How to evaluate sustainability of supply chains ? A dynamic network DEA approach

Vahid Shokri Kahi Department of Management and Accounting, Karaj Branch, Islamic Azad University, Karaj, Islamic Republic of Iran Saeed Yousefi Pegah Distribution Company, Karaj, Islamic Republic of Iran Hadi Shabanpour Karaj Branch, Islamic Azad University, Karaj, Islamic Republic of Iran, and Reza Farzipoor Saen Department of Industrial Management, Karaj Branch, Islamic Azad University, Karaj, Islamic Republic of Iran


Introduction
Supply chain is a purposeful combination of interconnected and interdependent organization.Although, units of supply chains are separated legally, they are connected by flow of material, information, and finance.Providing the best combination of resources via outsourcing is the main issue of supply chain management (SCM) in modern business enterprises (Ageron et al., 2012).Price, quality, flexibility, and supplier reputation are economic criteria in SCM evaluation (Bai and Sarkis, 2010).
Nowadays, sustainability factors play a critical role in long-term achievement of SCM; accordingly, purchasing process becomes more complicated with social and environmental pressures.As Govindan et al. (2013) addressed, over the past decades, due to rapid reduction of natural resources, concerns over wealth inequality and corporate social responsibility, sustainability has become important for researchers and scholars.In other words, there are pressures which force SCM to focus not only on economic but also on social and environmental criteria (Dyllick and Hockerts, 2002).Thus, sustainable supply chain may be of great importance attributable to surging environmental conservation and societal prosperity while reinforcing economic intention of organization.To do so, companies should conserve resources, optimize processes, uncover product innovations, save costs, increase productivity and promote corporate values by managing and improving environmental, social, and economic performances across supply chains (Seuring and Müller, 2008).On the other hand, there are several sustainability criteria which complicate SCM's evaluation.This means that decision-makers encounter some discretionary/free and even contradictory criteria while evaluating sustainability of SCM.Dual-role links, inputs, desirable and undesirable outputs are some of the main criteria.
In an accurate appraisal of supply chain, interactions among suppliers should be taken into consideration.To evaluate sustainability of SCM, dealing with multiple criteria has been one of the significant concerns in preceding models (Yousefi et al., 2016).To deal with multiple criteria, this paper develops a novel network-dynamic data envelopment analysis (DEA) model.Wong et al. developed two DEA models including technical efficiency and cost efficiency models to explain application of DEA in measuring supply chain performance.Traditional DEA models measure relative efficiency of decision-making units (DMUs) (Charnes et al., 1978;Tavassoli et al., 2014).
However, one of the most growing criticisms in these sorts of models is that the DMUs are assessed in a specific period.Given fluctuations in performance of DMUs during several periods, considering just a specific period is insufficient in comprehensive efficiency evaluation.In dynamic DEA presented by Tone and Tsutsui (2010), the DMUs' efficiencies are assessed during several periods.However, dynamic DEA does not deal with DMUs' internal structures.Network DEA (NDEA) deals with DMUs with internal structures (Yu and Lin, 2008).Tone and Tsutsui (2014) proposed network-dynamic DEA model in which the DMUs were assessed in multiple periods with internal interactions.Their model is based upon slack-based measure (SBM) model.
Objective of this paper is to improve the dynamic DEA model proposed by Tone and Tsutsui (2014).To this end, we develop a novel dynamic DEA model with network structure in which, for the first time, the overall carry-overs of a network in period (t) enter next period (t+1).Note that Tone and Tsutsui (2014) connected multiple stages in different periods.In Tone and Tsutsui (2014), the carry-overs of each stage inside period (t) enter the next period (t+1) separately and independently.However, in our proposed approach, the overall values of carry-overs for each network are obtained and enter the next period.In other words, Tone and Tsutsui (2014) connected the stages to each other (among periods) but we connect each network (as a whole) to other networks.Hence, as it is shown in Figure 1, we consider each period (t) as a network which enters next period (t+1) by carry-overs.Here, in addition, free links among periods are defined in terms of desirable and undesirable carry-overs.Figure 1 displays our novel network dynamic DEA structure.Note that all notations in Figure 1 are defined in Section 3.

