI ;EFI 2 arg maxð1 � xÞyV ðxyV þ lEÞ � 12
E2 ð5Þ
Where the profit expression for the firm comes from inserting (4) in (2) and (2) in (1).
Solving the conditions for maximum yields the equilibrium values.
Proposition 1: Under the full information scenario, the sharing rule is xFI ¼ ð1 � l2Þ=
ð2 � l2Þ; the efforts of the employee and firm are respectively, eFI ¼ yV ð1 � l2
Þ=ð2 � l2
ÞN and EFI ¼ yVl=ð2 � l2Þ; and the profit to the firm is PFI ¼ y2
V 2=2ð2 � l2Þ:
When l¼ 0, i.e., when the effort by employees is the only productive effort, the firm
shares half the gains with the employees, otherwise, as long as l< 1, it shares less than
half. The share of profit given to the employees must be substantial enough to induce the
effort necessary to achieve the company goals without excessively sacrificing firm revenue.
It can be shown (see appendix) that when l 5 1,eFI ¼ xFI ¼ 0; EFI ¼ lyV ; and
PFI ¼ l2y2
V 2=2; i.e., the firm finds it best to share no revenue with employees. As a
result, employees contribute no effort, and it is solely the firm’s effort that determines the
outcome. That the firm would ever want to undertake all the effort rather than sharing it
INTERNAL MARKETING AT CONTINENTAL AIRLINES 79
with employees is a surprising result given that the disutility for effort is convex. In
practice, the result implies that incentive schemes should be used only when effort by the
firm is less efficient than effort by employees. For example, in small proprietary
establishments where the owner is also involved in the operations, perhaps such schemes
would be less used. Even for large organizations, for tasks that require significant
management input, such schemes are not useful. For the rest of the paper we restrict
our attention to the domain of interest, l 2 ½0; 1� where these schemes are useful.
4. Asymmetric Information
Let us now examine the case where the firm correctly knows y but employees do not,
believing it instead to be ~yy: Performing the calculations as in the previous case reveals that
the effort choice of the employee is proportional to ~yy rather than y while the profit of the
firm is proportional to y~yy rather than y2, i.e., e ¼ x~yyV=N and P ¼ y~yyV 2=2ð2 � l2Þ:
From the profit expression, the firm always benefits when ~yy is higher. The firm will try
to induce a higher belief among employees, and if they believe the firm, it should lie to the
greatest extent. If it lies, and is believed, the effort level of the employees is higher than
what their optimal effort level should be given the expected reward, which is a function of
y only and not ~yy: Therefore, employees should rationally disbelieve the firm when it makes
an announcement greater than ~yy given that it has every incentive to lie.
An example will clarify this further. Let y 2 ½y; �yy� ½0; 1�: Let ~yy ¼ 0:5: To obtain a
contradiction, let there be a value yy > ~yy; say 0.7, which is the maximum reported value
that is believed. Then all firm types will report 0.7, since no type has reason to report less,
and no type has reason to report more (if they report more they are not believed and
instead they get ~yy ¼ 0:5:) Hence, a report of 0.7 is uninformative about the type and does
not update employees’ prior beliefs. Thus, there can be no yy > ~yy:To conclude, if employees have a prior belief ~yy less than y there is degradation of the
firm’s profit because of the wrong belief. (From the firm’s perspective, the situation where~yy is equal to or larger than y requires no intervention. We therefore concentrate on
y 2 ½~yy; �yy�; and for continuity in exposition refer to this as ½y; �yy�). One solution to this
inefficiency is to use internal marketing.
4.1. The Internal Marketing Solution
Managers can exert increased effort, referred to as internal marketing, to convince
employees about the probability of success. The effort is made prior to employees
choosing their level of effort. In the first stage, the firm announces the employees’
share, x, and exerts effort, E. Employees observe x and E and draw conclusions about ybefore choosing their effort level in the second stage.
