Postponement Strategies in the Supply Chain – How do the reasons underlying demand uncertainty affect the choice of an appropriate postponement strategy? Final Master’s Thesis Miriam Bartels University of Maastricht Faculty of Economics and Business Master of Science in International Business Supply Chain Management Miriam Bartels Student ID: i342793 Final Thesis Supervisor: Prof. Dr. Martin Wetzels Maastricht, 20 th of august 2010
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Postponement Strategies in the Supply Chain
– How do the reasons underlying demand uncertainty affect the
choice of an appropriate postponement strategy?
Final Master’s Thesis
Miriam Bartels
University of Maastricht Faculty of Economics and Business Master of Science in International Business Supply Chain Management Miriam Bartels Student ID: i342793 Final Thesis Supervisor: Prof. Dr. Martin Wetzels Maastricht, 20th of august 2010
I
Preface
In the first year of my Bachelor studies I read a case study on the postponement concept
implemented at the textile company called Benetton. Back then this concept already intrigued
me and it was in regard to my Master Thesis that I finally took the opportunity to explore this
concept in more detail. I feel privileged that my topic suggestion was supported by my
supervisor Prof. Dr. Martin Wetzels, whom I would like to thank for guiding me through the
complete journey of this Master Thesis.
Furthermore, I would like to thank everyone that supported my research and without whom
this research could not have taken place. In the first place, I would like to thank the three
interview partners who contributed to this thesis by providing me with valuable practical
insights and interesting discussions in reference to real life examples. In addition, I would like
to thank all survey participants for taking the time to fill in the questionnaire. Your
contribution enabled me to conclude my thesis by means of studying empirical data, which is
one of the first attempts in the history of postponement choice theory. Finally, I would like to
thank the Dutch Logistics Institute that supported me in spreading the invitation to answer my
survey. Many thanks! Without you all, this thesis could not have been accomplished!
Finally, I would also like to thank my family and friends whose encouragement, especially in
periods in which my thesis progress did not go smoothly, was a tremendous help. In particular,
I would like to thank my boyfriend Peter for his patience and support. Especially during the
last months, when I had already started with my first job 800 km away, it certainly was a
challenging situation.
Finally, I would like to conclude this preface and wish you a pleasant reading experience.
Hopefully this thesis gives you interesting and valuable insights and inspires you in your
future academic or business efforts!
Miriam Bartels
II
Abstract
Coping with demand uncertainties is one of the greatest challenges that companies are facing
in an ever-changing world. One instrument that supports flexibility in this unstable
environment is postponement. By postponing certain activities of the supply chain to a later
stage, flexibility can be achieved, deferring decision making until more knowledge has been
acquired. A variety of postponement strategies is defined in current literature, among which
the most widely used are price postponement, logistics postponement, production
postponement, purchasing postponement and product development postponement. With such
a broad range of postponement strategies available, the question arises of how to choose the
most adequate postponement strategy for a certain company facing industry specific demand
uncertainties. This is exactly what this thesis investigates. Presently, available literature points
out several postponement indicators, including the level of uncertainty. The research at hand
however takes it one step further by exploring how different reasons of uncertainty actually
influence postponement choice.
For this purpose, a research model connecting reasons of uncertainty and different
postponement strategies has been developed, adapted and verified by means of three
interviews with senior supply chain professionals. Subsequently, the adapted research model
has been tested by conducting an online survey among professionals in a variety of industries.
Finally, the survey data has been analysed by means of PLS Path Modelling and cross-
validated with the help of Canonical Correlation and PLS Regression.
This research concludes that different reasons for demand uncertainty require the application
of different situation-specific postponement strategies. Several hypotheses were confirmed
and additional significant factors identified. The findings are summarized by means of a
postponement decision matrix that combines demand uncertainty reasons with organizational
aims. This tool can lead managers through the analysis and decision making process for
•Increased customer orientation (long-term, e.g. regarding new product development)•Increased innovativeness
Postponement Aims
Performance Impact
Figure 9: Adapted research model
In order to test this construct, data is collected by means of an online survey, which will be
discussed in the following subsection. Subsequently, the analytical tools for testing the
construct as well as the construct validation process will be explained.
4.2. Online Survey
The next step in the research process is the empirical validation of the research model. For
this purpose an online survey was carried out among supply chain practitioners. The
following section will outline the questionnaire development as well as the execution process.
4.2.1. Questionnaire Development
The questionnaire was designed with the intention of collecting information from companies
about their application of different postponement strategies, their aims and the reasons for
demand uncertainty that the company is facing. Prior to asking the respondent to answer a
question regarding postponement strategies, he is offered a general explanation of
postponement and its variants to ensure a common understanding. Subsequently, three
question blocks follow, covering the aspects just mentioned. Additionally, some general
4. Chapter: Methodology
35
questions about company characteristics were included. The complete questionnaire can be
found in Appendix II.
The measurement scales could not be adopted from earlier research, because very little
empirical research on postponement strategies has been conducted. Therefore, similar to
previous research, a 7-point Likert scale was initially used for most questions, however with
the aim of testing its effectiveness throughout the pilot study. A 5-point Likert scaled turned
out to be the preferred measurement scale. About ten supply chain professionals and
professionals from other business professions were asked to fill in the questionnaire and
provide feedback on wording, question difficulty, layout, length of questionnaire, sequence,
spelling, mood upon completion of the questionnaire and its reasons. Seven of the ten
participants indicated a preference for a 5-point Likert scale, without being confronted with
this idea.
4.2.2. Execution of Survey
A web-based version of a survey was chosen for data collection. In this way, also respondents
that were located out of the physical reach of the researcher could be contacted. Furthermore,
the online version is convenient for respondents, because time and location for participation
can be chosen individually, and answer transmission is not time consuming. The
disadvantages of this approach are the lack of personal contact and potential technical
problems, leading to lower response rates due to incomplete questionnaires. For the
construction of the questionnaire the software program NetQuestionnaire was used. The
questionnaire was administered in English as well as in German, depending on the
respondent’s preference.
The survey was directed to supply chain professionals that have a general understanding of
the supply chain structure of their companies. Optimally, these contact persons were Supply
Chain Managers, but also other professionals that are for example working in the field of
materials management proved to be able to give very valuable input. Potential participants
were contacted by a variety of methods, including direct e-mail, forwarding through personal
contacts, forum discussions on Xing.com and LinkedIn.com, direct contacting through
xing.com, and through cooperation with the Dutch Logistics Association “Vereniging
Logistiek Management”. In total 57 professionals completed the online survey and 53
responses remained valuable after data cleansing. The majority is active in the industry
4. Chapter: Methodology
36
sectors technology, consumer goods, fashion and chemicals. An overview of the sample
characteristics will be provided in section 5.1.
The following subsection will describe the statistical tools, as well as validation of the
research construct that is conducted by means of the questionnaire data.
4.3. Analytical Tools
Partial Least Squares Path Modelling (PLS PM) is used to analyze the research model. The
findings are then cross-validated by means of Canonical Correlation and PLS Regression.
These three statistical approaches and their relationship will be explained briefly in the
following paragraphs followed by a short elaboration on the reasoning of the final choice of
methods.
Canonical Correlation
Horst’s Generalized Canonical Correlation Analysis can investigate the relationship between
two sets of variables (Thompson 1984). Thereby, the first link of the research model, as it was
depicted in Figure 9, can be analysed, that is the relationship between the set of reasons for
demand uncertainty and the set of different postponement strategies. Canonical roots can be
identified that explain an underlying latent variable. Horst’s Generalized Canonical
Correlation Analysis is based on several assumptions. In the first place the distribution of the
data is assumed to be multivariate normal. For larger sample sizes canonical correlation
however proved to be robust against other distributions. For small sample sizes, it is important
to note, that canonical correlations may only be detected, if they are very strong (> 0.7). That
means that in case of a small sample size a low significance may be detected although the link
may actually be significant (Guinot C., Latreille et al. 2001). Furthermore, outliers should be
detected in the data set because they can seriously influence canonical correlations. Finally,
the data set should not contain any redundant variables. Due to these limitations, canonical
correlations will only be used for cross-validation purposes. Instead PLS Path Modelling can
be used to emulate canonical correlation and to avoid these hard assumptions.
