Advertisement

Requirements for Relief Distribution Decision-Making in Humanitarian Logistics

  • Mohammad Tafiqur RahmanEmail author
  • Tim A. Majchrzak
Conference paper
  • 2 Downloads
Part of the Lecture Notes in Information Systems and Organisation book series (LNISO, volume 39)

Abstract

Making efficient and effective decisions in the chaotic environment of humanitarian relief distribution (HRD) is challenging. Decision-makers need to concentrate on numerous decision factors categorized into decision objectives, variables, and constraints. Recent HRD literature focuses on optimizing procedures while neglecting the quantification of essential requirements (decision factors) for information systems to provide decision-making support. In this article, we address this gap by accumulating affecting decision factors from both literature and practice. We investigated the practical implications of these factors in HRD decision-making by measuring the preferences of a Delphi panel consisting of 23 humanitarian experts. The results from our study emphasize the importance of the decision factors in the proposed process model for HRD in a large-scale sudden onset. Our work provides researchers not only with a comprehensive set of practically feasible decision factors in HRD but also with an understanding of their influences and correlations.

Keywords

Natural disasters Decision support system Decision factors Relief distribution Humanitarian logistics Delphi technique Expert preferences 

1 Introduction

Although saving lives is the main aim of humanitarian relief operations, it is important to concentrate on minimizing social tension, which increases due to imbalance (inefficiency) in relief distribution (RD). For example, if two distribution centers distribute different relief items, it may fuel tension among recipients depending on which center serves them. Hence, responders need to prepare to standardize relief packages by coordinating with other responding groups and communicate with the recipients to disseminate an RD plan and during the duration of response operations. However, to meet beneficiaries’ necessities, responders must know what the demanded items are, and where and when they are needed. For rapid, effective, and efficient response, they also require knowing the accessibility (to transport relief items), warehousing (to store them), and distributing arrangements (to reduce social tension) [1]. Moreover, for successful relief operations, understanding and assessing the overall disaster situation (e.g., environment, vulnerabilities, coping mechanisms) is necessary. Thus, responders must acquire geographical, topographical, and demographical knowledge before scheduling RD operations  [7].

Identifying such influential decision factors in emergency management – especially in RD – is a complex task [47]. The humanitarian logistics (HumLog) literature proposes plenty of mathematical models and objective functions development by focusing on specific disasters as cases. Researchers utilized diverse variables and constraints in their models and functions for achieving targeted objectives. These factors need to be properly managed and utilized for rapid and effective decision making as they influence the success of the operation [46]. Failure to understand their importance for the information system will make the decision-making process more complex and time-consuming, causing delayed and inadequate responses – or potentially an overall unsuccessful relief operation [29].

By following the work of MacCarthy and Atthirawong [15], Okoli and Pawlowski  [26], and Richardson, de Leeuw and Dullaert [34], we rigorously and systematically reviewed and analyzed humanitarian literature to develop a summarized list of decision factors for relief distribution. While sharing some common decision factors (objectives, variables, constraints), the review denoted that five other problem areas (DPA) influence RD decision making: facility location (FL), inventory management (IM), relief supply chain (RSC), transportation (Transp), and scheduling (Sched). For achieving better performance in the complex decision-making operation, decision-makers (DM) in RD need to concentrate on shared decision factors as well and assist DMs in other DPAs to achieve their objectives.

However, there has been no structured attempt in RD to identify comprehensive factors and their correlations systematically as well as to prioritize them. This study addresses this gap by empirically testing decision support requirements with the help of the Delphi technique. A worldwide Delphi panel was formed with experts from academia, governments, and national and international NGOs. Their evaluations facilitated consensus and prioritization for each factor and assisted us in answering the following research question: What decision factors do experts prefer for effective humanitarian relief distribution decision-making?

To answer this research question, we need to identify experts’ preferences in the literature- and field-based decision factors. This investigation will assist us in finding the essential decision factors and understanding their correlations while decision-making for relief distribution. The remainder of this article is organized as follows. We provide the research background in Sect. 2. Section 3 describes our research design. Section 4 presents the results from the Delphi study, and Sect. 5 synthesizes and discusses the findings. We subsequently notify the limitations to this research and suggest implications for future research. Section 6 concludes the article.

2 Research Background

To respond to disasters in a chaotic environment, practitioners conduct complex and challenging tasks. While making decisions on RD, they face uncertainty when identifying appropriate decision factors. Not much research concentrates on recognizing factors that influence decision making in relief distribution. Peres et al. [27] classify operational research (e.g., RD) in HumLog into three DPAs (FL, IM, and network flow and Sched) without presenting influential decision factors. Gralla et al.  [12] and Gutjahr and Nolz [14] respectively categorized and refined humanitarian aid operations into efficiency (refined into cost efficiency), effectiveness (refined into response time, travel distance, coverage, reliability, and security), and equity criteria. This classification, categorization, and refinement led towards identifying affecting decision factors and developing a comprehensive set of them. Although Roy et al.  [37] listed some factors by dividing the RD process into four sub-processes (FL, IM, Transp, and RD decision), it was not investigated in detail to guide researchers on selecting decision variables and constraints for achieving targeted decision objectives. Safeer et al. [38] and Özdamar and Ertem [47] mapped constraints for specific objectives mainly for transportation and relief distribution but lacked a comprehensive set of decision factors, their priorities, and correlations. We know no research investigating the influences of other DPAs on the decision factors of RD.

However, to improve the disaster management process, adequate decision-making is the key, where prioritized and correlated decision factors play vital roles [4, 22, 43]. According to Li et al. [22], influential factors and their relationships need to be accumulated through proper investigation and experts’ judgment. Instead of studying the entire system, current research mostly concentrates on optimizing certain procedures that are extensively case-specific and are rarely used (or unusable) in other cases. To get a holistic image, we accumulated the existing decision support models for humanitarian operations that were implemented in practice in the contexts of sudden natural disasters, thereby collecting practical decision factors. The decision factors accumulated from academic literature are evaluated and utilized in this article to develop a practice-oriented RD process model (Table 1).
Table 1.

