Supply Chain Forum: An International Journal

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tscf20

Response surface methodology of dual-channel

green supply-chain pricing model by considering
uncertainty

Januardi Januardi & Widodo E.

To cite this article: Januardi Januardi & Widodo E. (2020): Response surface methodology of
dual-channel green supply-chain pricing model by considering uncertainty, Supply Chain Forum: An
International Journal, DOI: 10.1080/16258312.2020.1788904

To link to this article: https://doi.org/10.1080/16258312.2020.1788904

Published online: 13 Jul 2020.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tscf20
SUPPLY CHAIN FORUM: AN INTERNATIONAL JOURNAL
https://doi.org/10.1080/16258312.2020.1788904

Response surface methodology of dual-channel green supply-chain pricing

model by considering uncertainty
Januardi Januardi and Widodo E.
Department of Industrial Engineering, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia

ABSTRACT KEYWORDS
Companies apply dual-channel supply-chain (DCSC) by allocating online facilities without Dual-channel supply chain;
leaving conventional retailers. As a result, carbon emissions from production activities have pricing; probabilistic;
increased in the company structure. Nowadays, DCSC researchers have a tendency to set their response surface
methodology; uncertainty
models in deterministic settings. These models cannot cope with the structure complexity. This
research used Response Surface Methodology to analyse uncertainty in DCSC. The decision
variables were the retailers and online prices, while the responses were price channel sensitiv­
ities (βo and βr), cross-channel sensitivities (γor and γro), profit estimation, and carbon emission
estimation. In illustrating the real-world competition under oligopoly games, the optimisation
phase used nonlinear programming under cooperative and non-cooperative scenarios. The
result showed that in each channel, price combination had a significant effect on the response
variables. Price differences on each channel could confirm the changes of customer preference
channel. This study finds that the application of a response surface methodology can describe
the significance of price to the DCSC system.

Introduction came from the energy sector including transportation

and logistics (KLHK 2017). Based on this data, the
Industry 4.0 phenomenon is a worldwide trend that
hypothesis can be generated that DCSC also contri­
makes various companies utilise online facilities to
butes to carbon emissions which strengthens that
support product marketing. However, the practice of
multi-channel supply chain needs to cope with the
indirect channel using retailers for market distribution
environmental approach to achieve sustainable busi­
is still in place. It means that the internet of things (IoT),
ness system (Besbes et al. 2013). Therefore, supply
data exchange, cloud computing, and computer cog­
chain modelling and optimisation need to consider
nitive usage have spread widely. Indonesians have
various environmental issues such as carbon trading
used the internet to support their business activities.
(Sinha and Modak 2019), carbon footprint (Gautier
A survey to internet users in Indonesia by APJII (APJII
et al. 2013), and greenhouse gasses. By combining
2017) shows that e-commerce business transactions
environmental issues in DCSC, researchers and compa­
have increased since 2010. The e-commerce transac­
nies may support the business world to a sustainable
tions began by two hundred thousand users in 2010
supply chain management (SSCM).
and increased to almost sixty million users in 2017.
At the same time, mathematical modelling
That is why the dual-channel supply chain or DCSC is
research for supply chain (SC) and pricing has been
one of the effective methods for a company to
developed, but only on the deterministic problems.
broaden its market by utilising retailers and online
The probabilistic and stochastic problems are
channels (Widodo 2015). Utilisation of more than one
included in the uncertain problems (Simangunsong,
channel in the supply chain can increase the total costs
Hendry, and Stevenson 2012). Uncertainty is a general
in a company’s operations, but it raises demands in
condition and situation that is included in the VUCA
unexpected ways (Lang and Bressolles 2013).
situation. VUCA situation is a volatile, uncertain, com­
DCSC strategy helps companies to gain more prof­
plex, and ambiguous problem (Siegfried, LeBris, and
its, its effect on the environment, however, should take
Stewart 2018). Supply chain is one of the fundamental
into account. The market expansion can increase logis­
factors in the modern economy. It is shown that
tic armada to satisfy more demands. Consequently, it
economic problems are more volatile and uncertain
may cause higher carbon emission effect on the envir­
to be forecasted (Campbell 2007). Economic uncer­
onment. A survey from the Ministry of Environment
tainty evolves throughout the time that could
and Forestry of Indonesia found that in 2016 Indonesia
increase forecast errors (Jo and Sekkel 2017). Of all
produced approximately 619 ton CO2, in which 18.94%
the SC optimisation studies in Google Scholar, there

CONTACT E. Widodo [email protected] Department of Industrial Engineering, Institut Teknologi Sepuluh Nopember, Surabaya,
Indonesia
© 2020 Kedge Business School
2 J. JANUARDI AND E. WIDODO

