Dodo6/Topic_Modelling_using_LDA

1Article2 3Overcoming Barriers in Supply Chain Analytics—4Investigating Measures in LSCM Organizations5Tino T. Herden *, Benjamin Nitsche and Benno Gerlach6Chair of Logistics, Technische Universität Berlin, Straße des 17. Juni 135, 10623 Berlin, Germany7* Correspondence: herden@logistik.tu-berlin.de8Received: 23 December 2019; Accepted: 17 February 2020; Published: 26 February 20209 10Abstract: While supply chain analytics shows promise regarding value, benefits, and increase in11performance for logistics and supply chain management (LSCM) organizations, those organizations12are often either reluctant to invest or unable to achieve the returns they aspire to. This article13systematically explores the barriers LSCM organizations experience in employing supply chain14analytics that contribute to such reluctance and unachieved returns and measures to overcome these15barriers. This article therefore aims to systemize the barriers and measures and allocate measures to16barriers in order to provide organizations with directions on how to cope with their individual17barriers. By using Grounded Theory through 12 in-depth interviews and Q-Methodology to18synthesize the intended results, this article derives core categories for the barriers and measures,19and their impacts and relationships are mapped based on empirical evidence from various actors20along the supply chain. Resultingly, the article presents the core categories of barriers and measures,21including their effect on different phases of the analytics solutions life cycle, the explanation of these22effects, and accompanying examples. Finally, to address the intended aim of providing directions23to organizations, the article provides recommendations for overcoming the identified barriers in24organizations.25Keywords: Supply Chain Analytics; Logistics; Supply Chain Management; Grounded Theory26 271. Introduction28The business of logistics and supply chain management (LSCM) is changing rapidly based on29new technologies and consumer trends that demand new and broad digital capabilities [1]. The30phenomenon that is taking hold of this business, transforming it, and that demands action from the31organizations in the market, is digitalization [2]. One of the inherent effects of digitalization is the32constant growth of data, describing all sorts of aspects of the world and, thus, possessing potential33insights about it and opportunities to act on those insights. This growth is not expected to halt or34decelerate but to accelerate exponentially—a recent study estimates a growth of the global datasphere35from 33 zettabytes in 2018 to 175 zettabytes in 2025 [3].36To leverage the growth of data and the opportunities of digitalization, the studies mentioned37above emphasize the use of analytics. Analytics is the use of various quantitative, explanatory, and38statistical methods on extensive amounts of data [4]. When focused on the context of logistics and39supply chain management, it is specified as supply chain analytics (SCA) [5]. Scholars have40established that SCA helps to improve organizational and supply chain performance, increase41efficiency, reduce supply chain costs, and contribute to competitive advantage [6–10].42The implementation and application of SCA is, however, an immense challenge for43organizations. Scholars and organizational reports alike have demonstrated a variety of barriers44organizations face and the resulting effect of low implementation rates. Scholars have presented45barriers ranging from management failures and lack of analytical knowledge to unwillingness to46Logistics 2020, 4, 5; doi:10.3390/logistics401000547 48www.mdpi.com/journal/logistics49 50Logistics 2020, 4, 551 522 of 2753 54commit IT resources [9–14]. Organizational reports highlight the barriers of cost, insufficient data,55and not achieving the aspired to benefits [15–17].56However, these studies did not explicitly intend to explore such barriers but identified them in57addition to their research objectives. Despite these studies clearly implicating the existence of various58barriers, scholars in the field of LSCM have rarely addressed the topic of measures to overcome these59barriers. An exception is the study of Schoenherr and Speier-Pero [12], who provide some insight on60factors contributing to the successful implementation of analytics in LSCM. However, research on61practices and measures has concentrated on the research field of analytics in general. As a result,62there has been no investigation of whether LSCM employs specific measures, which measures are63most relevant for LSCM managers, and what impact they have. Thus, this study intends to take a64specific look at the barriers that occurring for organizations investing in analytics in the field of LSCM65with the additional intend to identify which measures and practices are being applied to overcome66these barriers. This research is intended to support logistics and supply chain managers concerning67directions in which to go forward with SCA in a successful and impactful manner and support the68transformation of their supply chains into the digital age.69In particular, this study seeks to contribute to the following research objectives:70•71•72 73RO1: Identify and systematize the organizational barriers that appear when companies initiate,74perform, or deploy SCA initiatives within the organization.75RO2: Identify and allocate organizational measures that seek to cope with the depicted barriers.76 77In pursuit of these objectives, 12 in-depth interviews were conducted with different actors along78the supply chain, including suppliers, original equipment manufacturers (OEMs), retailers, and79logistics service providers, as well as additional analytics providers with strong experience in LSCM.80From the exploratory and extensive interviews, the study identified the barriers to initiating SCA81initiatives, applying SCA, and deploying SCA solutions and derived and