Literature-based decision factors for relief distribution decision-making

Categories

Decision factors

Literature

10 decision objectives

maximize coverage (cov), maximize transport quantity (tq), minimize travel time (tt), minimize distribution time (dt), minimize travel distance (td), minimize total cost (tc), minimize resource cost (rc), minimize penalty cost (pc), minimize number of distribution centers (ndc), minimize practical length of emergency route (pler)

[5, 6, 12, 23, 32, 33, 35, 42]

13 decision variables

travel distance (td), inventory flow and capacity (ifc), penalty cost (pc), transport cost (trc), operational cost (oc), set-up cost (stc), supply unit (su), beneficiaries access cost (bac), transport quantity (tq), demand time (det), travel time (tt), distribution time (dt), resource need (rn)

12 decision constraints

storehouse capacity (shc), road capacity (roc), inventory holding cost (ihc), number of storehouses (nsh), budget availability (ba), demand satisfaction (ds), replenishment cost (repc), load flow (lf), transport cost (trc), travel distance (td), operational cost (oc), resource availability (ra)

3 Research Methodology

3.1 Method Selection

Several techniques were advocated in the humanitarian literature for decision making in different problem areas. We used the Delphi technique to evaluate these factors and to identify new ones. It is suitable for this kind of exploratory research where researchers need to communicate with distantly located practitioners and field experts for dealing with complex and indispensable issues [24, 34]. Although the Delphi technique was successfully utilized by MacCarthy and Atthirawong [15] for investigating and understanding decision factors, it was not widely exploited in humanitarian research. Cottam et al. [8] used the Delphi technique to assess the potential benefit of outsourcing the trucking activities for relief distribution in developing countries. Richardson et al. [34] investigated affecting factors for global inventory prepositioning locations.

The Delphi technique provides an unbiased rating of the decision factors, which further go through ranking and consensus phases for identifying the importance and acceptance of each factor for effective decision-making in disaster-like uncertain situations [17]. Figure 1 illustrates the procedure for our Delphi study, including panel formation and research design.
Fig. 1.

The process model utilized in this Delphi Study

(inspired by [21, 25])

We invited 76 out of 96 identified experts to participate in the survey. The questionnaire for the first Delphi round was sent to them to confirm their participation. 38 experts replied positively, and 23 finally participated in the survey (i.e., formed the Delphi panel). 17 of the 23 participants completed and returned the questionnaire, the others preferring audio-recorded interviews. We sent the questionnaire for the second round to the 17 who answered the questionnaire experts, of whom 13 responded. The panelists were anonymized according to the guidance of the Norwegian Center for Research Data (www.nsd.no) and the participants themselves. Hence, while conceptualizing panelists’ thoughts in Sect. 5, we refer to them with their assigned participant identification numbers (PID in the form of P#).

3.2 Delphi Panel Formation

Due to their recency and severity, we targeted the Indonesia earthquakes of 2018 and the Nepal earthquake of 2015. While searching for involved experts having knowledge and interest in RD processes, we established contact with active practitioners and with their networks to gain updated knowledge on their usage of information systems (IS) for relief distribution. Besides, we utilized our contacts and the snowballing technique to bring more experts on-board. As a tentative list of potential participants was ready, we sent a study plan including information on the aim of the Delphi and its rounds, the extent and timing of the expected involvement, expected outcomes, and the potential social benefit to the ones who replied affirmatively. Finally, 23 experts from around the world participated in this Delphi study. The participating experts are listed in Table 2, along with their PID, medium of participation, affiliated organizations, countries, and contributed disasters. With an adequate panel size, according to Grim and Wright [13] and Okoli and Pawlowski  [26], we proceeded to the next step. The first-round survey questionnaire was electronically distributed, along with a consent form and a non-disclosure agreement.
Table 2.

The Delphi panel

3.3 Research Design

Data Collection Method.

Instead of starting the process with an open-ended questionnaire or brainstorming sessions, as in traditional Delphi, to identify decision factors in RD [34], we approached participants with existing academic knowledge on such factors. These factors were accumulated, summarized, and clustered into three categories (decision objective, variable, and constraints). We then added them to the questionnaire for experts’ evaluation. The factors were explained in the questionnaire that facilitated respondents to rate each decision factor on a six-point Likert Scale (inspired by [40]). Respondents were also given space to express their understanding of each of the factors and propose new factors from the practical field. However, if a participant found it complicated to answer the questionnaire, they had the opportunity to express their opinion through interview sessions (physical or online). As a result, we gained qualitative insights for the entire RD process (inspired by [44]). Additionally, to understand the depth of influences, participants were requested to mark the relationship of each decision factor of RD to the other five problem areas (FL, IM, RSC, Transp, and Sched). Thus, we incorporated relevant and in-depth information for the research quest (inspired by [18]).

Consensus and Stability.

To decide on achieving consensus, we adopted the Average Point of Majority Opinions (APMO) technique by Kapoor [20]. A decision factor would be considered as achieving consensus if its agreement or disagreement is above the cut-off rate of APMO. Instead of considering consensus achievement as a tool to decide on further Delphi rounds, we verified how a certain percentage of votes fall within a prescribed range, i.e., how the experts react to different decision factors. We identified no clear instruction on deciding on the number of Delphi rounds for studies. Hence, by following Dajani and Sincoff [9] and Strasser [41], we calculated the coefficient of variance (CV) to decide Delphi rounds and check their consistencies. Finally, we utilize SPSS software to calculated Kendall’s concordance coefficient (W) to measure the degree of agreement among panel members (W = 0 means perfect disagreement and W = 1 means perfect agreement). W = 0.7 is considered as an indication to achieve a higher level of general agreement in Delphi studies [39]. We demonstrate consensus and stability in Sect. 4 and discuss them in Sect. 5.