were 43% of supply chain researches that considered considering the variable of uncertainty and
uncertainty in 2010. This number grows every year. In environmental issue?
2018, there were 61% SC researches considering (3) What is the outcome of cooperative and non-
uncertainty from the author’s perspective. Research cooperative game scenarios from the system
on uncertainty has become a crucial issue both in that considers environmental constraints?
business and academics fields. The uncertainty pat­
tern can be analysed through simulation and fore­ This research aimed to analyse and model offline and
casting (Wu and Mortveit 2015). online pricing in DCSC as well as optimise pricing
This research discussed pricing decisions in the decisions in probabilistic problems with carbon emis­
dual-channel supply chain. There are three main sion limitation. Moreover, a sensitivity analysis was
aspects in DCSC, i.e. market channel decision, cross- conducted by changing the parameters. The scope of
channel coordination, and pricing policy of each chan­ research was only on one product, the actual demands
nel (Zhou 2016). Market channel decisions are usually were 100, the factors analysed were conventional retai­
selected between online facilities and retailers. The lers and online price, and the response variables were
coordination of each channel is performed to adjust channel price sensitivity, cross-channel sensitivity, and
the consumer demand. In DCSC, research on cross- profit estimation. The currency used in this research
channel coordination is essential to ensure that each was the Indonesian Rupiah (IDR). The environmental
channel gets the maximum profit. The game theory impact value in this research was carbon emission in
method is utilised to identify an optimal decision in the form of CO2, which resulted from the fulfiling
this phenomenon. Several DCSC researches also con­ demand. The last scopes were only one retailer and
sider product substitution (Widodo 2015, 2017). The one online facility. The assumptions of this research
game strategies used are the cooperative game where were the demand channel preference was probabilis­
retailers and online facilities make pricing decisions tic, the experiment was performed by the same con­
together and the non-cooperative game where each sumer, the online and offline resellers were ignored,
channel makes a pricing decision without any coordi­ the transportation of each channel was pickup trucks
nation. Meanwhile, the pricing policy is a fundamental for conventional retailer and motorcycles for the online
connection between demand and corporation. In addi­ facility, and the preference variable used was only
tion, price can also determine product demand pricing in each channel. This research can be used in
deterministically. science as probabilistic optimisation in DCSC. This
Research on deterministic DCSC has been presented research employed RSM to test the price factors and
in several papers (Widodo 2015, 2017; Nuha and the significant variables, and then optimise them in
Widodo 2016). Some supply chain research has con­ finding the highest profit in coalition or coordination
sidered carbon emission; however, the model is in scenarios. Research in the supply chain area that exam­
deterministic settings using the economic production ines the variables before optimising the response is
quantity (EPQ) and has not considered the pricing very limited. In modern problems, the manager can
factor (Sinha and Modak 2019). Moreover, there are make pricing decisions to maximise the profit in a real-
some DCSC research that has considered the environ­ world experiment.
ment (Li et al. 2016; Ji, Zhang, and Yang 2017; Heydari,
Govindan, and Aslani 2018; Li 2018; Modak et al. 2018).
Literature review
Pricing can affect DCSC structure as well. This research
was conducted in a small food enterprise in Surabaya Supply chain management has always been an inter­
that applied DCSC. In the offline channel, the products esting issue for companies and academic researchers.
are sold through private and public retailers. In the It manages the way to connect different parties in their
online channel, the products are sold through the business process system with their concern. The supply
website and social media. The probabilistic modelling chain is started from the suppliers and ended to the
in DCSC is rarely utilised until the optimisation phase; customers, either through the distributor or direct sell­
therefore, this research employed the Response ing. However, adding more channels to sales activities
Surface Methodology (RSM) to analyse DCSC uncer­ can grow more markets and profits. In this case, tradi­
tainty and optimise pricing to gain higher profits. tional companies use retailers as an indirect channel
From the problem statement above, the research ques­ and add a direct channel through their online facilities
tions of this research are as follows: (Mitra 2014).
In DCSC, the consumers’ need for a product is in the
(1) Does the pricing decision of the retailers and form of dmax, their online channel transaction prefer­
online facilities affect each channel’s profit and ence is in the form of ρ, and their conventional retailer
the environment? preferences are in the form of 1-ρ. Pricing decision of
(2) How does the company develop a pricing deci­ each channel (Pro for online price and Prr for retailer’s
sion in the dual-channel supply chain by price) has caused the real demand (Dmax) choose not to
 SUPPLY CHAIN FORUM: AN INTERNATIONAL JOURNAL 3

buy from their channel preference (βo for online sensi­ From the statistical view, there is an exception from
tivity price and βr retailer sensitivity price), or choose to the supply chain modellers to use the mathematical
switch their channel (γor for cross-sensitivity from method in stochastic, i.e. the significance of their vari­
online to retailer’s channel and γro for cross-sensitivity ables. The modellers always determine that their vari­
from retailer’s to online channel) (Widodo 2015). Those ables are usually significant to the real-world system.
factors can determine the demand for each channel. This kind of problem has enforced the variables into
Meanwhile, the demand in each channel can deter­ deterministic even though the data is stochastic.
mine the profit and the carbon emission from logistic Therefore, some of the supply chain modellers employ
armada used to fulfil the demand of each channel. The the statistical approach for their analysis of the variable
conceptual model of this problem is presented in of uncertainty. This results in another problem since
Figure 1. hypothesis testing of the significance of the variables is
As a matter of concern, adding more channels makes always in categorical variables. Categorical variables
an increase for the logistic armada. The increase of the complicate the problem analysis. Sometimes, the deci­
armada can generate higher carbon emission in the sion-maker needs quantitative variables for an abso­
logistic process for the demand (Liu 2018). From this lute decision. According to the problem statement, the
point of view, supply chain managers and researchers RSM is used for coping with such problems. Based on
need to make a decision that can gain more profits for uncertainty views, RSM is an embedded experimental
the company and lower the environmental impacts approach to test whether the variables and factors are
such as carbon emission. In accordance with the pre­ significant to the problems. The quantitative result can
vious statement, researchers in the world have devel­ be observed in the polynomial regression from the
oped many mathematical and statistical models and problems. The difference of the RSM model from the
analyses for managers to make decisions. There is, how­ original mathematical model is the RSM model makes
ever, a certain issue for supply chain management that a simpler model into a second-order model, unlike the
has bothered the researcher and supply chain modeller mathematical model which has many constraints and
for a long time, i.e. uncertainty. Uncertain supply chain variables.
in a mathematical model is modelled by using the As can be seen, supply chains have many variables
stochastic process where the number of each variable that can be considered uncertain (Mason-Jones and
is randomly generated. Usually, the generation of the Towill 2000), such as lead time, demand, and customer
number uses a distribution pattern. preference. Many kinds of research in the supply chain

Figure 1. Conceptual model of the research problem.