PLS Path Modelling
PLS Path Modelling is a specific type of Structural Equation Modelling (SEM), which is less
subject to constraints than Horst’s generalized canonical correlation analysis. In fact,
4. Chapter: Methodology
37
canonical correlation is a special case of PLS PM, as the stationary equations of Horst’s
generalized canonical correlation can also be found in the estimation of the path model
(Tenenhaus, Vinzi et al. 2005). Recently this statistical analysis became more popular among
operations management researchers. It is used for this research, because it can be employed as
a diagnostic tool that allows the simultaneous analysis of several causal relationships of
manifest (MVs) and latent variables (LVs) (Chin 1998). PLS PM is a component-based SEM
method, in contrast to a standard covariance-based SEM approach (Wetzels, Odekerken-
Schröder et al. 2009). The aim of this method is to minimize the residual variance of
dependent variables. PLS Path Modelling is adequate for the research at hand due to its
limited assumptions. It does not assume normal distribution, requires only a small data sample
and has minimal demands on measurement scale. However, according to Hsu, Chen et al.
(2006) it should be kept in mind that a small sample size leads to a downward bias in form of
underestimation of path coefficients. Furthermore, this method can not only be used for theory
verification, but also for theory development and can easily handle formative constructs
(Wetzels, Odekerken-Schröder et al. 2009).
The model specifications can be split into an outer and an inner model. The outer model, also
called the measurement model, describes the relationship between the MVs to their LVs
(Tenenhaus, Vinzi et al. 2005). Within this research it relates for instance the reasons of
uncertainty to a certain postponement strategy, which is a link that was illustrated in Figure 9.
The inner model, also called the structural model, describes the relationship among the LVs
(Tenenhaus, Vinzi et al. 2005). Furthermore, formative indicators are chosen, in contrast to
reflective indicators, because they reveal a causal link from the indicators to the latent
variable (Tenenhaus, Vinzi et al. 2005). This is in line with the research aim of this paper,
which strives to investigate, if potential reasons for demand uncertainty trigger the use of a
specific postponement strategy. To estimate the inner and outer parameters of the model, the
software SmartPLS is used, that can conduct PLS path modelling with a centroid weighting
scheme. Estimating LVs with formative indicators and a centroid weighting scheme actually
copies the stationary conditions of Horst’s generalized canonical correlation (Guinot C.,
Latreille et al. 2001). Finally, the bootstrapping procedure is used to estimate the significance
of the PLS estimates (Chin 1998).
4. Chapter: Methodology
38
PLS Regression
PLS Regression focuses on the estimation of the outer weights of formative models. In PLS
Path Modelling the outer weights are computed by a system of multiple and simple
regressions. The problem here is that formative blocks often violate the independence
hypothesis of the classical multiple and simple regressions, which leads to increased variance
of estimation, insignificant regression coefficients and not interpretable weights. Therefore
PLS Regression was developed as a new multivariate method of linear multiple regression for
formative models and is thus also applicable for the research model of this study (Esposito
Vinzi, Russolillo et al. 2009). PLS regression can handle multicollinearity which occurs when
components are not independent. It proved to be a powerful diagnostic tool detecting those
latent dimensions underlying a block that are useful for explaining LV (Tenenhaus 2005).
Finally, PLS regression preserves the asymmetry of the model opposed to canonical
correlation. As the aim of this research is to analyse the causal relationship between
predetermined variables, PLS Path Modelling is the preferred approach. PLS Regression will
however be valuable in cross-validating the PLS PM findings for a specific model link.
Thus, there are several arguments for using PLS PM as main statistical tool for the research at
hand. In the first place, PLS PM provides the researcher with greater flexibility for the
interaction between theory and data (Chin 1998). Hypothesized theoretical relationships,
including unobservable latent variables, can be modelled and then tested for significance,
whereas canonical correlation and PLS Regression do not allow for pre-specification of
relations. Canonical correlation cannot determine statistical significance of weights and
loadings resulting in a subjective and thus arbitrary cut-off value (Bagozzi, Fornell et al.
1981). However, estimating the LVs by employing mode B and the centroid scheme in PLS
PM, recovers the stationary condition of Horst’s generalized canonical correlation (Tenenhaus,
Vinzi et al. 2005). Furthermore, chains of causal relations instead of pairs of variable groups
can be analyzed taking causal effects into account. Last but not least, due to a small sample
size and hard assumptions, canonical correlations may prove to be insignificant although in
fact a significant relation exists, thereby not unrevealing all significant links of the model
(Guinot C., Latreille et al. 2001). In conclusion, PLS PM should be used as the primary
statistical tool, and canonical correlation and PLS regression will be employed for cross-
validation purposes.
4. Chapter: Methodology
39
4.4. Construct Validation
In this section construct validation will be performed. Since the model at hand includes
formative indicators, the guidelines for formative model assessment by Petter et al. (2007)
will be followed. It follows that construct validity and reliability will be examined. It is not
necessary to check for unidimensionality, as this only needs to be examined in case of
reflective models, because formative models are by definition multidimensional.
4.4.1.1.Validity
As we are analysing a formative model, the general validity tests, such as Average Variance
Extracted (AVE) and Goodness-of-Fit (GoF) that are applied on reflective models cannot be
applied. This is due to the fact that the indicators do not have to be correlated, because they
are giving rise instead of influencing the LVs (Hulland 1999). As an alternative a Principal
Component Analysis (PCA) could be conducted (Petter, Straub et al. 2007). However, in
order to preserve content validity a PCA is not conducted. This is due to the fact that all items
will be kept in the model to assure not to omit a part of the construct. A model with causal
indicators requires a “census of indicators, not a sample” (Bollen and Lennox 1991, p.308).
Additionally, clear theoretically backed judgement lead to a list of indicators that is assumed
to be complete. This was achieved by means of literature review, interviews and forum
discussions. Hence, it is assumed that all indicators contribute significantly to scale reliability
and content validity.
4.4.1.2. Reliability
The uncertainty and aims are examined for multicollinearity by means of the variance
inflation factor (VIF). The corresponding tables are provided in Appendix III. Severe
multicollinearity is said to exist, when VIF > 10 (Kutner, Nachtsheim et al. 2004). None of
the uncertainty factors and aims proved to suffer significantly from multicollinearity (VIF <
10) and therefore none was omitted from the subsequent analysis. Note, however, that
uncertainty factors C1 and D 2 , aims B 3 and C 4 and aims A 5 and D 6 indicate moderate
collinearity.
1 Uncertainty factor C = Changing customer preferences (product line) 2 Uncertainty factor D = Changing customer preferences (new product) 3 Achievement factor B = Lead time reduction 4 Achievement factor C = Inventory reduction
4. Chapter: Methodology
40
4.5. Conclusion
The initial research model has been verified as well as rounded up by means of expert
interviews. To test the resulting adapted model, an online questionnaire was developed and
executed. Finally, 53 responses resulted to be valuable for the subsequent statistical analysis.
PLS Path Modelling is used for the main analysis and Canonical Correlation as well as PLS
Regression as cross-validation. For the latter two approaches it is important to keep their
constraints and limitations in mind, while interpreting the results. The following chapter will
present the findings of the statistical analysis.
5 Achievement factor A = Overall cost reduction 6 Achievement factor D = Product price optimization
5. Chapter: Findings
41
5. Findings
This chapter will analyse the data collected by means of the online survey. In the first place,
general findings regarding postponement strategies and sample characteristics will be
presented. Thereafter, the findings of the PLS PM analysis of the research model as well as
the cross-validation by means of Canonical Correlation as well as PLS Regression will be
presented. Finally, a conclusion to this chapter is provided.
5.1. General Sample Characteristics
The data collected by means of the online questionnaire indicates a frequent application of all
five postponement strategies among the 53 respondents. Logistics, production and purchasing
postponement proved to be applied most frequently, namely by 47, 45, and 46 respondents
respectively as can be observed in Figure 10 below.