A dynamic network DEA approach
Despite Tone and Tsutsui (2014), our proposed approach has following advantages: our approach connects the periods by networks' carry-overs and not stages' carry-overs; we define nature of free links and apply them in calculating relative efficiency of supply chains (DMUs); we develop an additive NDEA model for evaluating sustainability of supply chains in several periods.
There are numerous criticisms directed at the model proposed by Tone and Tsutsui (2014).In the aforementioned model multiple networks are simultaneously introduced as efficient units; therefore, the model fails in ranking the supply chains (DMUs).Second, the types of link among periods have dominant weakness.Changing an inefficient period to efficient period by the solution suggested by Tone and Tsutsui (2014), causes other periods to become inefficient.In other words, by increasing output of a period, inputs of next periods will be increased.This causes inefficiency of next periods.This happens due to poor definition of links in Tone and Tsutsui (2014) model.To address this weakness, we use different method.Besides, in our model the links among networks in periods are connected in an integrated way.
It is noteworthy that Tone and Tsutsui (2014) did not determine nature of the free links and also ignored the free links in calculation of DMUs' efficiency.In other words, in the approach proposed by Tone and Tsutsui (2014), decision maker cannot identify whether a free link is a desirable or undesirable factor.Note that a free link connects two periods and consists of desirable and undesirable types (Yousefi et al., 2016).Tone and Tsutsui (2014) did not clarify nature of free links.Hence, in this paper not only do we define the nature of free links but we also use all the links in assessing sustainability of supply chains (DMUs).In this paper, the links are classified into three categories, i.e. desirable, undesirable and free links which join multiple periods in series form.Note that our proposed network-dynamic DEA model is based upon additive model in which both networks and links have series structure and each period is a network.
Next sections of the paper are as follows: Literature review is presented in Section 2. In Section 3, proposed model is given.Case study is discussed in Section 4. Obtained results and concluding remarks are presented in Section 5.In Section 5.2, managerial implications are given.
2. Literature review 2.1 DEA and sustainable supply chain evaluation Organizations are responsible for their environmental and social performance (Zhu et al., 2005).Sustainable development emphasizes on environment, economy, and society.Evaluating sustainable supply chain management (SSCM) performance is a key challenge (Tseng et al., 2015).SSCM has turned out to be a strategic process which enables firms to create competitive advantage.Customers' demands and products' complexity have increased competition among organizations.SSCM provides an opportunity for organizations to distinguish itself from its competitors.Thus, it provides a competitive advantage in market (Khodakarami et al., 2015).Numerous organizations have already commenced to develop a certain level of commitment toward sustainability practices to make their supply chains sustainable (Govindan et al., 2015).SSCM evaluation, hence, is a substantial duty for any types of organizations.Among evaluation methods, DEA appears to be a suitable technique to assess SSCM.
DEA measures relative efficiency of DMUs (Charnes et al., 1978).DEA has been used in assessing supply chains (Liang et al., 2006).To measure efficiency of supply chains, Chen et al. (2006) developed a DEA-game model.Yu et al. (2010) assessed efficiency of supply chains by cross-efficiency DEA model without considering internal structure of supply chains.To measure performance of supply chains, Wong and Wong (2007) determined technical efficiency and cost efficiency of supply chains.Wong et al. (2008) developed a method for assessing efficiency of supply chains by stochastic DEA model.One of the DEA applications is to evaluate efficiency of two-stage processes where all outputs of the first stage are intermediate measures which are considered as inputs of the second stage.As a result, two-stage DEA models assess both overall efficiency score of whole process and each of the individual stages (Khodakarami et al., 2015).DEA, furthermore, has been used to evaluate environmental and social performance of various settings (Sueyoshi and Goto, 2010;Noorizadeh et al., 2011;Rashidi et al., 2015;Azadi et al., 2015).Sueyoshi and Goto (2010) used DEA to assess effectiveness of US clean air act in controlling CO 2 emission.Rashidi et al. (2015) applied DEA to estimate energy saving and undesirable output abatement of 25 countries.To assess sustainability of supply chains, Wu et al. (2016) provided both theoretical and quantitative tool.
Nevertheless, as previously mentioned, in classical models of DEA internal interactions are neglected.Given that sustainable supply chains encompass a complex network of connections and interactions, traditional DEA models have been evolved to overcome the shortcomings.

