The relationship between network effects, new product pricing

Page 1 of 8
ANZMAC 2009
The Relationship between Network Effects, New Product Pricing Strategies and Sales
Performance: A Quadratic and Interaction Effect Estimation
Susanna Winter, Lappeenranta University of Technology, [email protected]
Sanna Sundqvist, Lappeenranta University of Technology, [email protected]
Abstract
Taking that new product launch decisions are critical to any company, the present paper
focuses on studying strategic and tactic launch decisions and examines the relationships
between pricing objectives (strategic launch decision), price setting (tactical launch decision),
network effects, and sales performance. We expect that products' network effects challenge
the current wisdom in launching new products. Our results from the analysis employing
structural equation modeling based on the data of 236 new product launch decisions show that
both strategic and tactical launch decisions have an effect on new product sales performance
and that the relationship between pricing and performance depends on the degree of network
effects.
Keywords: price, sales, innovation, network
ANZMAC 2009
The Relationship between Network Effects, New Product Pricing Strategies and Sales
Performance: A Quadratic and Interaction Effect Estimation
Introduction
New product launch decisions involve significant commitments of time, money and resources
(Urban and Hauser, 1993), and are thus critical to firms and managers making these decisions.
Launch decisions may also determine the success or failure of any new product (Hultink et
al., 1997) or even of a firm. Launch of a new product is a set of complex decisions like what
to launch, where to launch, when to launch, why to launch, and how to launch, and includes
both strategic and tactical decisions (Hultink et al., 1997). The focus of our present paper is
on new product pricing strategies and tactics as managers feel that finding the correct match
between pricing and product features is a major source of competitive advantage (Beard and
Easingwood, 1996). Additionally, as Lee and O'Connor (2003a) have noted, existing literature
offers little decision making guidance to managers on how to successfully introduce a product
that exhibits network effects, and thus a special attention will be given to product level
network effects. Network effects refer to a phenomenon in which the value of a product to any
single user increases as more people adopt the same product (Katz and Shapiro, 1985;
Economides, 1991; Lee and O'Connor, 2003a).The present paper focuses on studying the
relationships between pricing objectives (strategic launch decision), price setting (tactical
launch decision), network effects, and sales performance. Additionally, we control for product
and market level factors, such as product type and customer resistance. Besides contributing
to the launch strategy research stream, we also contribute to the networks effects domain by
applying a six-dimensional composite measure for product-level network effects.
Factors Affecting New Product Short-term Sales Performance
In new product performance literature, the relationship between product pricing and
performance has been studied. New product performance is reported to be higher when
relative price is lower (Choffray and Lilien, 1984; Lambkin, 1988). However, in order to link
strategic decisions to pricing, we propose that focus should be also on pricing objectives. It
has been suggested that it is necessary to study both strategic as well as tactical aspects of
product launch as both have an impact on new product performance (Hultink et al., 1997).
Especially for products characterized by network effects, it is critical that pricing strategy
supports the objective of growth in number of adopters so that critical mass will be achieved.
If companies highlight product-level profit margins too early, they may fail to create the
critical mass needed and thus will suffer from poor sales. Thus, we simply hypothesize that
H1: The more pricing objectives stress product-level profit margins (instead of sales
volume and market share), the lower the new product sales performance.
Network effects are expected to change the nature of consumption behavior (Lee and
O'Connor, 2003a). In the early stages of product life cycle, when there are very few adopters,
the value of a network effects product to any user is low. Since the incentives for adoption are
small, consumers prefer to wait to see whether a large enough installed base will emerge.
This, in turn, causes network effects products to be prone to the so-called late take-off
phenomenon (Lim, Choi, and Park, 2003): sales are at a lower level for a longer period of
time compared to non-network effects products (see e.g. Schoder, 2000). Network effects
have been discovered to be a major factor slowing new product growth (Goldenberg, Libai,
Page 2 of 8
Page 3 of 8
ANZMAC 2009
and Muller, 2005). Therefore, we propose that network effects have a negative impact on new
product sales, and present the following hypothesis:
H2: The more there are network externalities embedded in the product, the lower the
short-term new product sales performance.