05
101520253035404550
pric
epo
stpo
nem
ent
logi
stic
spo
stpo
nem
ent
prod
uctio
npo
stpo
nem
ent
purc
hasi
ngpo
stpo
nem
ent
prod
uct
deve
lopm
ent
post
pone
mnt
number of respondents
Figure 10: Number of respondents per postponement strategy
Furthermore, logistics, production and purchasing postponement are not only most frequently
applied but also receive higher importance and are more extensively applied (Figure 11).
Price and product development postponement show rather little or moderate importance for
the respondents.
5. Chapter: Findings
42
0
5
10
15
20
25
not applied littleimportance
moderatelyimportant
important veryimportant
importance of postponement strategy / extent of application
number of indications
price
logistics
production
puchasing
prod.dev.
Figure 11: Importance/extent of application within respondents’ companies
Referring to the length of the history of postponement strategy application, it can be stated
that price and logistics postponement have been frequently applied for more than ten years in
various companies, whereas production and purchasing postponement are on average used for
about 5 years. Product development postponement, on the other hand, indicates to be a rather
recent phenomenon as the majority of respondents indicates an application of less than 3 years.
The corresponding Figure is provided in appendix IV.
The impact of postponement on company performance was generally rated as positive to very
positive as can be observed in appendix IV. Only in very few instances, for product
development postponement and purchasing postponement, a negative influence was indicated
by two to three respondents. In case of product development postponement this can be due to
a negative impact on financial performance, even though a positive effect on innovation may
have been achieved. In case of purchasing postponement a negative impact may be due to the
risk of price variations.
The respondents are originating from different industries as shown in Figure 12. 22.2% of the
respondents are working in the technology industry, followed by 18.5 % in consumer goods
and 16.7 % in fashion as well as chemicals. There are eight respondents that fall into the
category “other”. Among these are four companies that could be classified as technology
companies and one as fashion company. In addition, two other companies indicate to produce
furniture and agricultural products, respectively.
5. Chapter: Findings
43
0
2
4
6
8
10
12
14
cons
umer
good
s
techn
ology
fashio
n
chem
icals
autom
otive
pharm
aceu
ticals
other
# of respondents
Figure 12: Distribution of survey respondents across industries
Moreover, 33% of the respondents indicated to operate internationally, whereas 61% operate
globally. The head quarter of 80% of the companies is located in Europe, 16% in America,
and 2 % in Asia. Only slight difference in postponement application could be detected among
European and North American companies. The p-values of the independent t-tests between
origin and logistics postponement, as shown in appendix IV, indicated that the means are
significantly different, namely significantly more important in the US compared to Europe.
This may be due to the longer distances in North America that require more frequent direct
deliveries compared to, for instance, milk runs, which are round trips by truck. However, the
findings should be interpreted with caution due to the small sample size of American firms. A
comparison to Asian companies could not be established, because the sample size was too
small. Logistics postponement also varies in importance between internationally and globally
active firms. At a 10% significant level international firms indicated a significantly lower
average importance of logistics postponement.
The relation between postponement strategies and level of uncertainty does not show stronger
postponement application in the presence of high uncertainty. The following PLS path
modelling analysis will therefore investigate, if the reason of demand uncertainty has a
decisive influence on the choice of postponement application.
5. Chapter: Findings
44
5.2. The Research Model
The following subsections will report the findings in the sequence of links established above:
(1) PLS PM for Link 1: Uncertainty – Postponement Strategies, (2) PLS PM for Link 2 and 3
Aims/Impact, (3) Cross-validation by means of Canonical Correlation and PLS Regression.
The main focus will be on the results of PLS PM, because it proved to be the most robust
measure. For each postponement strategy, one PLS PM model was created in the software
program SmartPLS. Figure 13 shows the PLS PM model for price postponement as an
example. Each model includes four LVs, being demand uncertainty, the postponement
strategy, aim as well as the performance impact, that are in turn influenced by various
formative indicators (MVs).
Figure 13: PLS PM Model for Price Postponement
In order to understand the data structure behind the model, the construction of each latent
variable will now be explained.
The latent variable uncertainty is influenced by ten formative indicators (MVs). These are the
reasons underlying demand uncertainty. Data regarding these indicators was collected by
asking the respondents to specify the importance of each of the ten indicators in influencing
the demand uncertainty their company is facing.
The latent variable demand uncertainty is hypothesized to influence a certain postponement
strategy; in the example above this is price postponement (Figure 13). Price postponement
has one formative MV that is determined by asking the survey respondents to specify the
importance/extent of application of price postponement at their company.
Subsequently, price postponement leads to a certain aim of price postponement, which is
illustrated by the third LV. Eight formative postponement strategy specific indicators
5. Chapter: Findings
45
influence the aim of price postponement. Each indicator represents a possible aim of
postponement strategies. Data for these indicators was collected by asking the survey
respondents to determine the importance of each of the eight possible aims, when applying
price postponement.
Finally, the aim of price postponement is modelled to have an impact on company
performance, which is represented by the fourth LV. The impact on company performance is
influenced by one formative indicator. Data regarding this indicator was collected by asking
the survey respondents to rate the impact of price postponement on company performance.
In summary, data for the indicators of the first LV (demand uncertainty) is collected by means
of general questions, whereas the data for the indicators of the other three LVs is collected by
means of postponement strategy specific questions. Finally, the PLS algorithm is run in
SmartPLS for each of the five postponement strategy models, employing the centroid
weighting scheme, as every factor should be given the same weight (Vinzi, Chin et al. 2010).
The bootstrapping procedure subsequently provides an insight into the significance of outer
weights and path coefficients by calculating t-statistics. A summary of the findings of the PLS
PM analyses are provided in table 3, including the weights, path coefficients between LVs, t-
statistics as well as p-values. The findings will briefly be explained throughout the following
subsections. A more detailed discussion is provided in chapter 6.
Reasons underlying demand uncertainty = positive significance expected A - price fluctuations (compet. Product) = significant at 5% B - price fluctuations (material/component) = slightly above 5% C - chang. customer preferences (pr. line) = - signifiance D - chang. customer preferences (new prod.) = + significance E - irregular purchases
F - innovationG - change in market shareH - many different customer groupsI - high seasonalityJ - weather
PLS PM
Table 4: Overview of cross-validation
5. Chapter: Findings
51
5.4. Conclusion
This chapter provided the findings of the empirical research which is investigating the
relationship between demand uncertainty and postponement strategies as well as the influence
on postponement aims and the impact on company performance. From PLS PM analysis it
can be concluded that only five of the eleven hypothesis regarding demand uncertainty
reasons were confirmed and seven additional relations were identified. Cross-validation by
means of canonical correlation and PLS Regression confirmed the findings from PLS PM to a
great extent. The following chapter will discuss the findings in more detail, including
discrepancies between expectations and findings. The discussion chapter will conclude with
the development of a decision matrix for postponement strategy choice.
6. Chapter: Discussion
52
6. Discussion
In the following sections each postponement strategy will be discussed by interpreting the
findings of the statistical analysis. Hereafter, the findings will be summarized by means of a
postponement decision matrix and recommendations for its application will be presented.
6.1. Postponement Strategies
The five postponement strategies will be discussed along the upwards direction of the supply
chain including all three model links.
6.1.1. Price Postponement
Uncertainties
Price postponement was expected to be positively influenced by demand uncertainty resulting
from price fluctuations of competitive products (H1.1). However, the findings in this thesis do
not confirm this hypothesis. A possible explanation for this outcome is that frequent price
adaptations, apart from promotional campaigns, do generally not occur. For this reason, price
postponement is not applied to existing products, but rather applied to new products that are
introduced into the market. The researcher however expects this factor to be significant for
service firms, which were not included in this study. By setting the price of a service, which
will take place at a fixed future time, as late as possible, revenue can be maximized. For
instance, in the airline industry, ticket prices are adapted according to demand development
throughout the time period prior to flight. Thereby, a low price can initially foster demand and
a high price can finally take advantage of the urgent need for this service. Revenue or demand
management is therefore a special case of price postponement and should be covered in future
research.