NDEA and dynamic DEA
To evaluate efficiency of a DMU, Färe andGrosskopf (1996, 2000) proposed an NDEA model with multiple production stages.Over the last decade, a couple of radial NDEA models have been presented (e.g.Lewis and Sexton, 2004;Färe et al., 2007;Kao and Hwang, 2010;Cook et al., 2010).Lewis and Sexton (2004) applied an NDEA model to evaluate organizational efficiency of companies with complex internal structure.Kao and Hwang (2010) utilized NDEA to measure efficiency of network systems taking into account IT impact on firm performance.Färe et al. (2007) developed an NDEA model by modeling data structures irregularities and structural complexities in DEA.Cook et al. (2010) selected the best supply chains by NDEA model.Liang et al. (2006) developed a non-linear model to determine efficiency of supply chains and their members.Chen and Yan (2011) presented a radial NDEA model to measure efficiency of supply chains.Yousefi et al. (2015) developed an NDEA model by goal programming to select the best supply chains.
For the first time, Sengupta (1995) introduced dynamic DEA model (D-DEA) and later it was extended based on the studies of Färe and Grosskopf (1996).They proposed a dynamic production frontier using an intermediate output which relates annual production processes.However, developing a new dynamic slacks-based measure (DSBM) model by Tone and Tsutsui (2010) was a turning point in the D-DEA literature.They suggested four types of carry-overs (links); desirable, undesirable, discretionary, and non-discretionary ( fixed) links to assess DMUs in different periods.The paper proposed by Tone and Tsutsui (2014) may be considered as one of the most comprehensive and reliable methods to evaluate sustainability of supply chains.They, for the first time, merged network SBM and DSBM models.By so doing, they proposed a dynamic DEA model with network structure.However, they did not specify some important items in their model.

1869
A dynamic network DEA approach

The proposed approach's necessity
In traditional DEA models as well as current dynamic DEA models, internal processes within the DMUs are not considered.In a comprehensive and realistic assessment of DMUs (networks), internal structures should be assessed.This means that all links among and inside the DMUs (networks) should be appropriately considered and defined.Unlike the real world, Tone and Tsutsui (2014) applied stages as periods and not networks and they ignored to explicitly define nature of free links in their model.In other words, they overlooked to determine whether free links are desirable or undesirable.Hence, they ignored to employ free links in calculating efficiency of DMUs.This paper intends to address these gaps.Indeed, very few papers so far have considered all the aforementioned criteria along with sustainability in supply chain using an optimization approach.For the first time, in this paper, a network-dynamic DEA model is proposed which is able to assess sustainability of supply chains in different periods.It should be mentioned that the connections among periods are as the links among networks and not stages.Besides, this model can assess sustainability of supply chains in multiple periods using additive model of network-dynamic DEA.In summary, contributions of this paper are as follows.
• for the first time, we present additive model of network-dynamic DEA; • we develop temporal link among networks based on series view; • we utilize the free links and determine their nature; and • for the first time, we outline the links in a way that carry-overs of networks are connected in different periods and not in different stages.

Proposed method 3.1 Algorithm
To introduce our proposed method, following steps are taken: In step 1, we provide data set as inputs, outputs, and links for each supply chain.In step 2, our dynamic NDEA model is run to evaluate sustainability of supply chains.In step 3, rank of supply chains is determined and the best sustainable supply chain is introduced.These steps are presented in Figure 2.
3.2 Definition of variables, network structure of supply chains, and dynamic DEA At this juncture, as is shown in Figure 3, the network model is defined.Here, The relevant multipliers (weights) of the above factors are defined as follows: • λ pr : multiplier of the output z p0 jrt exiting from stage p.
• β pk : multiplier of the output z pp þ 1 jkt which exits the stage p and enters stage p+1.
• α pi : multiplier of the input n 0 p jit entering the network at beginning of stage p from outside.Now, the efficiency of stage 1 in tth period is defined as follows.It should be noted that the first stage is the only stage in which all inputs enter the network from outside and not from prior stage: Numerator of Equation ( 1) expresses two terms, weighted sum of outputs that exit the network as final output and also weighted sum of output which exits from the first stage and enters the second stage as input.Besides, denominator indicates weighted sum of inputs that enter stage 1 from outside.Efficiencies of other stages are defined by the following equation.
Other stages can possess two types of inputs; that is inputs which enter the stage from outside and also the inputs that are outputs of prior stage and enter the stage p as input: Numerator of Equation ( 2) expresses the weighted sum of outputs which exit the network as final output and also weighted sum of outputs that exit the stage p and enter the stage p+1.Denominator indicates weighted sum of inputs that enter stage p+1 directly from stage p and weighted sum of inputs which enter the stage p from outside of network.Now, overall efficiency of jth DMU is defined in the following equation: Equation ( 3) indicates the overall efficiency of network.Numerator of Equation ( 3) expresses weighted sum of all outputs including the outputs which exit the network as final outputs; and the outputs which exit stage p but enter stage p+1 as input.Denominator indicates all inputs including inputs which go into stage p+1 after they leave stage p; and the inputs which enter the stage p from outside of network.