It should be noted that employees would expect that some effort, EbaseðyÞ will be spent
by the firm even when y ¼ y; and it is only effort beyond this, DE ¼ E � EbaseðyÞ that has
80 PRASAD AND STEFFES
a signaling effect in indicating y� y ¼ Dy: We use the notation f (DE) to denote the
assessment of Dy given observation of the level of internal marketing DE. Therefore, the
effort chosen by the employees in the second stage will be:
e ¼ ½yþ f ðDEÞ�xV=N ð5Þ
Anticipating this, the firm now has to determine x and DE. Since the revenue is obtained
in the second stage, it is discounted by a discount factor, d:
PIM ¼ ð1 � xÞdyV ½ðyþ f ðDEÞÞxV þ lE� � 12
E2 ð6Þ
The solution is given by the following proposition.
Proposition 2: Under internal marketing, the sharing rule is xIM ¼ ð1 � lffiffiffid
pÞ=
ð2 � lffiffiffid
pÞ; the efforts of the employee, firm, and the internal marketing are, respectively,
eIM ¼ yV ð1 � lffiffiffid
pÞ=ð2 � l
ffiffiffid
pÞN ; EIM ¼ yV
ffiffiffid
p=ð2 � l
ffiffiffid
pÞ; and DE ¼ ðy� yÞ=k; where
k ¼ ð1=2xV Þ
h�lþ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffil2
þ ð4x=dð1 � xÞÞ
q i: The profit of the firm is PIM ¼ y2
V 2d=
2ð2 � lffiffiffid
pÞ2:
Unlike in the previous case, where the sharing rule depended only on l, it depends here
on the discount factor as well. Note that a firm that observes a higher y finds it optimal to
do a greater amount of internal marketing to convince employees that this value is true.
4.2. The Money Burning Solution
There is another solution to the asymmetric information problem obtained by using a
mechanism design approach in which the firm shows that it has an incentive to tell the truth
at all times. The firm credibly communicates the true probability of success to employees
by spending money in an unproductive manner ahead of employee efforts. The amount of
money spent is so high that only a firm seeing a particular state would find it optimal to
spend such an amount. This is known as money burning. The money could be spent on new
buildings, furniture, celebrations, etc. Although the firm spends money unproductively, it is
still in its best interest to do so since it induces employees to work harder.
The firm announces a value of y, say yy; in addition to burning money to the amount yðyyÞthat will lead employees to believe yy is indeed the correct value of y. Thereafter, the
employees and the firm expend effort. It should be noted that if the money burnt is spent
on employees in the form of celebrations, say, it is spent prior to the choice of effort and
not as a reward; hence, it does not directly enter their effort function, which is
e ¼ xyyV=N : ð7Þ
Let Pðyy; yÞ be the expected profit when the real state is y and the announced value is yy:The required incentive compatibility condition for truth telling is:
INTERNAL MARKETING AT CONTINENTAL AIRLINES 81
Pðy; yÞ �Pðyy; yÞ � 0 for all y; yy: ð8Þ
If a firm in state y announces x, yðyyÞ and yy; which employees believe, the profit is:
Pðyy; yÞ ¼ ð1 � xÞyV ½yyxV þ lE� � E2=2 � yðyyÞ ð9Þ
By maximizing profit with respect to x, E and yy; subject to the incentive compatibility
requirement, we get the following proposition:
Proposition 3: Under money burning, the sharing rule is xMB ¼ ð1 � l2Þ=ð2 � l2
Þ; the
efforts of the employee and firm are respectively, eMB ¼ yV ð1 � l2Þ=ð2 � l2
ÞN and
EMB ¼ yVl=ð2 � l2Þ; and the profit to the firm is PMB ¼ ððy2
þ ð1 � l2Þy2
ÞV 2Þ=ð2ð2 � l2
Þ2Þ: The money burnt is yMB ¼ ðy2
� y2ÞV 2ð1 � l2
Þ=2ð2 � l2Þ2:
The sharing rule and the efforts by the firm and the employees do not change from the
full information scenario. It can be seen that PMB þ yMB ¼ PFI : Thus, there is a rather
large loss in profits from the full information scenario due solely to the necessity of
signaling to employees. Observe that when y is close to the actual value y, employees do
not need much convincing, the firm does not need to spend much money in signaling and
profits converge to the full information profit, PFI :
5. Comparison of Solutions
Based on these results, the following proposition follows from direct observation of the
profit expressions and provides a comparison of the profits in the three scenarios.