Delphi Rounds.

After finalizing the list of experts, we started round-1 by commencing the Delphi process by sending the questionnaire to each panel member in December 2018. Although an online survey is a typical mode for the Delphi technique [34, 40], emailing the questionnaire – e-Delphi – is also practical [2, 3, 25]. In addition to survey questions, the questionnaire captured the professional background for each respondent. We collected responses until February 2019. Data accumulated from the first round of the Delphi survey were extracted for descriptive analysis for finding frequencies and percentages. We utilized MS Excel and IBM SPSS to find correlations among factors and different statistics, such as the mean rank and Kendall’s W. Furthermore, we utilized APMO to determine whether each factor achieved consensus. In round-2, the result generated from the collective feedback in the first Delphi round was shared with all the panel members in March 2019. We redesigned the questionnaire to inform about the average rating, percentage of agreement and disagreement, overall ranking, and achieving consensus for each decision factor. The respondents were provided with their previous rating for each of the decision factors and allowed to update it (inspired by [36]). The newly identified practical decision factors from round-1 were added to the questionnaire to be evaluated.

4 Results

4.1 Descriptive Information on the Participants

Most panel members have extensive working experience, some of whom worked for more than 25 years in this sector. They participated or are participating in the response operations for large-scale natural and human-made disasters worldwide, for example, the South Iceland earthquakes 2000 and 2008, the Haiti earthquake 2010, the Gorkha (Nepal) earthquake 2015, the Indonesia earthquake 2018, different devastating hurricanes and floods, the 2014–2016 Ebola outbreak in West Africa, and the Syria crisis. Their heterogeneous experiences on responding to various crises and disasters assist us in evaluating the influential decision factors.

4.2 Measurement of Stability and Stopping Criterion

To achieve stability and to stop further rounding, English and Kernan [11] quantified 0 < CV ≤ 0.5. In the first Delphi round, we had four factors in three decision-making categories (one in decision objectives and constraints, and two in decision variables) that were in the border or out of the suggested range of achieving general agreement (CV ≥ 0.5). Besides, Kendall’s W value for each category was very low (for objectives W = 0.181, for variables W = 0.133, and for constraints W = 0.26). Therefore, we conducted the second round, where the four decision factors achieved a good degree of consensus with CV ≤ 0.39. Then, we measured the CV difference and defined the stopping rule as a CV difference of ≤ 0.3 (inspired by [41]). However, there were significant improvements (although still not high) in the degree of agreement in all categories in the second Delphi round (for objectives W = 0.194, for variables W = 0.213, and for constraints W = 0.470). Finally, receiving an absolute CV difference of ≤0.26 for each factor in every decision-making category and improved value for Kendall’s W constituted stability, we decided to terminate conducting any additional Delphi round (inspired by [9, 10]).

4.3 Results of the Delphi Rounds

Literature-Based Decision Factors.

Table 3 demonstrates the combined statistical results for two Delphi rounds. It illustrates the consensus and ranking for each decision factor incorporated into three categories for relief distribution (decision objectives, variables, and constraints). We easily compare the responses in both rounds and visualize the changes made by the respondents in the second round. For convincingly presenting the result, we clustered decision factors up to the third level of importance: achieving an average rating (AR) of ≥5.00 was considered as highly important decision-making factor and placed in cluster-1, whereas factors satisfying 5.00 > AR ≥ 4.00 were considered in cluster-2 as mediocre and the rest with AR < 4.00 were in cluster-3 as least affecting factors.
Table 3.

Combined statistical results for Delphi rounds 1 and 2 (inspired by [8, 41])

Decision Objective.

In Delphi round-1, 76.8% of the experts rated all listed decision objectives as important topics in the relief distribution decision-making, whereas 19.6% found them unimportant, and 3.6% abstained from commenting. Among those decision objectives, travel time minimization and coverage maximization were placed in cluster-1 as the most important objectives that responders try to achieve without considering minimizing different costs (total, resource, penalty) and the number of distribution centers, hence placed in cluster-3. The mediocre category (cluster-2) encompassed factors that were mostly related to transportation and distribution. The result suggested transporting a maximum quantity of relief items by choosing a practically short emergency route that would minimize travel distance and distribution time. In Delphi round-2, 78.5% of experts voted as important properties of decision making, and 21.5% voted not to consider. However, a significant change was observed in this round, where coverage maximization was downgraded, and all the topics from cluster-3 were upgraded to cluster-2. The only topic remained in cluster-3 was resource cost minimization.

Inspecting the consensus, we can identify that transport quantity from cluster-2 and all the topics in cluster-3 did not receive general agreement from the participants in the first Delphi round. However, they continued not to receive consensus in the second Delphi round as well, but for the topic of transport quantity. Its AR was upgraded to 4.8 and secured its consensus with 92.3% vote in round-2. Except for the down-graded topic of travel distance, all topics in cluster-1 and -2 gained their votes to be importantly considered in the relief distribution decision-making process. Finally, the voting for total cost was unstable (as CV > 0.5) in round-1 and achieved its stability in round-2.

Decision Variables.