Proposed dynamic DEA model's structure
At this juncture, we represent our proposed dynamic DEA model which evaluates the supply chains in which the networks have series structure.As depicted in Figure 1, we classify the links into three categories.Thus, we define desirable (good) link, undesirable (bad) link, and free link (dual-role link) as typical carry-overs among periods.Let us now define following links: (1) Desirable (good) link: this link indicates desirable carry-over which connects tth period to t+1 period.In our model, we maximize this kind of link.
(2) Undesirable (bad) link: this link corresponds to undesirable carry-over which connects tth period to t+1 period.In our model, we minimize this kind of link.
(3) Free (discretionary) link: free link signifies discretionary carry-over which connects tth period to t+1 period.This type of link has dual-role nature.This means that free link can be stated desirable or undesirable.One of the results of the model presented in this paper is that it defines nature of this type of link.
Now, factors and their weights are defined as follows: shows undesirable (bad) link of jth DMU in tth period (t ¼ 1, …, T) which enters into period t+1.
• ϑ tl is multiplier of desirable carry-over related to the free link c free jlt indicating degree of desirability in free link.
• ω tl is multiplier of undesirable carry-over related to the free link c free jlt indicating degree of undesirability in free link.
• γ tl is multiplier of desirable link c d jlt .
• ν tl is multiplier of undesirable link c und jlt .
Given the above discussions, overall efficiency of jth network under evaluation in tth period is defined as following objective function: Equation ( 4) represents jth network's efficiency ratio.In this equation, weighted values of the link are taken into account which connects the period t to period t+1 in jth network.
We can distinguish F t j as two merged fractions consisting of the fraction in the second part indicating Ρ j (Equation ( 3)) in which both numerator and denominator have parentheses and fraction without parentheses in the first part of Equation ( 4) which demonstrates weighted ratio of desirable links to undesirable ones.We wish to determine free links' nature in terms of being desirable or undesirable.Accordingly, we introduce the links with two multipliers.Each of these multipliers indicates whether the free links are desirable or undesirable factors.Here, we set sum of these multipliers to 1 in which multipliers signify degree of desirability or undesirability of free links: Hence, in general, ratio of links for networks in different periods is defined as follows: To optimize overall efficiency Φ, a multi-stage process is needed as restrictions ρ pj , ρ 1 , and ϕ should not be more than 1.At this juncture, given definitions and formulas, overall non-linear model is presented as follows: Max Formulas (2), ( 4), ( 5), and ( 6) are incorporated into model ( 7).Non-linear model (7) includes formula (4) in the objective function.Formula ( 2) is used as the first constraint and formula ( 5) and ( 6) are applied in second constraint.Model ( 7) is a non-linear model which should be converted to a linear model.Charnes-Cooper approach may be utilized to convert the

1873
A dynamic network DEA approach non-linear models to linear models (Charnes et al., 1978).The o in our proposed model ( 8) implies the DMU under evaluation: g tl X0; W tl X0; l pr X0; b pk X 0; n tl X0; o tl X 0; a pi X 0: The objective function in model ( 8) indicates DMU's efficiency score in the period under evaluation.Furthermore, to illustrate weighted average of network efficiency during several periods, so the following expression is presented: In next section, a case study is given.(1) Kadbanoo Co. with more than 61 years of experience in food industry does the production and outsourcing of Delpazir products encompassing more than 50 stock keeping units including mayonnaise and salad dressings, canned food, tea, tomato paste, and ketchup.

Case study
(2) Imports and exports are executed by Akam Co. Akam exports various products to neighbor countries.
(3) Sales and distribution is performed by Pegah Co.With 29 years of experience, Pegah Co. is operating with more than 150 SKU's and 4 major brands in fast moving consumer goods industry.Pakhsh Pegah was established in 1983 and is responsible for distribution, sales and marketing of domestic and foreign products across Iran.This company distributes products through 14 distribution centers across Iran and has always been a pioneer in distribution industry.
In this paper, sustainability of 17 supply chains is assessed during 2012 to 2014.To assess sustainability of supply chains, economic, environmental, and social factors are taken into account.Each supply chain has three stages including farms, canning factories, and chain stores.Inputs and outputs of supply chains given three stages are defined below: (1) Inputs and outputs of farms: • Inputs: -Environmental costs ($1,000) as an environmental criterion which represents amount of money that is paid for environmental damage in a year.-Cost of labor safety ($1,000) as a social criterion which represents amount of money that is for compensating labor damages in a year.-Other costs (economic criterion).