Proposition 4: Comparing the profits under the three scenarios, we obtain:
1. PFI > PMB except when y ¼ y or l ¼ 1 in which case PFI ¼ PMB.
2. PFI > PIM except when d ¼ 1 and l ¼ 1 in which case PFI ¼ PIM .
3. PIM ¼ PMB if d ¼ 1; y ¼ 0 and l ¼ 0 or l ¼ 1:4. PIM > PMB if
a. y ¼ 0 and d 2 ½1; dÞ where d ¼ 4=ð2 þ l� l2Þ2:
b. d ¼ 1 and y=y 2 ½0;mÞ where m ¼ lð4 � l� l2Þ=½ð1 þ lÞð2 � lÞ2�1=2:
The following points summarize the salient information from the proposition:
The full information scenario profit dominates the profits of the internal marketing and
money burning scenarios indicating that signaling is expensive.
When d ¼ 1; i.e., the first stage does not take much time, and l ¼ 1; PIM ¼
PMB ¼ PFI . When, d ¼ 1 and l ¼ 0; i.e., the firm can contribute no direct productive
effort, and y ¼ 0 the money burning solution generates identical results to the internal
marketing solution in every respect, with the cost of internal marketing to the firm being
82 PRASAD AND STEFFES
of equal value to money burnt. However, if y > 0; i.e. if there is guaranteed positive
probability of rewards, the money burning solution does better than the internal
marketing solution in this case. As l increases from zero, internal marketing begins
to dominate money burning because it provides not only signaling value but also
productive value.
There is a problem with money burning that makes internal marketing a much better
option, ceteris paribus. If management spends a great deal of money unproductively, it is
unclear whether employees will get the correct signal or whether they will make
inferences regarding poor management and become demoralized.
In group activities, the amount of effort put in is less than socially optimal due to free
riding (Holmstrom, 1982). Some authors have examined the effectiveness of mutual
monitoring, peer pressure and group norms in reducing free-riding (Barron and Gjerde,
1997; Kandel and Lazear, 1992). Whereas mutual monitoring may mitigate the effect of
free-riding it is unlikely to eliminate it. In the Continental Airlines example, it is
unrealistic that a group at one airport can monitor a group at another airport or that flight
attendants can monitor baggage handlers; therefore complete mutual monitoring cannot
exist. Perhaps the simplest way to incorporate mutual monitoring into the present model
is by reducing N to N=n where n is the number of employee groupings that can engage
in mutual monitoring. The overall analytical results and intuition from the analysis
remain unchanged. Thus, the incentive scheme should be implemented despite free
riding taking place.
6. Conclusions
Communicating firm-wide objectives and motivating employees to achieve them remains a
challenging task for today’s managers. Continental Airlines overcame this challenge by
designing a successful firm-wide incentive program focusing on timely flight arrival and
departure. It may appear self-evident that on time flights increase customer satisfaction and
airline profitability. However, before the implementation of the profit sharing plan, flight
attendants would routinely hold up a full flight to allow catering to locate a few more meals
just to avoid a small shortage. The result was delay for a plane full of passengers for the
sake of meals that many passengers would have happily exchanged for on time arrival and
a free drink (Bethune and Huler, 1998).
We investigated three scenarios, full information, internal marketing, and money burn-
ing. Although the full information scenario yields the highest profit by saving on signaling
costs, the assumption that employees and management are equally informed about the
probability of success of a scheme is not realistic. Under the two asymmetric information
scenarios, internal marketing and money burning, the firm signals its information to the
employees that their efforts will produce the desired firm-wide objective.
The profit depends on several parameters, such as relative efficiency of effort, the
discount factor, and employees’ belief about the probability of success. Consequently, the
internal marketing solution is not always more profitable than money burning.