4 J. JANUARDI AND E. WIDODO

cope with uncertainty by using stochastic settings in approach, the pricing equations (Widodo 2015) are as
the mathematical model and simulation approach. The follows:
stochastic setting is a random number generation that
Do ¼ ρDmax βo Pro þ γro Prr (1)
is forecasted by distribution patterns. However, they
determined that their variables are significant to their
Dr ¼ ð1 ρÞDmax βr Prr þ γor Prr (2)
system observation. Determining something without
any testing is the same as the deterministic model.
From this observation, there is no uncertainty model
Experimental design
that tests their variables and parameters until the deci­
sion-making phase. As it is known, variables testing can The experimental design was employed to analyse the
be approached by using statistical methods, such as type of uncertainty in DCSC because pricing experi­
hypotheses testing. One of the methodologies that ment can analyse the customer behaviour towards the
consists of hypotheses testing is an experimental product price (Kiefer, Kelly, and Burdett 1994; Kramer
design (Elbert, Bogusch, and Özsucu 2014). The experi­ 2016). The reason is that the retail and online price
mental approach can be used to support pricing deci­ factors were included in the continuous variable. The
sion by considering uncertainty (Kramer 2016). Pricing experiment was conducted through questionnaire dis­
experiment can perceive the effect of price to demand tribution in the survey analysis. The questionnaire was
quantity of product (Kiefer, Kelly, and Burdett 1994). distributed to 100 correspondents that wanted to pur­
The application of categorical variables in the experi­ chase the observed product. Then, the questionnaire
mental design of supply chain research can complicate was collected from all correspondents and analysed
the observations (Shil 2010). Due to these reasons, RSM from the arrange price of retail and online channel
is used to conduct experiments and modelling of con­ combinations and their decisions towards the product.
sumer preferences at dual-channel price, and model From their decisions, the data can show customer
the results into regression equations with continuous acceptance by considering the price to form a profit
variables for better observation. estimation and the carbon emission from an effort to
The research gap of this study is that the pricing distribute the product to the customers.
decision and carbon emission cost are discussed in The response variables were online price channel
deterministic settings (Modak et al. 2018). Dual- sensitivity (βo), retailer price channel sensitivity (βr),
channel supply chain research that considers uncer­ cross-price channel sensitivity from online to retailer
tainty has been carried out by many researchers using channel (γor), and cross-price channel sensitivity from
stochastic settings (Modak 2017; Modak and Kelle retailer to online channel (γro). Price sensitivity
2019a, 2019b); however, those researches did not becomes a major factor in pricing to understand the
cope with the uncertainty of the variable determina­ characteristic of demand elasticity (Kim, Blattberg, and
tion, such as using hypotheses testing in statistics to Rossi 1995). After the data were obtained, another
understand whether the variables are significant to the response variable could be determined, such as profit
real-world systems. This research mainly focused on and carbon emission estimation in each channel. The
the study of the effect of two channels in dual sales experimental design utilised a Central Composite
activities, unlike other research who applied for more Design (CCD) because the analysed factors were only
than two channels (Modak 2017; Niranjan et al. 2019). two. The experiment factors are presented in Table 3.
DCSC probabilistic research in considering uncertainty The price was tested from 14,000 IDR to 20,000 IDR.
from the variable testing to the optimisation phase has From the company perspectives, the price approach­
not been performed very frequently by academics and ing the unit cost was 14,000 IDR since it was profitable,
managers. and their customer willingness to purchase was 20,000
IDR. There were no price range differences between
online and retail channels because of the rotatability
Methods value (α) was 1.414 for two-factor design in CCD. The
experiment was conducted to answer the first research
Demand equations
question.
Pricing decision in the dual-channel supply chain is
affected by many factors. The factors discussed in this
Second-order regression model
study were price channel sensitivity and cross-price
channel sensitivity. Price channel sensitivity is The data were obtained from random correspondents.
a sensitivity value that refers to the number of lost This means that the questionnaire was not distributed
demands if the channel price increases. Cross-price to certain correspondents. The data type was primary
channel sensitivity is similar to price sensitivity, but with two factors and nine-level of design experiments.
the cross-demand value is determined when the The outermost points of 12,757 IDR and 21,243 IDR
price of the cross channel has been set. In deterministic were the extreme points used to examine the effect on
 SUPPLY CHAIN FORUM: AN INTERNATIONAL JOURNAL 5

Table 1. Research gap.

Uncertainty analysis
Response variables Variables significant
Pricing Price Environment Distribution Continuous Categorical
No Author decision sensitivity issue pattern variables variables
1 Elbert, Bogusch, and Özsucu √
(2014)
2 Kramer (2016) √ √
3 Li et al. (2016) √ √
4 Zhou (2016) √ √
5 Ji, Zhang, and Yang (2017) √ √ √
6 Modak (2017) √ √
7 Heydari, Govindan, and Aslani √ √
(2018)
8 Modak et al. (2018) √ √
9 Modak and Kelle (2019a) √ √
10 Modak and Kelle (2019b) √ √
11 Sinha and Modak (2019) √
12 This Paper √ √ √ √