To find important decision-making variables in round-1, 74.8% of panel members positively rated the factors in this category, whereas 21.3% finds them unimportant, and 3.9% did not vote. In round-2, 81.1% voted to list them as important decision factors. However, by analyzing the voting result, we identified that resource need was placed in cluster-1 in both rounds, whereas the transporting quantity of relief items accompanied it in round-2. All costing-related topics (penalty, transportation, operational, and set-up) secured their places in cluster-3 in round-1, except beneficiaries’ access cost. It was listed in cluster-2 along with travel distance, inventory flow and capacity, supply unit, transportation quantity, and demand, travel, and distribution time. There was no such significant change in round-2. Operational and set-up cost upgraded to cluster-2, and as already mentioned, transportation quantity joined resource need in cluster-1. Although travel distance was a mediocre affecting decision factor, it did not achieve general agreement along with all factors from cluster-3 in the first round. However, all the non-consensus factors in the first round remained unchanged in the second round, except beneficiaries’ access cost. It secured its consensus with 84.6% of the general agreement in the final round. Lastly, the rating for penalty cost and transportation cost were unstable (as CV > 0.5) in round-1 that became stable in round-2.

Decision Constraints.

The decision factors in this category already achieved stability as CV < 0.5 for each of them in Delphi round-1, and this stability became higher in round-2 as CV ≤ 0.29. However, the analysis found no highly important decision factor for cluster-1 in the first round. Seven out of 12 decision-making constraints were considered as mediocre and placed in cluster-2, while the remaining ones were encompassed in cluster-3. The factors constituted this category gained their maximum percentage of general agreement in round-1, which remained the same in round-2 as road capacity, and the number of storehouses switched their places in achieving consensus. However, five decision constraints (storehouse capacity, budget availability, demand satisfaction, travel distance, and resource availability) from cluster-2 gained higher importance in the second round and moved to cluster-1, which was the maximum content of this cluster. 72.3% of the panel members agreed to consider the listed factors as important decision-making constraints in round-1, whereas 23.1% were not convinced, and 4.6% were unable to comment. In round-2, 78.8% voted for enlisting these factors as decision-making constraints in the intended decision support system (DSS), whereas 21.2% voted not to.

Field-Based Decision Factors.

While evaluating the decision factors in round-1, the panelists were requested to recommend essential factors missing so far. Out of 23 panelists, 13 contributed to suggesting additional decision factors based on their experiences. After analyzing and refining, three new decision objectives were identified, whereas 13 new decision variables and ten new decision constraints were enlisted for further evaluation in round-2. The panel members were requested to follow similar evaluating procedures as that of in the first round. This evaluation procedure facilitated panel members with a chance to know and verify the new decision factors proposed by other members. Table 4 demonstrates the newly recommended decision factors, along with the results from the analysis that is subsequently discussed.
Table 4.

Field-based decision factors for relief distribution decision-making

After analysis, we identified that two decision objectives, six decision variables, and three decision constraints achieved consensus with over 90% vote and hence, prioritized into the list though their mean rank is lower (please consult Table 4 for detail). Although other decision factors did not achieve consensus, their importance in the decision-making process was significant as they scored over 76% vote. For example, what would be the reason for tending to reduce central control on the financial flow and other decisions? Again, from the general observation, it can be understood that most of the panelists suggested having decision-making flexibility in the field, but we do yet not know the actual reasoning. If the rating for this attribute is analyzed, it seems that participants were, somehow, confused to claim that flexibility because 8 out of 10 agreeing participants rated 4 out of 6, whereas 2 out of 3 disagreeing participants rated 3 and in the Likert scale of 6, rating of 3 and 4 are normally meant as confusing. Thus, further study is essential. Additionally, based on experts’ ranking, these new decision attributes were finally ranked by using the mean rank calculated by SPSS software. Furthermore, the degree of agreement among panel members (Kendall’s W) was also measured. The Kendall’s W for new decision objectives was measured as 0.644; 0.7 is considered as a higher level of general agreement. Hence, it was decided to conclude the Delphi survey though Kendall’s W for the other two categories were not that high – for decision variables W = 0.184 and for decision constraints W = 0.166.

Final Ranking.

Over 76% of the panelists in round-1, voted to include all the literature-based decision factors in the relief distribution decision-making; over 81% voted this way in round-2. The field-based decision factors received an overall vote of over 77% to accept them in the decision-making process. Thus, the importance of these comprehensive decision factors in the envisioned DSS for relief distribution was accomplished. Therefore, combinedly the number of enlisted decision factors became large: 13 decision objectives, 26 decision variables, and 22 decision constraints. This bigger list of decision factors is impractical to suggest to the decision-makers and will be challenging to manage in the crucial responding time. Hence, a comprehensive and prioritized list of decision factors is needed.

We further analyzed the consensus regarding highly ranked decision factors in both lists. To be enlisted as highly influential decision-making attributes, literature-based decision factors must accomplish over 80% vote in both Delphi rounds, whereas decision factors from the practice must ensure over 90% vote in round-2. We, thus, identified and enlisted six decision objectives, eight decision objectives, and eight decision constraints as the most influential decision factors for humanitarian relief distribution. Table 5 presents a comprehensive list of top-ranked decision factors, along with the vote percentage. To present them conveniently, we placed the field-based decision factors just after that of the scientific literature.
Table 5.

The most effective decision factors for relief distribution decision-making

5 Synthesis and Discussion

In this section, we synthesize our findings from the Delphi process and discuss them category-wise. Afterward, by exploiting the result, we draw a correlational matrix and propose a relief distribution process model. Finally, we conclude this section by discussing the challenges and portraying our future research directions.