INTERNAL MARKETING AT CONTINENTAL AIRLINES 83
Before choosing their communication mechanism, management must understand a
number of key issues. For example, it must obtain a clear understanding of the probability
of success, y. Without this, their objective may fail to be achieved despite optimal
employee effort, causing a loss of trust between management and employees. Second,
management must have an idea of how managerial activity affects productivity through the
relative efficiency parameter. Understanding this parameter will help managers determine
the level of profit sharing that will optimize employee effort. Table 2 compares the model’s
predictions with Continental’s successful implementation.
An area for future research is to examine whether the results would be modified if there
were two types of effort; for example, effort towards improving on-time flight arrivals and
departures and effort towards reducing baggage-handling complaints. At present all types
of effort are considered identical in their cost and their effect on productivity.
Notes
1. Another interpretation of (1) may be given in keeping with the Continental case, that rewards of V accrue if a
certain target is met (such as being above average in on-time flight performance), or 0 if the target is not met.
Under this interpretation, the probability of the target being reached is given by yg(�). It then becomes
necessary to ensure that the probability yg(�) remains bounded in [0, 1], but without taking a non-linear
transformation that would significantly complicate the analysis. We can do this conceptually by dividing yg(�)
by a scalar M to normalize it, and adjusting M into V. The present analysis is unaffected.
Table 2. Comparisons of Model Results to Practice at Continental
Research Issue Model Results
Practice at Continental
(from Bethune and Huler, 1998)
(RI.1) Should an incentive
scheme be used?
As long as employees are relatively
more efficient in determining the
outcome, a profit sharing incentive
scheme is appropriate (Proof of FI
scenario in appendix.).
It was clearly understood that flight
timeliness was dependent on
employee action with little
role for management. A profit
sharing incentive scheme was used.
What signaling method
should be used?
Internal marketing and money
burning yield identical results if
l ¼ 0; and d ¼ 1 as is likely for the
Continental case (Prop. 4).
Internal marketing. The CEO and top
management spent a great deal of
time and effort with employees
to communicate their scheme.
(RI.1) What sharing rule
should be used?
Since l ¼ 0 and d ¼ 1;
the share is x¼ 1=2 (Prop. 2).
x¼ 1=2
(RI.2) How much effort
should be spent on
internal marketing?
DE ¼ ðy� yÞV=2 when l ¼ 0
and d ¼ 1 (Prop. 2).
Not precisely known, but stated as
initially large. In subsequent months,
employees were more confident about
receiving their share (i.e. y higher) and
management spent less effort
propagating the scheme, in line with
the analysis.
84 PRASAD AND STEFFES
Acknowledgements
The authors thank Kissan Joseph, Barton Weitz and an anonymous reviewer for their
helpful suggestions.
References
Barron J. M., and Gjerde, K. P. (1997). ‘‘Peer Pressure in an Agency Relationship,’’ Journal of Labor Economics,
15(2), 234–254.
Basu A. K., Lal, R., Srinivasan, V., and Staelin, R. (1985). ‘‘Salesforce Compensation Plans: An Agency
Theoretic Perspective,’’ Marketing Science, 4(4), 367–291.
Bethune G., and Huler, S. (1998). From Worst to First: Behind the Scenes of Continental’s Remarkable
Comeback—A Flight Plan for Success. New York: John Wiley Sons, Inc.
Brenneman, G. (1998). ‘‘Right Away and all at Once: How We Saved Continental,’’ Harvard Business Review, 76,
162–179.
Chu, W., and Desai, P. S. (1995). ‘‘Channel Coordination Mechanisms for Customer Satisfaction,’’ Marketing
Science, 14(4), 343–359.
Cooper, J., and Cronin, J. J. (2000). ‘‘Internal Marketing: A Competitive Strategy for the Long-Term Care
Industry,’’ Journal of Business Research, 48, 177–181.
Coughlan, A. T., and Sen, S. K. (1989). ‘‘Salesforce Compensation: Theory and Managerial Implications,’’
Marketing Science, 8(4), 324–342.
Cowell, D. (1984). The Marketing of Services. London: Heineman.
George, W. R. (1990). ‘‘Internal Marketing and Organizational Behavior: A Partnership in Developing Customer-
Conscious Employees at Every Level,’’ Journal of Business Research, 20, 63–70.