the system. This is also known as the rotatability fac­ the problem with stochastic volatility and the effect of
tors. From the experimental design in Table 2, the uncertainty (Shin and Zhong 2018). The formulation of
questionnaire was used as a simulator to see what the experiment is shown in Equation (3) (Myers,
will happen to price sensitivities of each channel, cross- Montgomery, and Cook 2009).
price sensitivities, profits of each channel, and carbon
Yresponse ¼β0 þβ1 Pro þβ2 Prr þβ3 Pro 2 þβ4 Prr 2 þβ5 Pro Prr þε
emission generations if the online facility sells the
product in xx IDR and the retailer sells the product in (3)
xx IDR by using the above experimental design. where the notations are:
The obtained response was calculated from the The regression was continued with the validation
obtained demand that was converted from the ques­ test using a statistical test of analysis of variance
tionnaire survey. The second-order model was able to (ANOVA) to show the significance of the regression
answer the second research question. After the experi­ model to the data. This testing was employed to iden­
ment was conducted by distributing the questionnaire, tify the uncertainty of variable involvement in the
the calculation continued to form a second-order system. Then, the model continued to R-square test
regression model using Design Expert V.11. The to demonstrate the relationship of the model to the
model used regression because this model can identify data. The acceptance of the r-square used was 0.7
because the data type was primary data (Mukaka
2012). The validation was carried out because polyno­
Table 2. Demand equation parameters in DCSC. mial regression can be employed to analyse factors
Notation Description and their combination. In some experiment designs
Do Demand from online facility using a regression model, polynomial models cannot
Dr Demand from retailers
ρ Percentage of customer channel preference
be adjusted to noisy estimates of responses (Gelman
Dmax Product demand when the price is set near unit cost and Imbens 2017). The results of the probability value
βo Price sensitivity in online facilities from the hypotheses testing can be used to strengthen
βr Price sensitivity through retailers
γro Cross price channel sensitivity from retailer to online channel the answers to the first and second research questions.
γor Cross price channel sensitivity from online to retailer channel
Pro Product price in online facility (IDR) – Decision Variables
Prr Product price in retailer (IDR) – Decision Variables Optimisation phase
The optimisation phase in the regression model was
Table 3. Factor level of DCSC pricing experiment. performed to obtain the global optimum value. The
Coded variables optimisation utilised Non-linear Programming (Yeniay
Factor (IDR) −1.414 −1 0 1 1.414 2014) as the objective function to maximise the profit
Retailer price (Prr) 12,757 14,000 17,000 20,000 21,243 of each channel. The optimisation was performed by
Online Price (Pro) 12,757 14,000 17,000 20,000 21,243 using Lingo v.11. The optimisation was defined through
the cooperative and non-cooperative game scenarios. In
the cooperative game scenario, the optimisation was
Table 4. RSM equation parameters.
carried out simultaneously together for each channel.
Notation Description
βi Regression Constant (i)
In the non-cooperative game scenario, Stackelberg lea­
Pro Product price in online facilities (IDR) – Decision Variables dership was applied. The leader was the retailer channel
Prr Product price in retailers (IDR) – Decision Variables because according to (He et al. 2018) in the food
6 J. JANUARDI AND E. WIDODO

business, the existing firm is the leader while the enter­ γor ¼ 0:001 1:434 � 10 7 Prr þ1:437 � 10 7 Pro
ing firm is the follower. Therefore, the leader was the þ1:951 � 10 12
Prr Pro þ1:418 � 10 12
Prr 2 (6)
retailer channel, and the follower was the online chan­ 12 2
4:174 � 10 Pro
nel. The optimisation was done sequentially by maximis­
ing the profit in the follower channel and then the
leader was adjusting. It is carried out by the leader to γro ¼ 0:009 þ 2:014 � 10 7 Prr 1:231 � 10 6 Pro
understand the behaviour of the follower’s decision. 6:498 � 10 11
Prr Pro þ3:34 � 10 11
Prr 2 (7)
However, this statement is different from the 11 2
þ6:035 � 10 Pro
Stackelberg game between customers and their suppli­
ers (Gahbiche et al. 2016), where customers are the The sensitivity of retailer (Dr) and online (Do)
leader and there is a coalition between suppliers. It demands were determined from the sensitivity value
indicates cooperative and non-cooperative scenarios in of each price of the channels (βo and βr) and cross-
one business system. price channel sensitivity (γor and γro). The demand for
each channel can be transformed into profit and carbon
emission estimation. From the research findings, the
Result
pricing combination of online and retail channels
Uncertainty model of dual-channel supply chain showed many different values of the price and cross-
pricing price sensitivity. These results are not commonly found
in many dual channels of supply chain research. Usually,
After the variables of the experiment were set, the ques­
the price and cross-price sensitivity are considered as
tionnaire was distributed. The obtained response is pre­
the parameters. The response of those variables is pre­
sented in Table 5. According to the response in Table 5,
sented in Table 6.
the research continued to form a second-order regres­
sion model. The regression model can be employed to ProfitðOffÞ ¼ 3:12 � 10 6 þ191:77Prr þ278:12Pro
forecast uncertainty between each channel combina­ (8)
þ0:01Prr Pro 0:012Prr 2 0:01Pro 2
tion. First, the model of regression model for each price
sensitivity is displayed in Equations (4) and (5).
ProfitðonÞ ¼ 2:46 � 106 113:75Prr 180:19Pro
βr ¼ 0:011 þ 2:455 � 10 7 Prr 1:409 � 10 6 Pro (9)
12 12 2
0:01Prr Pro þ0:01Prr 2 þ0:01Pro 2
6:988 � 10 Prr Pro þ2:937 � 10 Prr (4)
11 2
þ3:808 � 10 Pro
Table 6. Experimental design and responses of profit and
βo ¼ 0:0002 1:978 � 10 8 Prr þ1:287 � 10 7 Pro carbon emission estimation.
12 12 Factors Response variables
þ5:952 � 10 Prr Pro 3:238 � 10 Prr 2
Carbon
12
6:626 � 10 Pro 2 Retailer Emission
price Online price Profit (IDR) (g CO2/km)
(5)
(IDR) (IDR) Retailer Online Retailer Online
After modelling the demand loss through price 14,000 14,000 978,959 93,730 4.68 1.99
20,000 14,000 608,672 664,104 2.04 13.80
channel sensitivity, the calculation continued. From 14,000 20,000 1,222,123 28,506 5.84 0.43
the demand loss of each channel, many demands 20,000 20,000 1,242,552 179,909 4.15 2.64
will go across the channel because of the price 12,757 17,000 1,169,870 16,927 6.14 0.31
21,243 17,000 695,359 612,267 2.19 10.47
combinations set in the form of cross-price channel 17,000 12,757 585,281 520,301 2.31 11.90
sensitivity. The model is presented in Equations (6) 17,000 21,243 1,317,965 57,118 5.18 0.80
17,000 17,000 1,189,815 85,095 4.68 1.49
and (7).