Furthermore, price postponement is found to be positively influenced by a high degree of
innovativeness within the industry at hand. In particular, final price determination in this kind
of situation turns out to be difficult. The reason is that market and competitor behaviour can
only be anticipated to a limited extent due to short life-cycles and a variety of customer
preferences. Hence, the pricing decision is postponed as much as possible to base it on the
most complete set of market information available.
On the contrary, the factor variety of different customer groups is and should be negatively
related to price postponement. Price discrimination may prove to be too complex and may
6. Chapter: Discussion
53
even be legally forbidden. Evidently, a teenager may not be charged a different price than a
senior. Due to different demand characteristics of various customer groups, a unique view of
demand can not be established, therefore not allowing price postponement to deliver benefits
when compared to prior to production price setting. However, price differentiation according
to a few very specific customer characteristics is allowed and even beneficial. This is as
mentioned above conducted by demand management that discriminates prices for example
according to early and late bookers.
Moreover, high seasonality is also negatively related to price postponement. In the presence
of high seasonality, prices rise and decrease according to season specific conditions. In
particular, the degree of demand uncertainty as well as customers’ price sensitivity is reduced
thanks to seasonality. Thus, as these price fluctuations can be forecasted relatively precisely,
price postponement is not necessary and does not lead to great benefits. The cost-benefit
analysis may well show that the costs are higher than the benefits associated with price
postponement in this case.
Aims
Moving one step forward in the causal link, the analysis in chapter 5 indicates that companies
aim to achieve product price optimization. Final price determination is postponed or is subject
to future adaptations in order to benefit from better market information leading to an
optimization of product pricing that maximizes revenue. Again not that in the service industry,
which was not subject to this study, continuous price adaptation prior to the service, referred
to as demand management, may also be seen as price postponement that maximizes revenues,
a potential additional aim.
Performance Impact
Even though the explanations stated above may sound logical, it was discovered that price
postponement actually had a negative impact on company performance. Although it was
assumed that price postponement would lead to price optimization and should have lead to
revenue maximization, the model showed that this was not the case. A possible explanation is
that price postponement could lead to unclear price communication to the customer, which
could result in confusion, irritation and annoyance on the side of the customer. Consequently,
customer satisfaction is reduced and the product buying decision could be postponed or
suspended.
6. Chapter: Discussion
54
6.1.2. Logistics Postponement
Uncertainties
Logistics Postponement was expected to be positively related to irregularity of purchases
(H2.2b). The results published in this thesis confirm this hypothesis. It was argued that in case
of short lead times being very important, direct shipments from a central distribution centre
could be beneficial in situations of irregular purchases. Nonetheless, it is important to take
into account that stocking all items at a distribution centre, while demand is highly variable,
will lead to a high level of space, handling and thus cost requirements. Obviously, customers
always prefer shorter lead times, but the actual benefits of this have to be precisely analysed
and compared to higher space and distribution costs to avoid unnecessary costly stock. In the
situation of irregular purchases, logistics postponement should therefore only be applied to
items for which short lead times are essential.
Additional uncertainty factors were also found to be significant. First of all, price fluctuations
of materials or components show to positively influence logistics postponement. It has to be
noted here that this result may well have been reached due to logistics postponement being
confused with purchasing postponement. Despite early purchases, the delivery of purchased
material may have been postponement instead of the final good. Therefore, this cannot be
considered as logistics postponement.
Thirdly, a changing market share also has a significantly positive effect on logistics
postponement. Despite this surprising outcome, a possible explanation is that in this situation,
product lines should be pooled in one distribution centre to offset the demand variability of
different products. As such, variability in stock requirements and hence warehouse space can
be balanced out. As a matter of fact, changes in market share could be translated into demand
variability across and between product lines. Consequently, the pooling effect of a distribution
centre also applies in this situation.
Finally, weather conditions are negatively related to logistics postponement. Evidently, this
result refers to products whose demand is sensitive to weather conditions. For instance, in
cold summers, sunscreen experiences a much lower demand than in warm summers. Hence,
high levels of inventory at distribution centres should be avoided, if weather conditions have a
strong influence on actual demand.
Aims
Observing the results of logistics postponement, the hypothesis with respect to lead time
reduction was confirmed. Naturally, direct deliveries from stock ensure the shortest possible
6. Chapter: Discussion
55
lead time. Furthermore, the aim of increased customer orientation in the long-run was found
to be significantly negatively related to logistics postponement. In fact, logistics
postponement rather focuses on short-run customer orientation by being flexible in quickly
fulfilling certain needs as opposed to a long-run customer orientation, where customer
preferences are taken into account with regard to new product development.
Performance Impact
The performance impact of logistics postponement is found to be significantly positive. In
particular, the higher the level of centralization of inventory, the higher are the pooling effects
and consequently the higher are also the associated cost savings. Furthermore, short lead
times will increase customer satisfaction and may thus generate more demand. Nevertheless,
it should not be neglected that logistics cost may increase significantly due to costly direct
shipments instead of transportation bundling.
6.1.3. Production Postponement
Uncertainties
Production Postponement was expected to be positively related to changing customer
preferences across product lines, irregular purchases, various different customer groups and
weather conditions. Apart from the factor irregular purchases, all factors have proven to be
significantly related to production postponement; the latter however shows the opposite sign.
Changing customer preferences across product lines (H2.1a) is indeed positively related to
production postponement. The reasoning is such that product modules or differentiations are
applied or added upon customer request and thus only the basic products need to be on stock.
Evidently, if demand varies a lot between differentiations within a product line, production
postponement can react to this flexibly with only limited amounts of stock levels.
The presence of various different customer groups (H4.1) also proves to have a significantly
positive effect on production postponement. These customer groups are characterised by
different customer characteristics and thus a segmented market. Production postponement
does not only achieve flexibility for local demand differences across product lines; but also
addresses geographic differences as well as variation between product lines. As a matter of
fact, this positive effect can be amplified by centralizing production at a distribution centre.
Hence, the benefits of logistics postponement and production postponement can be combined.
The mutual effect could be further investigated by future research.
6. Chapter: Discussion
56
Weather conditions (H5.1) are significantly negatively related to production postponement.
This is surprising as it was expected that uncertain weather conditions and the resulting
variable demand would require production postponement. Nonetheless, this negative relation
may be explained due to the fact that in this situation a demand decrease or increase requires
capacity flexibility and not production flexibility. In fact, demand in this situation is not
shifting, but increasing or decreasing, which renders production postponement as ineffective.
A second explanation may be that unstable weather conditions may influence materials and
components needed for product differentiations, which increases safety stock requirements.
As low stock levels should be maintained, while materials availability must be high,
production postponement may result to be inappropriate.
The forth hypothesis regarding irregular purchases (H2.2a) could not be confirmed. Similar
to the reasoning for the influence of weather conditions, capacity flexibility and not
production flexibility is required in this case. The reason is that irregular purchases lead to
demand swings. Consequently, capacity has to be flexible regarding output levels. Combined
with inventory methods based on pull principles, such as Kanban systems7, overproduction
can be avoided. Especially in cases of little product modularity or differentiations, production
postponement cannot cope with irregularity of purchases.
Aims
Increased customer responsiveness in the short-term is confirmed to be significant. The
reason is that customer preferences regarding product modules and differentiations can
thereby be satisfied quickly. The other two hypotheses relating inventory reduction and
overall cost reduction to production postponement were not confirmed. This is surprising;
nevertheless the reasoning could be explained as follows. A significant inventory reduction
may not be achieved, if the amount of differentiation or the modularity per product is
relatively low. The reason for this is that in such a case the benefit from merely stocking base
products instead of finished product is limited. Moreover, overall cost reduction may not be
accomplished as the production set-up needs to be flexible and their cost might offset a
potential inventory reduction.
Two further aims were found to be significant: product price optimization (positive relation)
and the benefit from decreasing material prices (negative relation). Product price optimization
7 A Kanban system is a materials management system, which “pulls” replenishment, if a certain inventory level is reached. This system is often managed by two buckets, where an empty bucket is handed over to replenishment, as soon as it is empty.