Firstly, distributing a maximum of relief items within a short period is the main objective of the humanitarian operations undertaken in response to disasters  [5]. For successful humanitarian operations, DMs always try a fast response and to meet as many demands as possible [16]. In doing so, the operation must be forecasted with adequate data for need assessment. P12 exemplified the context of the Indonesian Earthquake 2018 to point out that the process should prioritize acquiring and assessing demand data before focusing on serving maximum needs. According to the participant, this is sometimes absent in the process operated in the field. To speed up the process, P44 and P52 suggested focusing on fulfilling the basic needs with quality relief items instead of quantity of relief demand. P24 came with a unique idea of publicly forecasting the need information to serve maximum demand by incorporating the concept of social capital. After sudden-onset, initial responses come from the people inhabiting in neighboring communities when organizational support is still unavailable (P41, P42, P57). So, if they can be forecasted with frequently updated need information, more demands can be served to save more lives. By monitoring communal services, national or international responders can avoid allocating funds for relief items that may stay unused or become surpluses (P24, P25). This will provide flexibility to responders for meeting important demands that are still missing. However, P40 recommended to “…prioritize remote regions for relief operations as small and mediocre organizations keep those regions out of their distribution plans to minimize expenditure” though operational cost and social tension may increase. According to P20 and P71, the success of any relief operation largely depends on the instructions from the sourcing organizations (e.g., hosting government, United Nations) and their mission objectives and capacity.

Speed is one of the critical success factors of relief distribution [29]. When a responding team is planning to serve maximum demands, it needs to find its way(s) for faster mobilization of maximum relief items (transport quantity) to the affected population [16]. According to P26, minimizing travel time would ensure timely relief distribution (distribution time minimization) by increasing the potential number of trips of shipments. Although it is important to shorten travel time, the access constraints need to be considered during emergencies (P58). For example, extreme weather conditions made the relief operation challenging in the East part of Indonesia, where P12 participated. Hence, P24 suggested placing demand notation into a map so that central DMs can select the shortest practical length of the emergency route(s) (hence, shorter travel distance) and calculate minimum travel time to the demand points from the nearest distribution center(s). However, participants identified that minimizing travel time is more important than coverage maximization. Thus, the later factor was re-evaluated in round-2 and listed cluster-2. It would make the entire operation unsuccessful if maximum coverage is planned without minimizing travel time. Hence, P41 remarked, “…do well in one area rather than poorly in all areas”. Furthermore, the cost-related factors are theoretically important (P58), but practically “…saving lives and providing basic needs and medical treatment are of paramount importance as compared to the cost involved” (P3). However, although some participants were in favor of having reasonable (or more) distribution centers for serving affected people, others were not concentrating on this issue as this topic is directed to the central logistics hub.

Secondly, for achieving the objectives of humanitarian assistance and successfully distributing relief items, DMs are required to control some variables [37]. Among the 13 listed decision variables, panel members considered, in the first place, balancing resource need and relief transportation quantity to meet demands at the targeted point of distribution (POD). In doing so, multiple panel members suggested to categorize and prioritize peoples’ needs before dispatching relief vehicles, whereas P24 and P40 emphasized to share the distribution plan beforehand to gain beneficiaries’ satisfaction. For example, the relief packages can be standardized by categorizing the recipients by age, gender, location, households, family member, etc. and if they are informed earlier about the package (food/non-food, heavy/lightweight), they can ensure their arrangements (beneficiaries’ access cost) to receive relief package(s) and return home safely. This will ensure the reduction of social tension, which is one of the most critical and complex issues to tackle in the disaster-arisen chaotic field (P40). Furthermore, to face such challenges, it is also necessary to maintain reduced travel and distribution time that can be done by establishing supply unit(s) with sufficient storing capacities in shorter travel distance, accelerating inventory flow for shortening demand meeting time.

However, none of the cost related issues (penalty, transport, operational, and set-up cost) gained ultimate consensus and hence, rank low. According to the participants, achieving cost-benefit may be important in business logistics, not in HumLog. P3 expressed that “…importance should be given to the mechanism to transport the relief materials as quickly as possible and not the cost involved”. Nonetheless, P40 criticized the hidden cost-benefit issue in humanitarian operations that restrict NGOs to support remote communities. The participant suggested prioritizing those communities while planning for deployment as they are not covered in most of the cases, and if necessary, this can be negotiated with the donors for supporting responding operations in better ways.

Thirdly, to operate an effective and efficient relief distribution, DMs need to satisfy limiting constraints that are not directly controlled by them. For example, budget and resource availability, travel distance, and storehouse capacity gained maximum attention. Humanitarian operations largely depend on donors [19], and humanitarian organizations have no line of credit (P40). Although it is expected to have an adequate budget to support the entire relief distribution mechanism (P3), it is always difficult to convince donors to increase the budget, even if it is needed to cover more survivors in remote areas (P19, P41). Additionally, if the required items (resources) are unavailable in the hosting area (e.g., local market), the logistical costs become higher and affect the entire operation (P24). On the other hand, the number of storehouses and their capacities are centrally controlled and always face space unavailability to the upcoming shipments waiting in the port to be unloaded (P57, P58). Although P71 was mentioning to arrange mobile storages, it would, however, increase operational cost and relief distribution time. Furthermore, unavailable access points would delay the distribution process by limiting road capacity or traveling longer distances (P40, P44). This results in an irregular load flow; inventory holding cost and replenishment cost would increase significantly.

Moreover, geographical location, security, political instability, and weather of the hosting area(s) always bring uncontrollable situations to the operations. Besides, having support from the hosting government and the military, responding teams must be careful while tackling such situations. P19 and P41 suggested to incorporate local informants for continuous situational updates on further sections of a distributing network, and local transport provides as they know the local road-links. Hence, P24 was envisioning a technological system where local communities can post information on certain issues that are further refined by system analysts and graphically presented into a distribution network map. This would help DMs to find alternatives.

Fourthly, after getting a clear understanding of decision factors and their influences on the relief distribution process, it is important to know how each factor in decision objectives is correlated with that of in decision variables and constraints. Table 6 illustrates the positive and negative correlations. For positive correlation, we considered a correlation coefficient of ≥0.3; for negative correlation, we notated all of them although some values were insignificant. By doing so, we warn DMs, in case they intend to consider these factors for the process. The presented correlation matrix guides DMs to select appropriate variables and constraints for achieving certain objectives. By consulting the correlational values in the matrix, DMs can rapidly decide on the factors that are necessary for supporting decision-making and can thus produce decision alternatives for efficient and effective relief distribution.
Table 6.