Table 5. Experimental design and responses of price sensitivity.

Factor Response variables
Price channel sensitivity Cross channel sensitivity
Retailer price (IDR) Online price (IDR) βr βo γor γro
14,000 14,000 0.0018 0.0006 0.0004 0.0004
20,000 14,000 0.0030 0.0003 0.0001 0.0030
14,000 20,000 0.0007 0.0005 0.0006 0.0000
20,000 20,000 0.0017 0.0004 0.0003 0.0004
12,757 17,000 0.0005 0.0006 0.0007 0.0000
21,243 17,000 0.0028 0.0003 0.0001 0.0019
17,000 12,757 0.0033 0.0003 0.0001 0.0028
17,000 21,243 0.0012 0.0004 0.0005 0.0001
17,000 17,000 0.0015 0.0005 0.0004 0.0002
 SUPPLY CHAIN FORUM: AN INTERNATIONAL JOURNAL 7

The profit regression model for the retailer channel Table 8. Optimisation result.
reveals that higher prices in the online channel can Cooperative Non-cooperative
Description Notation scenario scenario
switch the demand from an online channel to the
Decision Retailer Price (IDR) 23,767.95 17,804.41
retailer, but it occurs only when the price in the retailer Variables Online Price (IDR) 24,479.66 19,054.92
channel is cheaper. A demand increase in the retailer’s Response βr 0.003202 0.001352
Variables βo 0.000134 0.000478
channel can also increase the profit of its channel. γor 0.000128 0.000364
Many demands will shift to online channels when the γro 0.001293 0.000000
Retailer Profit (IDR) 853,326 1,281,832
price in the retailers is higher than the online channel, Online Profit (IDR) 608,560 63,023
causing an increase in the demand in the online chan­ Total Profit (IDR) 1,461,886 1,344,855
nel. Although demand increase is a positive side for the Retailer Emission (g 1.93 4.81
CO2/km)
profit, it might have a negative effect on carbon emis­ Online Emission (g 8.92 0.69
sion. To support demand, logistics are required to CO2/km)
Total Emission (g 10.84 5.50
connect the actors and logistics dominantly use vehi­ CO2/km)
cles. From the price sensitivity response, the calcula­
tion showed a regression model for emission
production. The model is presented in Equation (11).
a cooperative scenario, the retailers and the firm in the
EmissionðoffÞ ¼ 5:48 1:04 � 10 4 Prr þ0:001Pro online channel had a coalition to gain higher profits.
þ2:65 � 10 8 Prr Pro 2:23 � 10 8 Prr 2 This profit was higher than the non-cooperative sce­
4:56 � 10 8 Pro 2 nario. For the pricing decision, each scenario had
a small price difference. It caused the value of switch­
(10)
ing channels from the retailer to online was so small.
Meanwhile, in the non-cooperative scenario, the price
EmissionðonÞ ¼ 46:62 0:001Prr 0:01Pro
difference was higher since as a leader, the retailers
2:67 � 10 7 Prr Pro þ1:85 � 10 7 Prr 2 made a pricing decision after their followers decided
þ2:38 � 10 7 Pro 2 the amount of price they used. That is the reason why
(11) the leader has more profits compared to other scenar­
ios, but their join profit is lower than the cooperative
The regression model was validated by testing the
scenario.
hypothesis to ensure that the model is significant to
the data. The hypothesis test utilised the analysis of
variance. Table 5 shows that all the forms of regression Discussion
models from the calculation were significant to the
The regression model in Equations (4) and (5) explains
experimental data. All the p-values of the models
that the retailer’s price sensitivity (βr) was higher when
were under the significance level (Myers,
the retailer price was high and the online price was
Montgomery, and Cook 2009). The R-squared values
low. This could increase the demand loss in the retailer
of all models had a strong relationship with the data
channel. In the regression model of online price sensi­
(Mukaka 2012). Table 7 shows the variable of certainty
tivity (βo), when both channels had the lowest price
for the first and second research questions about the
and the online price sensitivity was higher, the
variable of significance to the system.
demand loss in the online channel was higher, which
results in the decreased demand. The contour model
Pricing decision (Figure 2) is used to illustrate the model in the graphi­
cal image.
The regression model from the previous chapter was The model in the previous section explained that
employed to perform the optimisation using non- simultaneous higher or smaller pricing decisions on
linear programming. The optimisation was carried out online channel and retailers increase the value of cross-
to maximise profit in cooperative and non-cooperative channel from online channel to the retailers (γor). It
game scenarios. The result is displayed in Table 8. In indicates that when the value of each channel is the
same, consumers prefer retailers than the online facil­
Table 7. Validation testing of regression model. ity. In the cross channel from retailer to online (γro),
Regression model P-Value* R-squared Adj R-squared however, lower price in the online facility than the
βr (Equation 4) <0.001 0.969 0.946 retailers accommodates the channel switch to the
βo (Equation 5) <0.001 0.968 0.945
γor (Equation 6) 0.025 0.792 0.643
online facility. The survey analysis found that βr, βo,
γro (Equation 7) <0.001 0.93 0.881 γor, and γro can provide different values. The reason
ProfitOffline (Equation 8) <0.001 0.967 0.943 was due to the price combination in each channel. The
ProfitOnline (Equation 9) <0.001 0.984 0.973
EmissionOffline (Equation 10) <0.001 0.982 0.97 3d model in Figure 2 adequately presents a curve. It
EmissionOnline (Equation 11) <0.001 0.986 0.976 indicates that the extreme pricing point may not put
8 J. JANUARDI AND E. WIDODO

Figure 2. Contour model of βr and βo.