Outputs:
-Revenue from sales of tomatoes as raw material which are sent to tomato-paste factory ($1,000) as an economic criterion.-Revenue from sales of raw tomatoes in markets ($1,000) as the other economic criterion.
(2) Inputs and outputs of canning factories: • Inputs: -Cost of purchasing tomato from farm ($1,000) as an economic criterion.
-Environmental costs ($1,000) as an environmental criterion which is amount of money that is paid for environmental damage in a year.-Cost of labor safety ($1,000) as a social criterion which is amount of money that is paid for compensating labor damage in a year.
• Outputs: -Volume of tomato pastes which are ready for sales (economic criterion) in a year.
(3) Inputs and outputs of chain stores: • Inputs: -Volume of tomato pastes which are ready for sales (economic criterion) in a year.

1875
A dynamic network DEA approach -Cost of labor safety ($1,000) as a social criterion which represents amount of money that is paid for compensating labor damage during a year.
Degree of customer satisfaction (as a social criterion) which is measured by nine-point Likert scale.Table I shows Likert scale (Likert, 1932).Delpazir Co. utilizes opinions of customers to measure customer satisfaction.
Given the aforementioned definitions of inputs, outputs, and links, structure of supply chains is displayed in Figure 4.
Furthermore, as is shown in Figure 5, each period has three links to next period which are as follows: (1) Not collected revenue (desirable link).
(3) Green research and development (green R&D) ( free/dual-role link).In Delpazir Company, green R&D costs are paid for elimination of chemical and non-organic processes from tomato supply chain.By running model ( 7), the supply chains are assessed over three years.Efficiency scores of 17 supply chains in three years are shown in Table III.Besides, nature of the dual-role links is identified in all supply chains.According to Table III, if the ϑ becomes 1, the dual-role link has desirable nature.If the ϖ becomes 1, the link has undesirable nature.However, some sustainable supply chains may seem superior due to excellent efficiency scores only in a specific period.For example, in 2013, supply chains 15 and 16 are more sustainable than other supply chains.But, given overall efficiency, they are ranked as 7th and 8th, respectively.Though supply chains 17, 13, 9, and 4 are not sustainable in any period, they have received a better rank compared with supply chains 15 and 16.Given overall efficiency, the supply chain 5 is the most sustainable supply chain during all periods.