Correlational matrix of decision factors

Although most of the cost related topics did not achieve consensus and were ranked low, some of them show high correlational significance. For example, the operational cost has the highest impact when practitioners intend to transport maximum relief items to different PODs. It scored highest in both decision variables (0.78) and decision constraints (0.6) categories. This justifies that DMs working in the down-stream of the humanitarian supply chain are not fully independent while budgeting operational costs. They are controlled (to some extent) by donors and central authorities of respective organizations. They may face similar situations when deciding on transport costs and travel distance. However, DMs must be cautious while deciding on variables and constraints because some factors have high positive impacts to achieve certain objectives, whereas the same factor(s) affect other objectives to be accomplished. For example, operational cost and supply unit has a great influence on transporting maximum relief items, whereas they impact negatively on covering maximum demands. Hence, DMs should evaluate the applicability and impacts of those factors in their targeted context(s).

Fifthly, according to [26] and [45], instead of studying separately, all DPAs should be dealt with jointly and concurrently for effective disaster response. Therefore, by utilizing findings from this Delphi study and personal experiences, we have proposed an RD process model in Fig. 2. The model encompasses two distinct portions: information flow (denoted in solid arrows) and material flow (denoted in dotted arrows). To demonstrate processes more clearly, we assumed each DPA as an individual operational entity. The process starts by receiving (continuous) need information from the field that DMs analyze in the distribution centers. The assessed demand information is publicly forecasted immediately for informing neighboring communities to meet initial demand and to maximize coverage. The information on social capital is continuously assembled while preparing the responses by exploiting the decision factors evaluated in this research. By understanding the achieving objectives, DMs concentrate on utilizing necessary variables and constraints along with contextual ones. They consult and negotiate with other DPAs (if related) and plan for deployment.
Fig. 2.

The proposed relief distribution process model

(inspired by [4, 28, 37])

RSC receives initial demand notes and establishes communication with the logistics hub or local market for procuring necessary items. Parallelly, RSC communicates with IM for updates of FL status and Sched for scheduling items to be transported and vehicles to be utilized. Then, Sched contacts with Transp and IM for finalizing the shipment(s) to be stored in FL or sent to the distribution centers (DC). As soon as deploying arrangement(s) is finalized, DC shares the distribution plan with the PODs. After dispatching relief items either directly from the procurement or the selected FLs, DC monitors the entire shipment(s) and continually communicates with responsible ones for updating the safety and security of the selected distribution network. Along with official informants, DC may increase public involvement for a faster update on distribution networks (i.e., blocked road, collapsed bridge), political instability in the network, safety, and security.

While considering limitations, our study faced the typical weaknesses summarized by Hsu and Sanford  [18]: low response rates and high time consumption. Our study also faced the challenge of participation discontinuing in future rounds despite participants being motivated by providing information about the survey topic, method, rounds, outcomes, and the overall research theme. Since we exploited emails to communicate with geographically dispersed experts, it was difficult to reach them as we got no indicate whether we were using the right addresses until participants replied. The conducted interviews were informative, but it was laborious for us to convert them to a questionnaire-like format.

After tackling all these difficulties, these summarized findings allow us to identify paths for future research. Decision factors learned from our work can be translated as system requirements for developing future IS artifacts (e.g., DSS), where the prioritization by the experts can form the basis of a typical Must-Should-Could assessment. The step following this article will be a design-oriented pragmatic approach that will effectively support rapid decision making for efficient relief distribution in large-scale disasters [30]. Our research will focus on proposing an operational ecosystem for RD by examining the influences that it receives from other problem areas introduced in Sect. 1. This operational ecosystem could feedback DSS to produce effective and efficient decision-making support.

6 Conclusion

Relief distribution is the core task of the humanitarian logistics operations. It depends on qualified decision-making in facility location, inventory management, relief supply chain, transportation, and scheduling. Except for a few of them, decision factors in relief distribution are shared by different problem areas. Thus, decision-makers need to know the decision objectives and how and to what extent decision variables and constraints influence them. In this article, we generated and evaluated two generalized sets of decision factors. The first one encompasses decision factors that researchers exploited in their objective functions and models to solve case-specific or scenario-specific relief distribution problems, whereas the latter one incorporates expert-recommended decision factors from the field. To provide decision-makers with manageable and comprehensive advice, we proposed a shortened and prioritized list based on expert ranking. However, for receiving the operational benefit, we still suggest decision-makers to keep tracking all the enlisted decision factors instead of searching the top-ranked ones only. As humanitarian operations are highly contextual and decision-makers face severe uncertainty in information gathering, processing, and implementation [31], we expect that the enlisted decision factors will support them to visualize and understand the changes and quickly identify relevant ones necessary for fast humanitarian relief distribution. We also proposed a correlational matrix to assist decision-makers with an understanding of the influential relationship between decision factors, so that they can select essential decision factors based on their respective contexts.

The findings in this research have various implications. Empirically evaluating the decision factors has extended the current body of knowledge on the RD process in large-scale sudden onsets. Based on our findings, we have contributed to the HumLog literature by extending the existing models to accelerate decision-making in disaster-like deeply uncertain events, where information is infrequent and incomplete. Our research findings, along with the proposed process model, will support field-based decision making in the down-stream of the humanitarian relief supply chain, as well as in the center. Moreover, it serves as an input to information system development to support decision-making. Additional research is needed to refine the findings and extend the process model to prototype and develop a decision support system to assist decision-makers with decision alternatives for actual implementation.

Notes

Acknowledgement

The authors acknowledge the cooperation and valuable assistance received from Eli Hustad, Dag Håkon Olsen, and Rania El-Gazzar on the Delphi process. At the same time, the authors are grateful to all panel members, who provided insightful information, and evaluations and valuable comments for this research on relief distribution decision making.