those parameters in extreme points. Usually, many in channel switching from online facilities to the retai­
mathematical models set those parameters into fixed lers, which occurs when the price in the retail channel
value. This study contradicts many types of DCSC is cheaper. Hence, it produces higher carbon emissions
research. The model regression above is described in the retail channel. Many demands will move to the
using the contour model in Figure 3. online channel if the price in the retailers is higher than
The demand generation from price and cross-price that in the online channel, causing an increase in the
sensitivity of each channel increased because of the carbon emission of the online channel. The contour
price difference decisions from each channel manager. plot is presented in Figure 5.
The demand for each channel can simulate the Besides the parameter variation, the research also
obtained profit of the online facility held by the man­ found that price decisions in each channel can construct
ufacturer or suppliers and offline facility held by the a profit and carbon emission estimation. In addition,
retailers. Figure 4 shows that higher pricing leads to some of the models were forming a curve. In carbon
higher received profit in each channel. It means that emissions, the curve is not precisely formed. However,
the adjusted sets of factors from the experimental the model showed the real-world situation if a channel
design showed a higher number and response of the sets the price too low, and other channels set their price
profits. Even though the price was set at 20,000 IDR in too high. In this case, many demands will go to the
online and offline channels, customers were still willing lowest price channel. Increasing demand also causes
to purchase. The decreased number of demands kept higher carbon emissions. The research experiment was
making a profit for the company. performed to a product sold by a company through its
In emission regression, every result from profit online facility and disruption by retailers as the indirect
regression was directly proportional to carbon emis­ channel. The research simulated the price sensitivity,
sion. An increase in prices on the online channel results cross-price sensitivity, each channel profit, and carbon

Figure 3. Contour model of γor and γro.

 SUPPLY CHAIN FORUM: AN INTERNATIONAL JOURNAL 9

Figure 4. Contour model of retailer and online channel profit.

Figure 5. Contour model of retailer and online channel facility carbon emission.

emission development using a questionnaire survey. It this research can show the managers that price differ­
is similar to the DCSC mathematical model, only in ence of each channel provides an impact to the price
a simpler form. The research also found that different and cross-price sensitivity. Therefore, they may consider
price combinations can change the price and cross- the right price to compete with other channels. For their
price sensitivities of customer preference in the dual- decision-making process, this method uses the optimi­
channel supply chain. This variable is usually used as sation phase through the second-order model.
a parameter in many mathematical models of DCSC. The second-order model enforces the data to form an
With cooperative and non-cooperative scenarios, the optimum global form, so the computation becomes
models show that cooperation between the online facil­ faster. However, the model from this research can be
ity and retailers can make a higher profit than in the developed by adding more variables and game scenar­
non-cooperation scenario. ios. This development needs more experiments to make
For managerial implications, RSM can be utilised by a robust decision. This research is relevant to supply
company decision-makers, supply chain managers, and chain study, especially supply chain uncertainty. The
business analysts to understand customer behaviour supply chain parameter has a variation that causes
through the online facility and retail sales. As a bullwhip effect. This variation considers a VUCA pro­
a response, it can adjust to the business environment blem in supply chain. The application of experimental
system. They can also identify their hypothesis about studies, such as RSM, can cope with the supply chain
a variable that depends on their supply chain system uncertainty problem from the significance of the vari­
through the significance test. Performing experimental ables. From those statements, this research is having an
activities are simpler than making a mathematical impact on a novel way to conduct pricing decisions in
model with complex problems where the variables are the DCSC research and practice besides mathematical
not necessarily significant to the system. The variables of approach. The mathematical model usually enforces the
10 J. JANUARDI AND E. WIDODO