According to Porter (1985), there are three generic competitive strategies firms may aspire:
cost leadership, differentiation, or focus. In cost leadership, a firm put its emphasis on being
the low-cost producer. Thus, it might be able to squeeze prices down, and yet still have a good
quality image, and therefore can be an above average performer in its industry (Porter, 1985).
On the other hand, firms choosing a differentiation strategy select attributes that many buyers
value, and uniquely position themselves. These firms are rewarded for their uniqueness with a
premium price. Additionally, they can cultivate a winner image and build strong brands,
which further boost their sales. However, firms with no clear competitive strategy may end up
pricing their products in the middle. These companies suffer from poor sales because there is
no compelling reason for customers to buy. Relating our reasoning to the strategy literature,
we assume that the relationship between pricing and sales performance is quadratic.
H3: Pricing will have a positive quadratic (U-shaped) effect on sales performance.
Since the value of a network effects product to any user is dependent on the amount of other
users, the short-term goal for a company launching a new network effects product is to
maximize the installed base rapidly (Lee and O'Connor, 2003a), i.e. reach a critical mass of
users. To reduce potential customers' cost of adoption to meet their lowered value
expectations, the company needs to lower the initial price of the product. Indeed, a commonly
suggested strategy for reaching critical mass is price subsidies or discounts (Witt, 1997; Hill,
1997; Mahler and Rogers, 1999) or even giving the product away free. Penetration pricing can
help increase the growth in customer base (Hill, 1997), and, consequently, constitutes a
powerful tool in accelerating the take-off of new products (Montaguti, Kuester, and
Robertson, 2002). Thus, in the presence of network effects, penetration strategy becomes
critical (Lee and O'Connor, 2003a) because its effectiveness in accelerating new product
growth is increased (Montaguti, Kuester, and Robertson, 2002). Therefore, we suggest that in
the case of high network effects, the impact of low prices on sales performance is heightened.
In other words, network effects positively moderate the quadratic relationship between pricing
and short-term sales. Figure 1 summarizes our conceptual model and hypotheses.
H4: The quadratic (U-shaped) relationship between pricing and sales performance
becomes larger in magnitude as network effects increase.
H1 (-)
Pricing
Objective
Network
Effects
Pricing
Squared
H3 (+)
Figure 1. Conceptual Model and Hypotheses
H4 (+)
H2 (-)
Sales
Performance
ANZMAC 2009
Methodology
The sample was drawn from 1798 companies with more than 50 employees. After prenotification by telephone, 954 companies emerged as being eligible to participate (they had
introduced a new product or service with three years of data collection). In total, 236 usable
questionnaires were obtained, yielding an effective response rate of 24.7% of the qualified
sample. Where possible, the constructs of interest were measured using scales drawn or
adapted from prior studies on five-point Likert-type scales. The two items for both pricing
objective and price level were drawn from Lee and O'Connor (2003b). The 19 items for
network effects were adapted and extended from a survey item used in Lee and O'Connor
(2003b). The five items for sales performance were drawn from Griffin and Page (1996) and
Lee and O'Connor (2003b). Additionally, we controlled for customer resistance and product
type. Mohr (2001) notes that customer perceptions of the cost/benefit of a new technology
affect pricing strategy. Because of customers' fear, uncertainty, and doubt, they may postpone
adoption and thus firms' sales performance is deteriorated. Thus, customers provide a ceiling
above which marketers should not price. The three items for customer resistance were drawn
from Zhou, Yim, and Tse (2005) and Lee and O'Connor (2003b). Furthermore, we controlled
for product type (measured as a categorical variable) as we expected that initial sales levels of
new industrial products are much lower than for consumer products.