References

  1. 1.
    Afshar, A., Haghani, A.: Modeling integrated supply chain logistics in real-time large-scale disaster relief operations. Socio-Economic Plan. Sci. 46(4), 327–338 (2012)CrossRefGoogle Scholar
  2. 2.
    Aljamal, M.S., Ashcroft, D., Tully, M.P.: Development of indicators to assess the quality of medicines reconciliation at hospital admission: an e-Delphi study. Int. J. Pharm. Pract. 24(3), 209–216 (2016)CrossRefGoogle Scholar
  3. 3.
    Avery, A., Savelyich, B., Sheikh, A., Cantrill, J., Morris, C., Fernando, B., Bainbridge, M., Horsfield, P., Teasdale, S.: Identifying and establishing consensus on the most important safety features of GP computer systems: e-Delphi study. J. Innov. Health Inform. 13(1), 3–11 (2005)CrossRefGoogle Scholar
  4. 4.
    Balcik, B., Beamon, B.M., Krejci, C.C., Muramatsu, K.M., Ramirez, M.: Coordination in humanitarian relief chains: practices, challenges and opportunities. Int. J. Prod. Econ. 126(1), 22–34 (2010)CrossRefGoogle Scholar
  5. 5.
    Barahona, F., Ettl, M., Petrik, M., Rimshnick, P.M.: Agile logistics simulation and optimization for managing disaster responses. In: Proceedings 2013 Winter Simulation Conference: Simulation: Making Decisions in a Complex World, pp. 3340–3351. IEEE Press (2013)Google Scholar
  6. 6.
    Chunguang, C., Dongwen, C., Xiaoyu, S., Bo, G.: Logistics routes optimization model under large scale emergency incident. In: 2010 International Conference on Logistics Systems and Intelligent Management, Harbin, pp. 1471–1475 (2010)Google Scholar
  7. 7.
    Comes, T., Schätter, F., Schultmann, F.: Building robust supply networks for effective and efficient disaster response. In: ISCRAM Conference, Baden-Baden, pp. 230–240 (2013)Google Scholar
  8. 8.
    Cottam, H., Roe, M., Challacombe, J.: Outsourcing of trucking activities by relief organisations. J. Humanitarian Assistance 1(1), 1–26 (2004)Google Scholar
  9. 9.
    Dajani, J.S., Sincoff, M.Z., Talley, W.K.: Stability and agreement criteria for the termination of Delphi studies. TFSC 13(1), 83–90 (1979)Google Scholar
  10. 10.
    El-Gazzar, R., Hustad, E., Olsen, D.H.: Understanding cloud computing adoption issues: A Delphi study approach. J. Syst. Softw. 118, 64–84 (2016)CrossRefGoogle Scholar
  11. 11.
    English, J.M., Kernan, G.L.: The prediction of air travel and aircraft technology to the year 2000 using the Delphi method. Transp. Res. 10(1), 1–8 (1976)CrossRefGoogle Scholar
  12. 12.
    Gralla, E., Goentzel, J., Fine, C.: Assessing trade-offs among multiple objectives for humanitarian aid delivery using expert preferences. Prod. Oper. Ma. 23, 978–989 (2014)CrossRefGoogle Scholar
  13. 13.
    Grime, M.M., Wright, G.: Delphi method. Wiley StatsRef, 1–6 (2014)Google Scholar
  14. 14.
    Gutjahr, W.J., Nolz, P.C.: Multicriteria optimization in humanitarian aid. Eur. J. Oper. Res. 252(2), 351–366 (2016)MathSciNetzbMATHCrossRefGoogle Scholar
  15. 15.
    MacCarthy, B.L., Atthirawong, W.: Factors affecting location decisions in international operations–a Delphi study. IJOPM 23(7), 794–818 (2003)CrossRefGoogle Scholar
  16. 16.
    Han, Y., Guan, X., Shi, L.: Optimal supply location selection and routing for emergency material delivery with uncertain demands. In: 2010 International Conference on Information, Networking and Automation (ICINA), pp. V1-87–V1-92. IEEE (2010)Google Scholar
  17. 17.
    Hasson, F., Keeney, S., McKenna, H.: Research guidelines for the Delphi survey technique. J. Adv. Nurs. 32(4), 1008–1015 (2000)Google Scholar
  18. 18.
    Hsu, C.C., Sandford, B.A.: The Delphi technique: making sense of consensus. Pract. assess. Res. Eval. 12(10), 1–8 (2007)Google Scholar
  19. 19.
    Humanitarian assistance from Non-State donors (2015). http://www.devinit.org/wp-content/uploads/2015/05/Private-Funding-2015_May2015.pdf. Accessed 30 Mar 2019
  20. 20.
    Kapoor, P.: Systems approach to documentary maritime fraud. University of Plymouth (1987)Google Scholar
  21. 21.
    Kobus, J., Westner, M.: Ranking-type Delphi studies in IS research: step-by-step guide and analytical extension. In: 9th IADIS IS, pp. 28–38. IADIS, Vilamoura (2016)Google Scholar
  22. 22.
    Li, Y., Hu, Y., Zhang, X., Deng, Y., Mahadevan, S.: An evidential DEMATEL method to identify critical success factors in emergency management. A. Soft C. 22, 504–510 (2014)Google Scholar
  23. 23.
    Liberatore, F., Ortuño, M.T., Tirado, G., Vitoriano, B., Scaparra, M.P.: A hierarchical compromise model for the joint optimization of recovery operations and distribution of emergency goods in humanitarian logistics. Comput. Oper. Res. 42, 3–13 (2014)MathSciNetzbMATHCrossRefGoogle Scholar
  24. 24.
    Linstone, H., Turoff, M.: The Delphi Method: Techniques and Applications. Addison-Wesley, Reading (2002)zbMATHGoogle Scholar
  25. 25.
    McMillan, S.S., King, M., Tully, M.P.: How to use the nominal group and Delphi techniques. Int. J. Clin. Pharm. 38(3), 655–662 (2016)Google Scholar