Hauser, J. R., Simester, D. I., and Wernerfelt, B. (1997). ‘‘Side Payments in Marketing,’’ Marketing Science, 16(3),
246–255.
Heskett, J. L., Jones, T. O., Loveman, G. W., Sasser, E. W., Jr., and Schlesinger, L. A. (1994). ‘‘Putting the
Service-Profit Chain to Work,’’ Harvard Business Review, (March=April), 164–172.
Hermalin, B. E. (1998). ‘‘Towards an Economic Theory of Leadership: Leading by Example,’’ American
Economic Review, 88(5), 1188–1206.
Holmstrom, B. (1982). ‘‘Moral Hazard in Teams,’’ The Bell Journal of Economics, 13 (Autumn), 324–340.
Kandel, E., and Lazear, E. P. (1992). ‘‘Peer Pressure and Partnerships,’’ Journal of Political Economy, 100(4),
801–817.
Kotler, P. (2000). Marketing Management The Millennium Edition. New Jersey: Prentice-Hall.
Knez, M., and Simester, D. I. (2001). ‘‘Firm-Wide Incentives and Mutual Monitoring at Continental Airlines,’’
Journal of Labor Economics, 19(4), 743–772.
Lovelock, C. H. (1983). ‘‘Classifying Services to Gain Strategic Marketing Insights,’’ Journal of Marketing, 47,
9–20.
Pitt, L. F., and Foreman, S. K. (1999). ‘‘Internal Marketing Role in Organizations: A Transaction Cost
Perspective,’’ Journal of Business Research, 44, 25–36.
Schlesinger, L. A., and Heskett, J. L. (1991). ‘‘The Service-Driven Service Company,’’ Harvard Business Review,
(September=October), 71–81.
Appendix
Proof of Full Information Scenario
From (5) the firm’s objective is to maximize with respect to x and E the following profit
function:
INTERNAL MARKETING AT CONTINENTAL AIRLINES 85
ð1 � xÞyV ðxyV þ lEÞ � 12
E2 ða1Þ
The first order condition with respect to E is:
ð1 � xÞyVl� E � 0 c:s: E � 0 ða2Þ
where c.s. stands for the Kuhn-Tucker complementary slackness condition. Since the first
term on the LHS is non-zero, (a2) holds with equality, i.e., E ¼ lyV ð1 � xÞ: The first order
condition for x is:
yV ½ð1 � xÞyV � ðxyV þ lEÞ� � 0 c:s: x � 0 ða3Þ
The second order condition for maxima is that the Hessian be negative definite requiring
ð�1Þð�2y2V 2Þ � ð�lyV Þ
2 > 0; which is satisfied if l2 < 2: Supposing this is so, solving
(a2) and (a3) simultaneously yields:
if l2 > 1 then x ¼ e ¼ 0;E ¼ lyV ; P ¼ l2y2V 2=2: ða4Þ
if l22 ½0; 1� then x ¼
ð1 � l2Þ
ð2 � l2Þ; e ¼
yV ð1 � l2Þ
ð2 � l2ÞN
;
E ¼yVl
ð2 � l2Þ
and P ¼y2
V 2
2ð2 � l2Þ: ða5Þ
Proof of Internal Marketing Scenario
The objective function to be maximized with respect to x and DE is:
PIM ¼ dð1 � xÞyV ½ðyþ f ðDEÞÞxV þ lE� � 12
E2 ðb1Þ
The first order condition for x gives:
x ¼ 12� lE=2ðyþ f ðDEÞÞxV ðb2Þ
The first order condition for DE gives:
dð1 � xÞyV ½ f 0ðDEÞxV þ l� � E ¼ 0 ðb3Þ
This implies:
y ¼ E=dð1 � xÞV ½ f 0ðDEÞxV þ l� ðb4Þ
86 PRASAD AND STEFFES
Given truth telling, the LHS above must be yþ f ðDEÞ: In (b4), replacing y by
yþ f ðDEÞ and E by EbaseðyÞ þ DE results in a differential equation that can be solved
for DE.