6. Chapter: Discussion
57
can be achieved by adapting the product modules to the customer that will achieve optimal
cost-benefit for the customer. The automotive industry exemplifies this observation by taking
pricing decision upon defining the modularity of a car during customer contact. In fact,
modules, price and customer satisfaction is balanced to reach the optimal product price.
Furthermore, benefiting from decreasing material prices is not an aim of production
postponement. Production postponement focuses on flexibly meeting customers’ needs and
not on increasing the cost structure of production. Decreasing material prices should instead
be taken into account throughout the purchasing process.
Performance Impact
With regard to company performance it could be confirmed that production postponement has
a positive impact. In particular, increased customer satisfaction due to increased
responsiveness as well as inventory reduction in case of high product modularity, improve
company performance.
6.1.4. Purchasing Postponement
Uncertainties
Purchasing postponement was expected to be positively related to price fluctuations of
materials, changes in market share and seasonality. Only price fluctuations of materials
(H1.2) turned out to be significant, however with a negative relationship. For example, a
harvest of wheat depends on weather conditions and determines the available quantity of
supply and sets price. Therefore, purchasing of wheat is expected to be postponed until a
forecast on the harvest can be made and therefore on the price that it can be bought for. On
the contrary, explaining the resulting negative relationship, it can be argued, that an early
purchase with direct full delivery or later delivery arrangements are submitted to hedge
against potential price fluctuations. Evidently, this constitutes rather an advancement of
purchasing instead of postponement.
Frequent changes in market share (H4.2) were assumed to lead to purchasing postponement
to assure flexibility in case of unstable market share division. This relation was not confirmed
by the findings in this thesis. One could assume that not many companies go through quick
market share shifts. In addition, the market introduction of new competitive products should
usually be managed by market studies and forecasts, good product positioning and fast
reaction by product differentiation, which renders purchasing postponement to have a minor
importance in this situation.
6. Chapter: Discussion
58
Furthermore, high seasonality (H4.3) did also not prove to be significant. As a matter of fact,
in this situation forecasts usually are relatively precise due to the fact that high season timing
is known and planning can account for seasonality in advance. Hence, the use of purchasing
postponement realizes no incremental benefits.
A factor that may render the forecast of a season as imprecise is unforeseen weather
conditions, which is another indicator found to be significant by this research. In case that the
demand for certain products is influenced by weather conditions, such as for ice cream, the
purchase of materials (ingredients) may be postponed to the forecasted season start.
Aims
The expected aims, namely overall cost reduction and benefit from decreasing materials
prices, are not found to be significant in the case of purchasing postponement. The aim of
overall cost reduction may be rather substituted by risk or loss avoidance. Traceable cost
reductions are not achieved. Moreover, the benefit from decreasing material prices seems to
be of minor importance. A possible explanation is that the benefit of purchasing
postponement rather results from shorter inventory ownership than from the benefit from
decreasing material prices. However, two other aims were found to be significant in the
executed research instead. Inventory reduction is another significant aim. Due to shorter
material ownership, inventory levels and thus costs can be decreased. One way of maintaining
low inventory levels is by managing close supplier cooperation such as by means of VMI.
Materials are thereby only ordered, when needed which is continuously controlled by the
supplier. It follows that no bulk purchases are required that cover a certain demand horizon.
Finally, increased customer orientation in the long-run proves to be significant. If purchasing
of new components during new product development is postponed, more accurate
specifications can be taken into account, when developing a new product. Similarly, if the
purchase of materials for product differentiation can be postponed, new and more economic
materials, or materials of higher quality, could be used. This can in turn lead to continuous
improvement and increased customer satisfaction.
Performance Impact
Finally, the impact of aims and thus achievements of purchasing postponement on company
performance is only significant at a significance level of slightly above 5%. The impact is
furthermore negative. It can be argued that, as mentioned before, purchasing postponement is
frequently used to mitigate risk instead of proactively reducing cost. Furthermore, purchasing
6. Chapter: Discussion
59
postponement is only indirectly influencing customer satisfaction, and thus the benefit is
difficult to measure.
6.1.5. Product Development Postponement
Demand Uncertainties
Product Development Postponement was expected to be positively related to changing
customer preference in the long term and little market information in innovative industries.
Both hypotheses were however not confirmed by the findings of this thesis. This is surprising;
however potential explanations could be as follows. Changing customer preferences may not
demand for product development postponement, because new market information gained
through postponement may already indicate another new potential product. In innovative
industries with little available market knowledge, product development of any kind should not
be postponed, due to the extreme importance of the first-mover advantage. Furthermore,
innovations are completely new products, something that a customer could not really imagine
yet. It follows that most information necessary for new innovations does not necessarily come
from the target market but rather from the ideas of developers or from other markets.
Finally, irregularity in purchases is found to be significantly negatively related to product
development postponement. This result should be interpreted with caution. Irregular
purchases can only occur, when a product already exists and an adaptation or new module is
developed. In this situation, it could be argued, that product development should not be
postponed. The reason is that adapting the product as soon as possible and thereby positioning
it better in the market may smoothen demand. On the other hand, the negative outcome can
mean that the bullwhip effect created by irregular purchases should be taken care of by using
other and potentially more effective methods.
Aims
As expected a significant aim of product development postponement is increased
innovativeness. In particular, the product development process is optimized since the project
subtasks are organized in a way that takes as much market information into account as
possible. Thereby, the quality of new product development is maximized; hence innovation is
accelerated.
6. Chapter: Discussion
60
Performance Impact
The overall performance impact of product development aims is positively significant.
Although the performance impact of product development postponement does not have a
direct impact on the bottom line of a company, it has a clear impact on demand as well as
customer satisfaction. Thereby it clearly results in a positive company performance impact.
Concluding on the discussion of the five postponement strategies, this subsection has
provided a deeper insight into the findings of the statistical analysis and offered the
opportunity to understand the relationship between uncertainty, aims, performance impact and
the respective postponement strategy in more depth. An explanation for all significant factors,
expected or unexpected, was provided. Note that uncertainty factors D and I as well as aims E,
G and H did not show any positive, significant influences. Thus, these factors are no triggers
for certain postponement strategies. In fact, for product development postponement this study
could not find any significant reason for uncertainty that positively triggers its implementation.
Summarizing, Figure 14 shows the research model adapted for the discussed findings of this
thesis. All positive effects and thus triggers are displayed. In order to make the findings
adequate for practical use, the following subsection will develop a postponement decision tool.
Purchasing Postponement
Product Development
Postponement
Production Postponement
Logistics Postponement
Price Postponement
Postponement Strategies1.) Operative
1.1 Price of competitive products (substitutes + complements)
1.2 Price of materials/components
2.) Customers
2.1 Changing customer preferences/tastes across product lines
2.2 Changes in customer preferences/tastes in regard to new or differentiated product
2.3 Irregularity of purchases (bullwhip etc.)
3.) Product Characteristics
3.1 Insufficient information for product development/Innovations
4.) Market
4.1 Various customer groups with different characteristics
•Increased customer orientation (long-term, e.g. regarding new product development)•Increased innovativeness
Postponement Aims
Performance Impact
Figure 14: Research model adapted for the findings of this study
6. Chapter: Discussion
61
6.2. The Decision Matrix
After discussing the findings of this research throughout the previous section, this subsection
summarizes the findings by means of presenting a postponement strategy decision matrix
based on the findings of this thesis as shown in Figure 15. In fact, the goal of this decision
matrix is to support managers, when analyzing and deciding on an appropriate postponement
strategy for their company. Based on the findings of this thesis, the matrix gives an indication
on which postponement strategy is and which one is not appropriate for the demand
uncertainty and postponement aim structure of the company at hand. Before being able to
evaluate the decision matrix for a specific company, the following steps have to be carefully
taken to retrieve the necessary input.
Demand Uncertainty Reasons
do not apply/ negative impact of uncertainty
Lead time reduction
Inventory reduction
Product price optimization
Benefit from decreasing
material prices
Increased customer
responsiveness (short-term)
Increased customer ori-
entation (long-term)
Increased innovativeness
Price fluctuations (material/component)
Purchasing Postponement
Logistics Postponement
Chang. customer preferences (prod. line)
Production Postponement
Production Postponement
Chang.customer preferences (new prod.)