model to have significance and usually has difficulty in Notes on contributors

coping with uncertainty.
Januardi Januardi was a master graduate of Industrial System
Optimization of Institut Teknologi Sepuluh Nopember,
Surabaya, Indonesia. His has a bachelor’s degree in
Conclusion Agroindustrial Technology from IPB University, Bogor,
Indonesia. Currently, he continues his PhD study at National
This study represents an analytical pricing model in Taiwan University of Science & Technology, Taipei, Taiwan.
the dual-channel of the green supply chain by using His research interests are in quantitative modelling in opera­
a response surface methodology to handle the tion research, decision-making using statistical approach,
uncertainty problem. The model has successfully and experimental design especially the application of
response surface methodology.(ORCID ID: 0000-0003-0772-
represented the structure of DCSC in real-world
5939)
data by using variance and r-squared value analysis.
The variables and the factors show high probability, Erwin Widodo Erwin Widodo was granted his Master Degree
from Ritsumeikan University of Kyoto and his Doctoral
which is like the real-world system. In accordance
Degree from Hiroshima University both in Japan, in 2006
with the data calculation and optimisation, the sys­ and 2012, respectively. His expertise is of multi-player deci­
tem under discussion in the cooperative game sce­ sion-making in industry especially by using game theory. He
nario had higher financial performance than that in is lecturer from Quantitative Modeling and Industrial Policy
the non-cooperative scenario. Moreover, the carbon Analysis (QMIPA) in Department of Industrial Engineering,
emission of production was also higher than that in Institut Teknologi Sepuluh Nopember (ITS), Surabaya,
Indonesia. He actively joins several international associations
the non-cooperative scenario. This finding implies related to his field, i.e. IDGS, Informs, IAEng, IEOM and the
that coordination also has a major part in DCSC to likes. Moreover, he also joins as board member in several
maintain profit, and at the same time, may lower international journals such as IJISE, OSCM, IJTech, Cogent
carbon emission to the desired level. Engineering, etc. (ORCID ID: 0000-0002-4772-0465)
Moreover, the experiment shows that there were
some differences between those two scenarios, which
are the customer preference of each channel, the profit ORCID
estimation, and carbon emission production. However, Januardi Januardi http://orcid.org/0000-0003-0772-5939
the model shows that the differences in price combina­ Widodo E. http://orcid.org/0000-0002-4772-0465
tion in each channel affect the demand level of each
channel, especially the price in each channel, and cross-
price sensitivity. Sensitivity analysis shows that the coop­ References
erative scenario might accommodate one particular APJII. 2017. Penetrasi & Perilaku Pengguna Internet Indonesia.
channel to support another channel and gain more Jakarta: Asosiasi Pengguna Jasa Internet Indonesia.
demands than the non-cooperative scenario. That is the Besbes, K., H. Allaoui, G. Goncalves, T. Loukil, et al. 2013.
reason why the non-cooperative scenario has less profit “A Green Supply Chain Design with Product Life Cycle
Considerations.” Supply Chain Forum 14 (2): 18–25.
and carbon emission than the cooperative scenario.
doi:10.1080/16258312.2013.11517312.
Cooperative and non-cooperative scenarios are describ­ Campbell, S. D. 2007. “Uncertainty in the Great Moderation
ing the real business competition in oligopoly games. The Macroeconomic Volatility, Predictability, and Uncertainty in
study discovers that experimental design can be used for the Great Moderation : Evidence from the Survey of
supply chain system analysis. The study also shows that Professional Forecasters.” Journal of Business & Economic
price variations that customers consider can affect many Statistics 25 (2): 191–200. doi:10.1198/073500106000000558.
Elbert, R., C. Bogusch, and Ö. Özsucu. 2014. “Revenue Sharing
variables, such as profit and carbon emission. Moreover, as an Incentive for Increased Information Disclosure in
the study implies that price decision by the managers Outsourcing Relationships: An Experimental
must consider their customer behaviour. In this research, Investigation.” Supply Chain Forum 15 (3): 30–36.
the response surface was employed as the consideration. doi:10.1080/16258312.2014.11517349.
In the future, the research authors plan to continue Gahbiche, I. E., K. Hadjyoussef, Z. Jemai, A. Dogui. 2016.
“‘Analysis of a Stackelberg Game between a Customer
the uncertainty modelling by using an experimental
and Several Cooperating Suppliers: Stability and
design approach, especially RSM. This research, how­ Efficiency.” Supply Chain Forum 17 (2): 78–86. Taylor &
ever, has some shortages, such as the limitation on Francis. doi:10.1080/16258312.2016.1172826.
dealing with stochastic data. When stochastic data Gautier, F., P. Lacomme, P. Pariente, S. K. Tchomte,
have been incorporated, the discussion will be more N. Tchernev. 2013. “Linear Model for Supply Chain
interesting, and the real-world problem will be well Operational Planning and Carbon Footprint
Optimization.” Supply Chain Forum 14 (2): 40–53.
represented by using a probability distribution pattern. doi:10.1080/16258312.
2013.11517314.
Gelman, A., and G. Imbens. 2017. “Why High-order
Disclosure statement Polynomials Should Not Be Used in Regression
Discontinuity Designs.” Journal of Business & Economic
No potential conflict of interest was reported by the authors. Statistics 1–22. doi:10.1080/07350015.2017.1366909.
 SUPPLY CHAIN FORUM: AN INTERNATIONAL JOURNAL 11