Obtained results
The traditional DEA models do not explicitly ascertain any of key sub-processes involved in divisions of supply chains.Furthermore, one of the drawbacks of these models is oversight of internal structure of supply chains (DMUs).As an illustration, numerous companies embrace several divisions that are linked to each other having division-specific inputs and outputs along with links to other divisions.It is recognized that supply chains (DMUs) may have multi-stage structures where initial stage utilizes inputs to produce outputs that become inputs of the second stage and the second stage employs the first stage outputs to produce its own outputs.On the other hand, there exist several sustainability criteria complicating SCM's evaluation.Accordingly, these make decision-makers encounter some discretionary/free and even contradictory criteria while evaluating sustainability of SCM.Dual-role links, inputs, desirable and undesirable outputs are some of main criteria.
To deal with multiple criteria, we proposed a novel method to consider interactions among supply chains and overall efficiencies in network-dynamic DEA.This enables decision makers not only to acquire the overall efficiency of supply chains (DMUs) over entire periods, but also to observe dynamic change of period efficiency and dynamic change of divisional efficiency of DMUs.Our approach excellently overcomes deficiencies of previous methods.We have also extended and improved Tone and Tsutsui (2014) dynamic DEA model.In our proposed approach, overall values of carry-overs for each network are obtained and afterwards are entered to succeeding period.In other words, Tone and Tsutsui (2014) connected the stages to each other (among periods) but we connect each network (as a whole) to other networks.Here, in addition, free links among periods are defined in terms of desirable and undesirable carry-overs.Numerical illustrations have been provided in a case study to demonstrate  208,500 225,200 200,500 231,000 285,300 223,800 212,400 202,600 206,500 215,600 209,600 210,500 202,100 207,000 226,900 225,400 217,800 Chain stores Inputs Tomato pastes ready to use 208,500 225,200 200,500 231,000 285,300 223,800 212,400 202,600 206,500 215,600 209,600 210,500 202,100 207,000 226,900 225,400 217,800 Shipping cost  16,400 12,100 11,800 11,900 21,500 12,800 14,200 13,100 11,000 12,400 12,300 10,900 11,200 11,600 10,900 12,400 11,500 Degree of customer satisfaction    200,900 206,900 223,600 222,000 203,100 203,800 208,800 205,000 206,900 207,300 203,200 204,900 219,900 204,400 203,700 245,200 237,400 Chain stores Inputs Tomato pastes ready to use 200,900 206,900 223,600 222,000 203,100 203,800 208,800 205,000 206,900 207,300 203,200 204,900 219,900 204,400 203,700 245,200 214,300 209,300 202,700 206,100 229,500 203,500 248,900 205,800 200,300 205,800 206,000 202,400 203,700 223,500 209,000 202,500 204,200 (continued )  applicability of the proposed approach.We examined performance of seventeen supply chains of Delpazir Company in 17 provinces of Iran in 2012-2014.
Case study was run to evaluate performance of supply chains in several periods.Results of analysis ( feedback) are put into practice by decision makers.The most sustainable supply chains are benchmarked by other supply chains.Based on the efficiency scores of Table III, the supply chain No. 5 is ranked 1st, 6th, and 4th in 2012, 2013, and 2014, respectively.Furthermore, the supply chain No. 5 is ranked as the most sustainable supply chain with overall efficiency score of 0.95.Similarly, the supply chains Nos 7 and 17 are ranked as 2nd and 3rd, respectively.The reason why the supply chain No. 5 is superior in ranking is based on the fact that its efficiency score in all years is at a high and acceptable level.Therefore, the overall efficiency of this supply chain is the highest in comparison with the other supply chains.It is noteworthy to state that if we wish to rank the supply chains only based on the efficiency scores in 2014, the supply chain No. 5 would be ranked 4th and in 2013 would be ranked as 6th.The supply chains Nos 15 and 16 are introduced as superior supply chains in 2013, but since this superiority is not sustainable in the years before and after, these supply chains cannot get an overall rank better than 7th and 8th, respectively (Figure 6).

Managerial implications
Note that the main noticeable outcome of this study is to present a more reliable and realistic model as a novel integrated network dynamic DEA framework for dealing with some shortcomings of the previous models in assessing SSCM.Our proposed method offers a means for managers to monitor and measure efficiency of their production processes.
In recent years, SCM has increasingly become of great importance.Process of evaluation and selection of SCM is one of the most dominant tasks of managers in field of sustainable supply chain.Since decision makers should take into account economic, social, and environmental criteria, these criteria should be identified by decision makers.A variety of mathematical modeling approaches have been utilized in assessing sustainability of supply chains.Mathematical models convert data into information to facilitate strategic and operational decisions.All former approaches, except for DEA, rely on subjective weights of criteria determined by managers.The approach presented in this paper evaluates sustainability of supply chains over multiple periods.
In addition to assessing sustainability of supply chains, in this paper, our model can propose improvement solutions for each inefficient network in any period.Moreover, using the proposed model, weaknesses in each network and in its stages are specified.In Table IV, Figure 1.Typical structure of dynamic model with network structure

Start
Figure 2. Steps of the proposed framework Figure 5. Topology of links among periods Figure 6.Supply chains' efficiency score Delpazir Company is a private holding company in food industry which was established in 1980 in Iran.The company's main products include tomato paste and tomato sauce.
Delpazir Company has 17 branches in different provinces of Iran.Each branch has a separate supply chain which produces tomato paste.The supply chains are assessed and controlled under supervision of Delpazir Company holding.Delpazir group has the most advanced and up-to-date food processing factories across the Middle East enjoying the most advanced European machinery with top technology in food industry.Delpazir group is comprised of three major sectors as follows:

Table I .
Historical data of 17 supply chains from 2012 to 2014 are reported in Table II.The data set is obtained through recorded documents of Delpazir Company as well as the company's executives.Note that there are 17 supply chains of Delpazir Company in 17 provinces of Iran.They operate separately and independently while being directed by a central ownership.Likert scale for measuring customer satisfaction