Irregular purchases
Product Development Postponement
Logistics Postponement
Innovation
Price Postponement
Change in market share
Logistics Postponement
Various different customer groups
Price Postponement
Production Postponement
Production Postponement
High seasonality
Price Postponement
Weather
Logistics/ Production
Postponement
Purchasing Postponement
Purchasing Postponement
Consider applying/ positive impact on company performance
+ Logistics Postponement
Purchasing Postponement
Price/Production Postponement
Price Postponement
Production/ Purchasing
Postponement
Product Development Postponement
Do not apply/ negative impact on company performance
- Production Postponement
Logistics Postponement
Aims/Achievements
Figure 15: Postponement strategy decision matrix
In the first place, it is important to gain a clear understanding of the reasons underlying the
demand uncertainty that the company is facing. The following questions help to gather the
necessary information:
How variable is the demand uncertainty? Which data indicates demand variability?
Why do we face demand uncertainty? What are the reasons of demand uncertainty?
6. Chapter: Discussion
62
In the second step, the aims and desired achievements, which the company is striving for in
order to effectively cope with the effects of demand uncertainty, have to be defined. The
following questions support this process:
Which difficulties/challenges do we face with regard to demand uncertainty?
How do we currently respond to demand uncertainty?
What do we want to achieve in the near future? On which aspects do we want to improve?
Subsequently, after these two aspects are clearly analysed and defined, the postponement
strategy decision matrix can be evaluated to determine, if and how postponement could
support the business goals. If the matrix indicates a certain postponement strategy to be
appropriate, the explanations regarding this type of postponement strategy throughout this
thesis can be considered to evaluate in what way this strategy could fit to the company needs.
In addition, several case studies can be found in literature, that are also included in the
references of this thesis, providing ideas on how to implement such postponement strategies.
The argumentation along the matrix evaluation can be elaborated upon by using interview
results that were acquired from the interview with Company 1. In the first place, reasons for
demand uncertainty as well as aims that shall be achieved in order to reduce the demand
uncertainty effects need to be identified. Referring to the reasons of demand uncertainty,
Company 1 indicated that the weather has a great influence on the variability of their demand.
If the weather is unexpectedly bad during the summer months, demand will decrease, in
contrast to the forecast. In order to cope with this kind of demand uncertainty, Company 1 put
forward the goals, which they want to achieve in coping with the identified demand
uncertainty. In the first place, their aim is to achieve increased customer-orientation in the
short term to be able to react to customer needs in a flexible way. Furthermore, they aim to
ensure that inventories remain minimal, even in periods, when demand is lower than expected.
On the other hand, availability needs to be guaranteed, even when demand turns out to be
higher than expected.
Subsequent to the identification of reasons for demand uncertainty and aims, the decision
matrix is evaluated. As illustrated in Figure 16, it can be observed that seasonality in
combination with the aims of inventory reduction and increased customer orientation in the
short-term indicate purchasing postponement to be the adequate postponement strategy. The
reasoning for this combination has been explained in subsection 6.1 above. Company 1 has
6. Chapter: Discussion
63
already implemented purchasing postponement in form of vendor managed inventory. In this
case, not Company 1 but the supplier is responsible for replenishing inventory as needed,
thereby reducing safety stock requirements due to shorter lead times. On the other hand, the
supplier receives accurate and timely demand data and can thereby adapt his offering in order
to satisfy Company 1’s needs in the long-term. Additionally, other ways of applying
purchasing postponement should be explored. For instance, the product range could be
analysed for potential opportunities for just-in-time supplies. Also the potential from
deepened supplier relationships should be analysed to reduce inventory. At this point of the
matrix evaluation, it is important to search for implementation ideas of the identified
postponement strategies and to examine their suitability for the specific company processes.
Finally, the matrix also indicates that production postponement may also reach one of their
aims, but in presence of the given demand uncertainty reason, production postponement
should not be applied. Also Logistics postponement is not appropriate in the presence of
changing weather conditions, as indicated in the second column. This empirical research
could unfortunately not conclude on the adequacy of price and product development
postponement in this study, as no significant results were found.
Demand Uncertainty Reasons
do not apply/ negative impact of uncertainty
Lead time reduction
Inventory reduction
Product price optimization
Benefit from decreasing
material prices
Increased customer
responsive-ness (short-term)
Increased customer ori-
entation (long-term)
Increased innovative-ness
Price fluctuations (material/component)
Purchasing Postponement
Logistics Postponement
Chang. customer preferences (prod. line)
Production Postponement
Production Postponement
Chang.customer preferences (new prod.)
Irregular purchases
Product Development Postponement
Logistics Postponement
Innovation
Price Postponement
Change in market share
Logistics Postponement
Various different customer groups
Price Postponement
Production Postponement
Production Postponement
High seasonality
Price Postponement
Weather
Logistics/ Production
Postponement
Purchasing Postponement
Purchasing Postponement
Consider applying/ positive impact on company performance
+ Logistics Postponement
Purchasing Postponement
Price/Production Postponement
Price Postponement
Production/ Purchasing
Postponement
Product Development Postponement
Do not apply/negative impact on company performance
- Production Postponement
Logistics Postponement
Aims/Achievements
Figure 16: Example of matrix evaluation
6. Chapter: Discussion
64
Concluding, the postponement strategy decision matrix indicates which postponement
strategy is appropriate in the company specific situation, as well as which strategy is not
appropriate. Postponement strategies that are not indicated for a certain combination of
indicators, neither positively nor negatively, should be subject to further investigation as no
significant conclusion in this respect could be drawn from this research. ^
6.3. Conclusion
This chapter discussed the research findings for each postponement strategy in detail. For all
significant relations, suggestions for interpretation were provided. The findings were
additionally summarized by means of a postponement strategy decision matrix, which links
demand uncertainty reasons and aims to postponement strategies. Guidelines for how to apply
this matrix were provided, which enable managers to evaluate postponement opportunities
within their company. Note that it is very important to evaluate and assess the outcome of the
matrix in light of company specific processes and market specific characteristics.
7. Chapter: Conclusion
65
7. Conclusion and Future Research
This final chapter will in the first place conclude this research by answering the research
question, followed by an overview of the theoretical as well as managerial contributions.
Finally, the limitations of this research are discussed and future research opportunities
highlighted.
7.1. Problem Statement
The sub-questions8 developed for answering the problem statement, can be answered as
follows. First of all, this thesis defines postponement as the action of delaying certain
activities of the supply chain downstream of the customer-order decoupling point in order to
achieve more flexibility and responsiveness to demand uncertainty. A variety of
postponement strategies exist which can focus on any activity of the supply chain. From the
literature review it resulted that the most widely discussed and applied postponement
strategies are price postponement, logistics postponement, production postponement,
purchasing postponement and product development postponement. This group of strategies
was chosen for this research, also because it covers extensive parts of a companies supply
chain. Furthermore, this thesis positions demand uncertainty into relation with the just
mentioned postponement strategies. “Uncertainty is best understood as an information defect”
(Spender 1993, p.16). When incomplete demand information is available, demand cannot be
predicted with certainty. The literature review in combination with field interviews resulted in
a list of ten reasons underlying demand uncertainty that were expected to affect the choice of
postponement strategies, and which became part of the research model.
Based on the findings and discussion of the previous chapters, the research question,
including the last three sub-questions, can be answered. The research question asked: “How
do the reasons underlying demand uncertainty affect the choice of an appropriate
postponement strategy?” It can be concluded that demand uncertainty reasons indeed have a
8 Sub-questions: - How is postponement defined?
- Which postponement strategies do exist and what is their specific purpose? - How is demand uncertainty defined? - Which are the most important reasons for demand uncertainty? - Which postponement strategy reacts to which demand uncertainty dimension? - Which postponement strategy has what operational goal? - What is the impact of postponement strategies on company performance?