Analyses and Results
Measurement items were entered into a confirmatory factor analysis using LISREL 8.50, and
the measurement model fit indexes show an excellent fit (see Table 1). Our network effects
measure is comprised of six first-order factors (direct, indirect, social network, user generated
content, customer expectations, extended product value). In constructing the measure for
network effects, a composite measure was created. As can be seen in Table 1, all average
variances extracted (AVE) exceed the squares of the correlations between latent variables,
providing support for the discriminant validity of the scales. Furthermore, the composite
reliabilities are all large (near 0.80) and well exceed the recommended level of 0.60 (Bagozzi
and Yi, 1988). The average variances extracted are almost all above 0.50, and therefore our
measures have adequate levels of convergent and discriminant validity.
Because H3 argues for a U-shaped relationship between pricing and product sales
performance, we created a quadratic term by squaring the pricing variable. In addition,
because H4 argues that network effects change the form of the quadratic relationship between
pricing and product sales performance, we created a product term by multiplying the pricingsquared by network effects. All direct effects and lower-order interactions were included as
control variables to ensure that the proposed model is parsimonious (Aiken and West, 1991).
Model complexity was reduced by using single observed scores for the variables involved in
the quadratic and multiplicative terms (following suggestions of e.g., Jöreskog and Yang,
1996; Ping, Jr., 1995). After creating the quadratic and multiplicative interaction terms, but
before testing the model, Little, Bovaird and Widaman's (2006) recommended procedure for
orthogonalizing (residual-centering) observed quadratic and interaction terms was followed to
handle possible multicollinearity issues.
Page 4 of 8
Page 5 of 8
ANZMAC 2009
Table 1. Model Fit Measures, Correlation Matrix, and Scale Properties
Model
Measurement model
Model 1 (Constrained Model)
Model 2 (Unconstrained Model)
R2
χ2 (d.f.)
325.67 (307)
63.07 (51)
50.13 (48)
2.
3.
∆χ
χ2 (∆ df) RMSEA CFI
–
0.016 0.991
0.032 0.992
12.94 a (3)
0.014 0.998
4.
5.
6.
NNFI
0.989
0.986
0.996
7.
0.16
0.21
1.
1. Pricing
–
2. Pricing objective
0.29** –
3. Network effects
-0.04 -0.06 –
4. Product type (single-item measure) 0.00
0.05 0.15*
–
**
5. Customer resistance
-0.05 -0.07 0.18
0.24** –
6. Sales performance
0.01
-0.11 0.03
-0.25** -0.20** –
7. Pricing squared
0.00
0.07 0.08
0.07
0.04
0.12
–
Mean
3.34
2.71 2.65
1.64
2.06
3.08
0.00
Standard deviation
0.78
1.01 0.80
0.45
0.88
0.81
0.90
Composite reliability
0.81
0.67 0.74
n.a.
0.80
0.86
n.a.
Average variance extracted
0.69
0.51 0.33
n.a.
0.57
0.61
n.a.
RMSEA = root mean square error of approximation; CFI = Comparative Fit Index; NNFI = NonNormed Fit Index; * p < 0.05; ** p < 0.01 (two-tailed)
a Relative to Model 1 (constrained model), Model 2 shows a significant improvement in Chi-square at 5%.
n.a. Composite reliability and average variance extracted are not meaningful in this case.
In Model 1 (the constrained model), only the main effects are allowed to be estimated freely
while the quadratic and interaction terms are fixed at zero. In Model 2 (the unconstrained
model), the quadratic and interaction terms are freely estimated. As can be seen from Table 1,
the reduction in χ2 from the constrained to the unconstrained model is significant, indicating
that the unconstrained model is a better fit to the data. The unconstrained model returns
excellent fit indexes (see Table 2), and was, thus, used for assessing our hypotheses.