yþ f ðDEÞ ¼ ½EbaseðyÞ þ DE�=dð1 � xÞV ½ f 0ðDEÞxV þ l� ðb5Þ
To solve this differential equation, observe that a solution of the following form will work:
f ðDEÞ ¼ kDE ðb6Þ
where, k is a constant to be determined. By substituting (b6) into (b5) the value of k is
found.
k ¼1
2xV�lþ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffil2
þ4x
dð1 � xÞ
s" #ðb7Þ
Thereafter, using (b2), and noting that from (b4) we can express the ratio y=V as a
function of k alone, the solution for x is obtained:
x ¼1
2�
xl
½�lþffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffil2
þ 4x=dð1 � xÞ
q�
ðb8Þ
Equation (b8) can be explicitly solved for x. The following solution is obtained after a
few intermediate steps:
xIM ¼ ð1 � lffiffiffid
pÞ=ð2 � l
ffiffiffid
pÞ ðb9Þ
From (b2) the value of E can be obtained directly:
E ¼ Vyð1 � 2xÞ=l ðb10Þ
and substituting x from (b9) yields
EIM ¼ yVffiffiffid
p=ð2 � l
ffiffiffid
pÞ ðb11Þ
From (b6), noting that f ðDEÞ y� y ¼ kDE we get
DE ¼ ðy� yÞ=k ðb12Þ
From equation (5) that e ¼ xyV=N ; we substitute the value of x to obtain
eIM ¼ yV ð1 � lffiffiffid
pÞ=ð2 � l
ffiffiffid
pÞN ðb13Þ
INTERNAL MARKETING AT CONTINENTAL AIRLINES 87
Finally, second order conditions are verified, and substituting all values into (b1)
PIM ¼ dy2V 2=2ð2 � l
ffiffiffid
pÞ2
ðb14Þ
Proof of Money Burning Scenario
We have from equation (4) that effort e exerted by employees depends on x and yy:
e ¼ xyyV=N ðc1Þ
If, when the real state is y, the firm announces x, yðyyÞ and yy; which is believed, the
expected profit is:
Pðyy; yÞ ¼ ð1 � xÞyV ½yyxV þ lE� � 12
E2 � yðyyÞ ðc2Þ
This is to be maximized by the firm with respect to E, x and yy: We get, as in the proof to
Proposition 1:
E ¼ yVlð1 � xÞ ðc3Þ
x ¼ ð1 � l2Þ=ð2 � l2
Þ ðc4Þ
For yy the following condition is obtained:
ð1 � xÞyVxV � y0ðyyÞ ¼ 0 ðc5Þ
For the claim to be believed, the yy satisfying the first order condition must be set to equal
to y, in other words, y ¼ arg maxyy Pðyy; yÞ: This is the required incentive compatibility
condition for truth telling: Pðy; yÞ �Pðyy; yÞ � 0 for all y; yy: Then condition (c5) is a
differential equation that can be solved to yield the appropriate y(y):
y ¼
Zyy
xð1 � xÞV 2ydy ¼ ðy2� y2
Þxð1 � xÞV 2=2 ðc6Þ
To see if the solution is a maximum we evaluate the Hessian. The terms
@2P=@E2; @2P=@x2; @2P=@yy2; evaluated at their solution points are respectively �1,
�2y2V2 and �ð1 � l2Þ=ð2 � l2
Þ2 which are all negative. We have shown, in the proof of
Proposition 1, @2P=@E2 @2P=@x2 � ð@2P=@E@xÞ2 > 0: The third principal minor can be
88 PRASAD AND STEFFES
calculated to be �y2V 2ð1 � l2
Þ=ð2 � l2Þ; which is unambiguously negative. The required
condition for maxima is therefore satisfied.
The expected profit can now be calculated.
PMB ¼ y2V 2=2ð2 � l2
Þ � ðy2� y2
ÞV 2ð1 � l2Þ=2ð2 � l2
Þ2
) PMB ¼ ½y2þ ð1 � l2
Þy2�V 2=2ð2 � l2
Þ2
ðc7Þ
INTERNAL MARKETING AT CONTINENTAL AIRLINES 89