7. Chapter: Conclusion
66
significant effect on the choice of postponement strategies. Demand uncertainty and its
reasons are affecting a company’s supply chain and specific postponement strategies that can
be applied in these situations in order to mitigate the uncertainty risk or to improve company
performance under these conditions. Five hypotheses could be confirmed of which however
two proved to have a negative impact on the respective postponement strategy. In addition
eight other significant factors were found to be significant, four with a negative impact. These
results emphasize that demand uncertainty reasons cannot only foster postponement
implementation, but may also render a postponement strategy inappropriate. All positive
triggers are visualized by the adapted research construct in Figure 14. Furthermore, the
postponement strategy decision matrix as shown in Figure 15 shows in detail, which demand
uncertainty reasons have a significantly positive or negative impact on a certain postponement
strategy. In addition, the matrix visualizes which operational goals are achieved with which
postponement strategy. Finally, the findings of this thesis indicate a positive impact of
logistics, production and product development postponement and a negative impact of price
and purchasing postponement on company performance. The negative impact should however
not be overestimated, as price and purchasing postponement are frequently seen as a means to
ensure instead of create high profit.
7.2. Theoretical contributions
First of all, this study contributes to the understanding of postponement strategies in currently
published literature, by explaining their specific purpose and providing a discussion on
existing strategy decision tools. One aspect that is discussed several times in the literature
regarding postponement strategy choice is the link between postponement application and the
level of demand uncertainty. The research at hand extends this aspect by not only considering
the level of demand uncertainty, but also additionally analysing the reasons underlying
demand uncertainty, when investigating the question of which strategy to chose in which
specific situation. A causal chain including demand uncertainty – postponement strategy –
postponement aims – performance impact was developed and tested empirically. The causal
chain together with the resulting decision matrix results in a clear contribution to current
literature and to a deeper understanding of demand uncertainty in regard to postponement
strategy choice. Moreover, currently very little empirical research has been conduct in the
field of postponement strategies. This study conducted an empirical investigation and thereby
contributes to overcoming this gap in current literature.
7. Chapter: Conclusion
67
7.3. Managerial contributions
The thesis focuses on postponement strategy choice and thus leads to several managerial
implications. In the first place, when initiating a postponement evaluation process, managers
should first analyze their environment as well as their processes. In particular, this study
encourages managers to understand the reasons and sources of demand uncertainty their
company is facing, in contrast to merely focusing on the level of demand uncertainty.
Depending on the discovered reasons underlying uncertainty, certain postponement strategies
will result to be most appropriate and whose application possibilities should be explored. First
of all, managers should follow a postponement choice process that conducts certain analysis
steps that were explained in section 6.2. The evaluation of their analysis can in turn be based
on the postponement strategy decision matrix that was developed by means of the empirical
investigation of this study. The matrix gives a clear indication of which postponement
strategy appears most appropriate within the companies demand uncertainty situation and
thereby offers managers a good starting point for in depth investigation of potential
postponement applications. The literature review of this thesis supports practitioners to
understand, what postponement strategies are and provides them with insights into specific
postponement application. Furthermore, the presentation of existing postponement decision
tools provides management the opportunity to reflect on and evaluate their own possible
postponement implementations.
7.4. Limitations and future research
This final section will shed light on the limitations of this study and the resulting future
research opportunities. The first shortcoming concerns the choice of uncertainty indicators for
this research. Due to the fact that reasons and types of demand uncertainty are only rarely
discussed in literature, the reasons underlying demand uncertainty used in this thesis could not
be drawn on an established list of factors. The final list of indicators, which was created by
means of a literature review and the conducted interviews, may not be complete. Further
research should therefore focus on verifying and extending the list of uncertainty factors that
influences postponement strategy choice.
Furthermore, some surprising results of the online survey indicate that there might well have
been one or more possible misinterpretations of questions by respondents. As a consequence,
slightly inaccurate results may have occurred. This possibly stems from the fact that even
7. Chapter: Conclusion
68
though the concepts of postponement and demand uncertainty are generally applied, these
could well be named and defined differently throughout different companies and industries.
Therefore, future empirical research in the field of postponement as well as demand
uncertainty could focus on refining the questionnaire of this thesis, ensuring a clear common
understanding of stated concepts and survey questions among respondents.
Moreover, the focus of this study was not on one or two specific industries but on any non-
service firm. As not much research has been conducted in the field of postponement strategy
choice, the aim was to retrieve a broad picture that in turn offers the starting point for more
focused studies within specific industries. This high level point of view should be kept in
mind, when applying the proposed decision matrix. Furthermore, the sample size of this
research (n=53) could be considered to be relatively small, even though the study was not
limited to a certain industry. The limited amount of data may have caused some links of the
model to be incorrectly insignificant. Future research should therefore check this model by
using a larger sample and focus on specific industries to discover differences in postponement
applications across industries. This will help to improve the effectiveness of the postponement
strategy decision matrix.
Following from the argumentation above, it is important to note that the research at hand did
not include service firms into its analysis. The reason is that their business processes and thus
their postponement application may evolve quite differently. As mentioned throughout the
discussion of this thesis, demand and revenue management is increasingly gaining importance
in the service sector and can be categorized as a special kind of price postponement.
Therefore, future research could focus on the application of postponement strategies in service
firms as well as the applicability of postponement choice parameters, especially with regard to
demand management.
Furthermore, postponement strategies can differ for different company processes, depending
on the company size, main tasks and the supply chain structure. Thus, future research could
investigate what postponement strategies mean for different supply chain structures.
Furthermore, the findings discovered a common factor between logistics and production
postponement, which may also have caused certain misunderstanding of the postponement
concepts. If production postponement is located and applied in the same facility as a
centralized DC, meaning logistics postponement, it is difficult to clearly distinguish the
7. Chapter: Conclusion
69
effects of these two postponement strategies. In addition, future research could examine, if in
such a situation these two postponement strategies create a mutual positive effect.
Finally, this research could only draw conclusions on the effect of indicators on their LVs as
well as on the effect of LVs among each other. It could not draw clear conclusions on the
relation between indicators of different latent variables. Future research could accomplish this
by analysing mediation models, in which the relation between different indicators is mediated
by a latent variable, using a set of bootstrap methods as explained by Shrout and Bolger
(2002).
Despite these limitations, this study has provided valuable theoretical and practical
contributions and offers a good starting point for future research. Hopefully, this study will
foster further research efforts into the still widely unrevealed field of postponement strategies
and further support management activities in companies facing demand uncertainties.
Postponement success stories such as of the company Benetton are surely going to proliferate.
In light of the current progress in present literature on the topic of postponement facing
demand uncertainties, this thesis will help company decision making until future models can
be established.
Appendix
70
Appendix
Appendix I: Unstructured questionnaire for interviews
Guideline for interviews
1. Which product segments does your company serve?
2. How does the corresponding supply chain look like?
3. What degree of demand uncertainty is your division facing?
4. What are the reasons for this demand uncertainty?
5. How do you cope with this demand uncertainty?
6. Are you also using postponement strategies to cope with this demand uncertainty?
a. Or to cope with other phenomena?
b. Which postponement strategies?
c. Examples
7. In general, what do you think about the relations of the research model?
Appendix
71
Appendix II: Online questionnaire
Appendix
72
Appendix
73
Appendix
74
Appendix
75
Appendix
76
Uncertainty:
No uncertainty:
Appendix
77
Appendix
78
Appendix III: Variance Inflation Factors (VIF)
1.) Variance Inflation Factors for reasons underlying demand uncertainty
Uncertainty VIFA - Price fluctuations (compet. Product) 1.544B - Price fluctuations (material/component) 1.306C - chang. customer preferences (pr. line) 4.584D - chang. Customer preferences (new product) 5.125E - irregular purchases 1.205F - innovation 1.825G - change in market share 1.467H - many different customer groups 1.893I - high seasonality 1.795J - weather 1.525
2.) Variance Inflation Factors for postponement aims per strategy
Not taking the postponement variables for granted as latent variables, we can evaluate the
latent factors resulting from the PLS Regression instead. Analysing the table “proportion of
variance explained”, it can be concluded that it is sufficient to evaluate the first five latent
factors, because the adjusted R² decreases for the subsequent factors, indicating a penalty for
complexity. The values for the first five latent factors are shown in Table 1.