Table 2. Model Path Coefficients and t-values
Hypotheses
H1
H2
H3
H4
Controls
Pricing objective
Network effects
Pricing-squared
Pricing-squared × network effects
Pricing
Pricing × network effects
Customer resistance
Product type
Parameter estimates and t-values
Unstandardized Standardized
t-values
estimates
estimates
-0.017
-0.022 -2.185b
0.111
0.084
0.787
0.234
0.238
2.533 a
0.004
0.109
1.528 c
0.121
0.093
0.969
0.004
0.005
1.459 c
-0.230
-0.159 -1.905 b
-0.224
-0.282 -3.101 a
a: Critical t-value (1%) = 2.326; b: Critical t-value (5%) = 1.645; c: Critical t-value (10%) = 1.282
Discussion
Table 1 shows that the effect of pricing objective on sales performance is negative and
significant (β = -0.022, p < 0.05), supporting H1. The more the company emphasizes per-unit
profits as opposed to market share objectives, the lower the resulting short-term sales will be.
Further, the positive and significant coefficient (β = 0.238, p < 0.01) for pricing-squared
ANZMAC 2009
supports our hypothesis H3, indicating that the relationship between price level and sales
performance is U-shaped: sales are higher at both low and high price points; however, midlevel prices lead to poorest performance. Companies without a clear strategy either way may
find themselves struggling to reach their sales objectives.
Our hypothesis H2 suggested that higher product-level network effects lead to lower sales
performance. Contrary to hypothesized, the coefficient for network effects is positive
although not significant, and thus H2 is not supported in our analysis. Network effects do not
seem to have a direct impact on product sales. Instead, they moderate the relationship between
pricing and sales performance, as hypothesized in H4. The positive and significant coefficient
(β = 0.109, p < 0.10) for the pricing-squared × network effects interaction term indicates that
when product-level network effects are high, the impact of price level on sales performance is
magnified further. Therefore, in the case of network effects, it pays off for companies to
introduce the product with a penetration price strategy even at the expense of (short-term)
profits. The heightened impact of lower prices in inducing initial sales and seeding the
creation of installed base will help the product reach a critical mass of users – something that
is essential for the long-term success of network effects products. Regarding our control
variables, the coefficients for customer resistance and product type were negative and
significant, as expected. The direct (linear) effect of pricing on sales performance is nonsignificant, further confirming their U-shaped relationship. The positive and significant
coefficient for the lower-order interaction (pricing × network effects) provides additional
support for the moderating effect of network effects.
This paper addressed several basic managerial questions, such as what is the relationship
between network effects, firms' pricing objectives, pricing strategies, and short-term product
sales performance. Additionally, we contributed to the academic research on network effects
by (1) offering first rich measure for the multi-dimensional nature of network effects, (2)
examining the impact of network effects on product performance, and (3) studying the
relationship between network effects, pricing, and new product sales.
Limitations and Further Research
This study is not without limitations. However, limitations provide interesting areas for future
studies. First, we should note that product sales performance is not created by just one aspect
of marketing mix but through series of decisions across multiple launch variables (Hultink et
al., 1997), and thus future studies should also include other marketing mix variables. What it
comes to our study design, we used the single informant approach in data collection (which
may create biased results); therefore, the use of multiple respondents is recommended.
Further, our measures are not free of limitations. For example, we created a composite, singleindicator measure for network effects to be able to analyze interaction effects (e.g., Ping, Jr.,
1995). However, based on the AVE score, we propose that in the future, network effect
dimensions should be studied separately. Additionally, in order to enhance the robustness and
generalizability of our results, continued refinement is necessary. Our results should be
validated further by using different samples of respondents, products, or environments.
Page 6 of 8
Page 7 of 8
ANZMAC 2009
List of References
Aiken, L.S., S.G. West, 1991. Multiple Regression: Testing and Interpreting Interactions,
Sage, Newbury Park, CA.
Bagozzi, R.P., Yi, Y., 1988. On the evaluation of structural equation models. Journal of the
Academy of Marketing Science 16 (1), 74-94.
Beard, C., Easingwood, C., 1996. New product launch: Marketing action and launch tactics
for high-technology products. Industrial Marketing Management 25 (2), 87-103.