He, Z., G.Han, T.C.E. Cheng, B. Fan, J. Dong. 2018. Demand.” Supply Chain Forum 18 (4): 218–230. Taylor &
“Evolutionary Food Quality and Location Strategies for Francis. doi:10.1080/16258312.2017.1380499.
Restaurants in Competitive Online-to-offline Food Modak, N. M., D. K. Ghosh, S. Panda, S. S. Sana, et al. 2018.
Ordering and Delivery Markets : An Agent-based “Managing Green House Gas Emission Cost and Pricing
Approach.” International Journal of Production Policies in a Two-echelon Supply Chain.” CIRP Journal of
Economics xx 215 (2018): 61–72. Elsevier Ltd. Manufacturing Science and Technology 20: 1–11. CIRP.
doi:10.1016/j.ijpe.2018.05.008. doi:10.1016/j.cirpj.2017.08.001.
Heydari, J., K. Govindan, and A. Aslani. 2018. “Pricing and Modak, N. M., and P. Kelle. 2019a. “Managing a Dual-channel
Greening Decisions in a Three-tier Dual Channel Supply Supply Chain under Price and Delivery-time Dependent
Chain.” International Journal of Production Economics 202 Stochastic Demand.” European Journal of Operational
(November): 1–12. Elsevier B.V.. doi:10.1016/j.ijpe. Research 272 (1): 147–161. Elsevier B.V.. doi:10.1016/j.
2018.11.012. ejor.2018.05.067.
Ji, J., Z. Zhang, and L. Yang. 2017. “Carbon Emission Modak, N. M., and P. Kelle. 2019b. “Using Social Work
Reduction Decisions In The Retail-/dual-channel Supply Donation as a Tool of Corporate Social Responsibility in a
Chain With Consumers ’ preference.” Journal of Cleaner Closed-loop Supply Chain considering Carbon Emissions
Production 141 (2017): 852–867. Elsevier Ltd. doi:10.1016/ Tax and Demand Uncertainty.” Journal of the Operational
j.jclepro.2016.09.135. Research Society: 1–17. Taylor & Francis. doi:10.1080/
Jo, S., and R. Sekkel. 2017. “Macroeconomic Uncertainty 01605682.2019.1654413.
through the Lens of Professional Forecasters Mukaka, M. M. 2012. “Statistics Corner : A Guide to
Macroeconomic Uncertainty through the Lens of Appropriate Use of Correlation Coefficient in Medical
Professional Forecasters.” Journal of Business & Economic Research.” Malawi Medical Journal 24 (3): 69–71.
Statistics 37 (3): 436–446. doi:10.1080/ Myers, R. H., D. C. Montgomery, and C. M. A. Cook. 2009.
07350015.2017.1356729. Response Surface Methodology: Process and Product
Kiefer, N. M., T. J. Kelly, and K. Burdett. 1994. “Menu Pricing : Optimization Using Designed Experiments. 3rd ed. Animal
An Experimental Approach Menu Pricing : An Experiments Genetics. New Jersey: John Wiley & Sons.
Approach.” Journal of Business & Economic Statistics 12 (3): Niranjan, T., P. Parthiban, K. Sundaram, P. N. Jeyaganesan.
329–337. doi:10.1080/07350015.1994.10524548. 2019. “Designing a Omnichannel Closed Loop Green
Kim, B., R. C. Blattberg, and P. E. Rossi. 1995. “Modeling the Supply Chain Network Adapting Preferences of Rational
Distribution of Price Sensitivity and Implications for Optimal Customers.” Sādhanā 44 (3): 1–10. Springer India.
Retail Pricing.” Journal of Business & Economic Statistics 13 (3): doi:10.1007/s12046-018-1038-0.
291–303. doi:10.1080/07350015.1995.10524603. Nuha, H., and E. Widodo. 2016. “Inventory Model in Dual
KLHK. 2017. Laporan Inventarisasi GRK Dan Monitoring, Channel Supply Chain considering Drop Shipping.”
Pelaporan, Dan Verifikasi 2017. Jakarta: Kementerian Proceedings of the International Conference of Logistic and
Lingkungan Hidup dan Kehutanan. Supply Chain Management System, 1–8. doi: 10.13140/
Kramer, A. 2016. “Using Experimental Survey Designs to RG.2.1.3423.0005.
Support Pricing Decisions.” Bussiness and Management Shil, N. C. 2010. “Customized Supplier Selection
Horizons 4 (1): 22–38. doi:10.5296/bmh.v4i1.9355. Methodology: An Application of Multiple Regression
Lang, G., and G. Bressolles. 2013. “Economic Performance and Analysis.” Supply Chain Forum: An International Journal 11
Customer Expectation in E-fulfillment Systems: A (2): 58–70. doi:10.1080/16258312.2010.11517232.
Multi-channel Retailer Perspective.” Supply Chain Forum Shin, M., and M. Zhong. 2018. “A New Approach to
14 (1): 16–26. doi:10.7137/scfij/14.1.16. Identifying the Real Effects of Uncertainty Shocks.”
Li, B., M. Zhu, Y. Jiang, Z. Li. 2016. “Pricing Policies of Journal of Business & Economic Statistics 38 (2): 367–379. .
a Competitive Dual-channel Green Supply Chain.” Journal doi:10.1080/07350015.2018.1506342.
of Cleaner Production 112 :2029–2042. doi:10.1016/j. Siegfried, R., S. G. LeBris, and M. Stewart. 2018. “Engineering
jclepro.2015.05.017. Students Ready for a Vuca World? a Design Based
Li, P. 2018. “Environmental Factor Analysis of Green Supply Research on Decisionship.” Proceedings of the 14th
Chain Based on Pricing Decision.” IOP Conference Series: International CDIO Conference, 872–881.
Materials Science and Engineering, 6th Annual International Simangunsong, E., L. C. Hendry, and M. Stevenson. 2012.
Conference on Material Science and Engineering “Supply-chain Uncertainty : A Review and Theoretical
(ICMSE2018), 397, 012098. Suzhou, Jiangsu, China. doi: Foundation for Future Research.” International Journal of
10.1088/1757-899X/397/1/012098. Production Research 50 (16): 4493–4523. doi:10.1080/
Liu, P. 2018. “Pricing and Coordination Strategies of Dual- 00207543.2011.613864.
Channel Green Supply Chain considering Products Green Sinha, S., and N. M. Modak. 2019. “An EPQ Model in the
Degree and Channel Environment Sustainability.” Perspective of Carbon Emission Reduction.” International
International Journal of Sustainable Engineering: 1–11. Journal of Mathematics in Operational Research 14 (3): 338.
Taylor & Francis. doi:10.1080/19397038.2018.1488891. doi:10.1504/IJMOR.2019.099382.
Mason-Jones, R., and D. R. Towill. 2000. “Coping with Widodo, E. 2015. “A Model Reflecting the Impact of Product
Uncertainty: Reducing “Bullwhip” Behaviour in Global Substitution in Dual- Channel Supply Chain Inventory
Supply Chains.” Supply Chain Forum: An International Policy.” Procedia Manufacturing 4 (2015): 168–175.
Journal 1 (1): 40–45. doi:10.1080/16258312.2000. Elsevier B.V.. doi:10.1016/j.promfg.2015.11.028.
11517070. Widodo, E. 2017. “Joint Decision of Pricing and Order
Mitra, S. 2014. “A Framework for Research on Green Supply Quantity by considering Product Substitution in Dual
Chain Management.” Supply Chain Forum 15 (1): 34–51. Channel Supply Chain.” AIP Conference Proceedings, 3rd
doi:10.1080/16258312.2014.11517332. International Materials, Industrial And Manufacturing
Modak, N. M. 2017. “Exploring Omni-channel Supply Chain Engineering Conference (MIMEC 2017) 6–8 December 2017,
under Price and Delivery Time Sensitive Stochastic 1–8. Miri, Malaysia. doi: 10.1063/1.5010622.
12 J. JANUARDI AND E. WIDODO

Wu, S., and H. S. Mortveit. 2015. “A General Framework for Yeniay, Ö. 2014. “Comparative Study of Algorithms for Response
Experimental Design, Uncertainty Quantification and Surface Optimization.” Mathematical and Computational
Sensitivity Analysis of Computer Simulation Models.” Application 19 (1): 93–104. doi:10.3390/mca19010093.
Proceedings of the 2015 Winter Simulation Conference Zhou, C. 2016. “Pricing Model for Dual Sales Channel with
(WSC), 1139–1150. Huntington Beach, California: IEEE. Promotion Effect.” Mathematical Problems in Engineering
doi: 10.1109/WSC.2015.7408240. 2016 (1804031): 1–20. doi:10.1155/2016/1804031.