The factor loadings indicate certain relations between postponement strategies and reasons
underlying demand uncertainty. A cut-off value of 0.4 is used for latent factors loadings,
which proved to be adequate for exploratory research purposes (Hulland 1999). These results
show that different postponement strategies may be influenced by the same uncertainty
factors.
As also the table “variable importance in the projection” illustrates, reason for uncertainty H
is of minor importance (all VIP < 0.8). Reason A also shows low VIPs. However, as the
values are only slightly below the cut-off value and prove to have a positive effect in the
loadings analysis, this factor should not be neglected.
Table 1
Appendix
83
Table 2
Table 3
Table 4
References
84
References
Alderson, W. (1950). "Marketing efficiency and the principle of postponement." Cost and Profit Outlook 3: 15-18.
AMR Research (2009). Factors impacting demand variability.
Bagozzi, R. P., C. Fornell, et al. (1981). "Canonical correlation analysis as a special case of a structural relations model." Multivariate Behavioral Research 16(4): 437.
Bernhardt, I. (1977). "Vertical integration and demand variability." The Journal of Industrial Economics 25(3): 213-229.
Biller, S., A. Muriel, et al. (2006). "Impact of price postponement on capacity and flexibility investment decisions." Production & Operations Management 15(2): 198-214.
Bollen, K. and R. Lennox (1991). "Conventional wisdom on measurement: A structural equation perspective." Psychological Bulletin 110(2): 305.
Boone, C. A., C. W. Craighead, et al. (2007). "Postponement: an evolving supply chain concept." International Journal of Physical Distribution & Logistics Management 37(8): 594-611.
Bowersox, D. J., T. P. Stank, et al. (1999). "Lean Launch: Managing product introduction risk through response-based logistics." Journal of Product Innovation Management 16(6): 557-568.
Bucklin, L. P. (1965). "Postponement, speculation and the structure of distribution channels." Journal of Marketing Research (JMR) 2(1): 26-31.
Chin, W. W. (1998). Issues and opinion on Structural Equation Modeling. MIS Quarterly: 1-1.
Chin, W. W. (1998). The Partial Least Squares Approach to Structural Equation Modeling, Lawrence Erlbaum Association.
Cooper, R. G. and E. J. Kleinschmidt (1987). "New products: What separates winners from losers?" Journal of Product Innovation Management 4(3): 169-184.
Dapiran, P. (1992). "Benetton - Global logistics in action." International Journal of Physical Distribution & Logistics Management 22(6).
Diamantopoulos, A. and J. A. Siguaw (2006). "Formative versus reflective indicators in organizational measure development: A comparison and empirical illustration." British Journal of Management 17(4): 263-282.
Duncan, R. B. (1972). "Characteristics of organizational environments and perceived environmental uncertainty." Administrative Science Quarterly 17(3): 313-327.
References
85
Esposito Vinzi, V., G. Russolillo, et al. (2009). "A joint use of PLS Regression and PLS Path Modelling for a data analysis approach to latent variable modelling." Retrieved 03-03, 2010, from http://www.statssa.gov.za/isi2009/ScientificProgramme/IPMS/0185.pdf.
Granot, D. and S. Yin (2008). "Price and order postponement in a decentralized newsvendor model with multiplicative and price-dependent demand." Operations Research 56(1): 121-139.
Guinot C., J. Latreille, et al. (2001). "PLS Path Modelling and multiple table analysis. Application to the cosmetic habits of women in Ile-de-France." Chemometrics and Intelligent Systems 58: 247-259.
Heskett, J. L. and S. Signorelli (1984). "Benetton (A)." Harvard Business School Cases.
Hsu, S.-H., W.-H. Chen, et al. (2006). "Robustness testing of PLS, LISREL, EQS and ANN-based SEM for measuring customer satisfaction." Total Quality Management & Business Excellence 17(3): 355-371.
Hulland, J. (1999). "Use of partial least squares (PLS) in strategic management research: A review of four recent." Strategic Management Journal 20(2): 195.
Kumar, S. and S. Craig (2007). "Dell, Inc.'s closed loop supply chain for computer assembly plants." Information Knowledge Systems Management 6(3): 197-214.
Kutner, M. H., C. J. Nachtsheim, et al. (2004). Applied Linear Regression Models, McGraw-Hill Irwin, inc, Boston.
Lee, H. L. (2002). "Aligning supply chain strategies with product uncertainties." California Management Review 44(3): 105-119.
Lee, H. L. and V. Padmanabhan (1997). "Information distortion in a supply chain: The bullwhip effect." Management Science 43(4): 546.
Lee, H. L. and C. S. Tang (1997). "Modelling the costs and benefits of delayed product differentiation." Management Science 43(1): 40.
Lee, H. L. and S. Whang (2001). "Winning the Last Mile of E-Commerce. (cover story)." MIT Sloan Management Review 42(4): 54-62.
Liker, J. K. and T. Y. Choi (2004). "Building deep supplier relationships." Harvard Business Review 82(12): 104-113.
Matthews, P. and N. Syed (2004). "The power of postponement." Supply Chain Management Review 8(3): 28-34.
Mullins, J. W. and D. J. Sutherland (1998). "New Product Development in rapidly changing markets: An exploratory study." Journal of Product Innovation Management 15(3): 224-236.
References
86
Olhager, J. (2003). "Strategic positioning of the order penetration point." International Journal of Production Economics 85(3): 319-329.
Pagh, J. D. and M. C. Cooper (1998). "Supply Chain Postponement and Speculation Strategies: How to choose the right strategy." Journal of Business Logistics 19(2): 13-33.
Petter, S., D. Straub, et al. (2007). "Specifying formative constructs in information systems research." MIS Quarterly 31(4): 623-656.
Prockl, G., A. Bauer, et al., Eds. (2004). Entwicklungspfade und Meilensteine moderner Logistik, Gabler.
Shrout, P. E. and N. Bolger (2002). "Mediation in experimental and nonexperimental studies: New procedures and recommendations." Psychological Methods 7(4): 422-445.
Spender, J. C. (1993). "Some frontier activities around strategy theorizing." Journal of Management Studies 30(1): 11-30.
Stigler, G. (1939). "Production and distribution in the short run." The Journal of Political Economy 47(3): 305-327.
Tenenhaus, M. (2005). "PLS methodology to study relationships between hedonic judgements and product characteristics." Food Quality and Preference 16: 315-325.
Tenenhaus, M., V. E. Vinzi, et al. (2005). "PLS path modeling." Computational Statistics & Data Analysis 48(1): 159-205.
Thompson, B. (1984). Canonical Correlation Analysis: uses and interpretations, Sage Publications, Thousand Oaks, CA , .
van Hoek, R. I. (2001). "The rediscovery of postponement a literature review and directions for research." Journal of Operations Management 19(2): 161-184.
Van Mieghem, J. A. and M. Dada (1999). "Price versus Production Postponement: Capacity and Competition." Management Science 45(12): 1631.
Vinzi, V. E., W. W. Chin, et al. (2010). Handbook of Partial Least Squares, Springer Handbooks of Computational Statistics.
Weele, A. J. v. (2005). Purchasing and Supply Chain Management. London, UK, Thomson Learning.
Wetzels, M., G. Odekerken-Schröder, et al. (2009). "Using PLS Path Modeling for assessing hierarchical construct models: guidelines and empirical illustration." MIS Quarterly 33(1): 177-195.
Yang, B. and N. Burns (2003). "Implications of postponement for the supply chain." International Journal of Production Research 41(9): 2075-2090.
References
87
Yang, B., N. D. Burns, et al. (2004). "Management of uncertainty through postponement." International Journal of Production Research 42(6): 1049-1064.
Yang, B., N. D. Burns, et al. (2005). "An empirical investigation into the barriers to postponement." International Journal of Production Research 43(5): 911-1005.
Yang, B., Y. Yang, et al. (2007). "Postponement: an inter-organizational perspective." International Journal of Production Research 45(4): 971-988.