Choffray, J.M., Lilien, G.L., 1984. Strategies behind the successful industrial product launch.
Business Marketing 69 (11), 82-95.
Economides, N., 1991. Compatibility and the creation of shared networks. In: GuerrinCalvert, M.E., Wildman, S.S. (Eds.), Electronic Services Networks: A Business and Public
Policy Challenge, Praeger Publishing, New York (NY), pp. 39-55.
Goldenberg, J., Libai, B., Muller, E., 2005. The chilling effect of network externalities on new
product growth.
Griffin, A., Page, A.L., 1996. PDMA success measurement project: Recommended measures
for product development success and failure. Journal of Product Innovation Management 13
(6), 478-496.
Hill, C.W.L., 1997. Establishing a standard: Competitive strategy and technological standards
in winner-take-all industries. Academy of Management Executive 11 (2), 7-25.
Hultink, E.J., Griffin, A., Hart, S., Robben, H.S.J., 1997. Industrial new product launch
strategies and product development performance. Journal of Product Innovation Management
14 (4), 243-257.
Jöreskog, K., Yang, F., 1996. Non-linear structural equation models: The Kenny-Judd model
with interaction effects. In: Marcoulides, J., Schumacker, R. (Eds.), Advanced Structural
Equation Modeling, Erlbaum, Hillsdale, NJ, pp. 57-88.
Katz, M.L., Shapiro, C., 1985. Network externalities, competition, and compatibility. The
American Economic Review 75 (3), 424-440.
Lambkin, M., 1988. Order of entry and performance in new markets. Strategic Management
Journal 9 (S1), 127-140.
Lee, Y., O'Connor, G.C., 2003a. New product launch strategy for network effects products.
Journal of the Academy of Marketing Science 31 (3), 241-255.
Lee, Y., O'Connor, G.C., 2003b. The impact of communication strategy on launching new
products: The moderating role of product innovativeness. Journal of Product Innovation
Management 20 (1), 4-21.
ANZMAC 2009
Lim, B.-L., Choi, M., Park, M.-C., 2003. The late take-off phenomenon in the diffusion of
telecommunication services: Network effect and the critical mass. Information Economics
and Policy 15 (4), 537-557.
Little, T.D., Bovaird, J.A., Widaman, K.F., 2006. On the merits of orthogonalizing powered
and product terms: Implications for modeling interactions among latent variables. Structural
Equation Modeling 13 (4), 497-519.
Mahler, A., Rogers, E.M., 1999. The diffusion of interactive communication innovations and
the critical mass: the adoption of telecommunications services by German banks.
Telecommunications Policy 23 (10-11), 719-740.
Mohr, J., 2001. Marketing of High-Technology Products and Innovations, Prentice Hall, New
Jersey.
Montaguti, E., Kuester, S., Robertson, T.S., 2002. Entry strategy for radical product
innovations: A conceptual model and propositional inventory. International Journal of
Research in Marketing 19 (1), 21-42.
Ping, R.A., Jr., 1995. A parsimonious estimating technique for interaction and quadratic latent
variables. Journal of Marketing Research 32 (3), 336-347.
Porter, M.E., 1985. Competitive Advantage, The Free Press, New York.
Schoder, D., 2000. Forecasting the success of telecommunication services in the presence of
network effects. Information Economics and Policy 12 (2), 181-200.
Urban, G.L., J.R. Hauser, 1993. Design and Marketing of New Products, Prentice-Hall,
Englewood Cliffs, NJ.
Witt, U., 1997. 'Lock-in' vs. 'critical masses': Industrial change under network externalities.
International Journal of Industrial Organization 15 (6), 753-773.
Zhou, K.Z., Yim, C.K., Tse, D.K., 2005. The effects of strategic orientations on technologyand market-based breakthrough innovations. Journal of Marketing 69 (2), 42-60.
Page 8 of 8