  26. 26.
    Okoli, C., Pawlowski, S.D.: The Delphi method as a research tool: an example, design considerations and applications. Inform. Manag. 42(1), 15–29 (2004)CrossRefGoogle Scholar
  27. 27.
    Peres, E.Q., Brito Jr., I., Leiras, A., Yoshizaki, H.: Humanitarian logistics and disaster relief research: trends, applications, and future research directions. In: 4th International Conference Information Systems, Logistics and Supply Chain, Quebec, pp. 26–29 (2012)Google Scholar
  28. 28.
    Pettit, S.J., Beresford, A.K.: Emergency relief logistics: an evaluation of military, non-military and composite response models. Int. J. Logistics Res. Appl. 8(4), 313–331 (2005)CrossRefGoogle Scholar
  29. 29.
    Pettit, S., Beresford, A.: Critical success factors in the context of humanitarian aid supply chains. IJPDLM 39(6), 450–468 (2009)CrossRefGoogle Scholar
  30. 30.
    Rahman, M.T., Comes, T., Majchrzak, T.A.: Understanding decision support in large-scale disasters: challenges in humanitarian logistics distribution. In: International ISCRAM-MED Conference, Xanthi, pp. 106–121 (2017)Google Scholar
  31. 31.
    Rahman, M.T., Majchrzak, T.A., Comes, T.: Deep uncertainty in humanitarian logistics operations: decision-making challenges in responding to large-scale natural disasters. Int. J. Emerg. Manag. 15(3), 276–297 (2019)CrossRefGoogle Scholar
  32. 32.
    Rancourt, M.È., Cordeau, J.F., Laporte, G., Watkins, B.: Tactical network planning for food aid distribution in Kenya. Comput. Oper. Res. 56, 68–83 (2015)zbMATHCrossRefGoogle Scholar
  33. 33.
    Ransikarbum, K., Mason, S.J.: Multiple-objective analysis of integrated relief supply and network restoration in humanitarian logistics operations. Int. J. Prod. Res. 54(1), 49–68 (2016)CrossRefGoogle Scholar
  34. 34.
    Richardson, D.A., de Leeuw, S., Dullaert, W.: Factors affecting global inventory prepositioning locations in humanitarian operations—a Delphi study. J. Bus. Logistics 37(1), 59–74 (2016)CrossRefGoogle Scholar
  35. 35.
    Rottkemper, B., Fischer, K.: Decision making in humanitarian logistics – a multi-objective optimization model for relocating relief goods during disaster recovery operations. In: 10th International ISCRAM Conference, Baden-Baden, pp. 647–657 (2013)Google Scholar
  36. 36.
    Rowe, G.: A guide to Delphi. Foresight 8, 11–16 (2007)Google Scholar
  37. 37.
    Roy, P., Albores, P., Brewster, C.: Logistical framework for last mile relief distribution in humanitarian supply chains: considerations from the field. In: Proceedings of the International Conference on Manufacturing Research, Birmingham, pp. 11–13 (2012)Google Scholar
  38. 38.
    Safeer, M., Anbuudayasankar, S.P., Balkumar, K., Ganesh, K.: Analyzing transportation and distribution in emergency humanitarian logistics. Pro. Eng. 97, 2248–2258 (2014)CrossRefGoogle Scholar
  39. 39.
    Schmidt, R.C.: Managing Delphi surveys using nonparametric statistical techniques. Decis. Sci. 28(3), 763–774 (1997)CrossRefGoogle Scholar
  40. 40.
    Stewart, D., Gibson-Smith, K., MacLure, K., Mair, A., Alonso, A., Codina, C., et al.: A modified Delphi study to determine the level of consensus across the European Union on the structures, processes and desired outcomes of the management of polypharmacy in older people. PLoS ONE 12(11), 1–17 (2017)CrossRefGoogle Scholar
  41. 41.
    Strasser, A.: Design and evaluation of ranking-type Delphi studies using best-worst-scaling. Technol. Anal. Strateg. Manag. 31(4), 492–501 (2018)CrossRefGoogle Scholar
  42. 42.
    Tofighi, S., Torabi, S.A., Mansouri, S.A.: Humanitarian logistics network design under mixed uncertainty. EJOR 250(1), 239–250 (2016)MathSciNetzbMATHCrossRefGoogle Scholar
  43. 43.
    Trkman, P.: The critical success factors of business process management. Int. J. Inf. Manage. 30(2), 125–134 (2010)CrossRefGoogle Scholar
  44. 44.
    Varho, V., Tapio, P.: Combining the qualitative and quantitative with the Q2 scenario technique—the case of transport and climate. TFSC 80(4), 611–630 (2013)Google Scholar
  45. 45.
    Vitoriano, B., Ortuño, M.T., Tirado, G., Montero, J.: A multi-criteria optimization model for humanitarian aid distribution. J. Glob. Optim. 51(2), 189–208 (2011)MathSciNetzbMATHCrossRefGoogle Scholar
  46. 46.
    Yadav, D.K., Barve, A.: Analysis of critical success factors of humanitarian supply chain: an application of interpretive structural modeling. IJDRR 12, 213–225 (2015)Google Scholar
  47. 47.
    Özdamar, L., Ertem, M.A.: Models, solutions and enabling technologies in humanitarian logistics. Eur. J. Oper. Res. 244(1), 55–65 (2015)MathSciNetzbMATHCrossRefGoogle Scholar

Copyright information

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020

Authors and Affiliations

  1. 1.University of AgderKristiansandNorway

Personalised recommendations