the organizations’ measures82for coping with such barriers. The collected data was analyzed using the Grounded Theory approach83[18]. This approach was supplemented by the Q-Methodology [19,20] to derive a theoretical84framework of measures and barriers in SCA, mapping barriers and measures of SCA initiatives and85their relationships. This contribution, in terms of systemizing barriers and measures, extracting core86categories, and identifying proposed relationships, provides an important basis for further deductive87research.88The remainder of the article is structured as follows. Section 2 provides an overview of the89relevant theoretical background to this study. Section 3 presents and details the methodology. Section904 contains the results and their discussion. The conclusion, including implications, further research91directions, and limitations, is presented in Section 5.922. Theoretical Background93To overcome barriers that prevent the successful use of SCA solutions, this section introduces94SCA with exemplified effects and its previously documented barriers. Further, this section presents95the recommended practices discussed in non-domain-specific analytics literature as a preliminary96analysis of available measures.972.1. Supply Chain Analytics98SCA is understood as the domain-specific application of business analytics (or analytics), which99is “concerned with evidence-based problem recognition and solving that happen within the context100of business situations” [21]. In organizations, it is applied by analysts (“data scientists”) using the101most advanced and complex tools and methods [4] or by process experts in the form of self-service102analytics exploiting the accessibility of analytics in respective software for self-service [22]. Analytics103is an essential component of generating value from “Big Data” [23]. Analytics has been reported to104increase decision-making effectiveness, can generate organizational agility (given a fit between105analytics tools, data, people, and tasks), and can contribute to providing competitive advantage [24–10626].107 108Logistics 2020, 4, 5109 1103 of 27111 112SCA is specific to the domain of LSCM. LSCM is concerned with efficiently integrating the actors113of the supply chain (suppliers, manufacturers, logistics service providers such as warehouses and114transportation, retail) such that products are manufactured and distributed to the customer in the115right amount, at the right time, at the right quality, and with system-wide minimal cost [27]. As a116pivotal perspective to the definition based on actors and objectives, Christopher’s definition [28]117introduces the various activities and managed entities of LSCM. These include management of118procurement and movement and storage of products in the different stages of their life cycles as119materials, parts, and finished inventory. While the multi-objective trade-off under the influence of120various actors, tasks, and entities already indicates the need for advanced methods to support121decision-making, further aspects increase this need. Dynamic vertical, horizontal, and diagonal122interactions among actors create hard to control complexity [29], complexity increases the occurrence123of disruptive events and, thus, operational and financial risks [30], and due to the focus on the124customer, common methods for managing disruptive risks from other domains are potentially a poor125fit to LSCM [31]. Further, besides disruptive events, volatility results in further mismatch of supply126and demand [32], adding to the need for advanced methods to support decision-making.127SCA caters to the various decision-support needs of LSCM as it exploits the variety of analytics128methods for the multitude of issues [5]. Thus, it is concerned with applying quantitative and129qualitative analytical methods to recognize and solve problems evidence-based in the context of130LSCM [5,33,34]. In this regard, SCA helps to measure and monitor performance, understand and131control poor performance, and improve performance [6–8]. However, with the variety of tasks132requiring support come a variety of applications that are hyper-specialized and also provide a133challenge for organizations in LSCM [10]. Further, adoption of SCA in organizations in LSCM is slow,134unwillingness to share data with supply chain partners is widespread, and organizations experience135barriers [12,13,35].1362.2. Barriers of Supply Chain Analytics137In this section, barriers to adopting and employing analytics are reviewed. Several studies have138discussed individual barriers in the domain of LSCM. However, these barriers are widely spread139across studies with no study systemizing them or exploring barriers explicitly to create a140comprehensive catalog. The presented literature overview is intended as a basis for the data141collection. This literature overview is limited to the field of LSCM.142Following the cycle of an adoption process, an early barrier to adopting analytics is the143approachability of analytics. Efforts to adopt and employ analytics may be slowed down by a lack of144consensus regarding the terms surrounding the concept of analytics and the subsequent confusion145[13,14] as well as the perception of complexity and the difficulty of its management [11,12]. This is146accompanied by a constant change of technologies and methods, which organizations would need to147be able to keep up with [6,10]. Further, managers see the lack of LSCM specific solutions as a barrier148[12]. Sanders [10] describes this inability to adopt analytics due to overwhelming complexity as149“analysis paralysis.”150When approaching analytics despite the perceived complexity, organizations in LSCM151encounter a lack of experience of employees in analytics and utilizing data [10,12]. As opposed to the152paralyzing complexity described above, the missing literacy concerning data and analytics is153expressed in missing knowledge and creativity. The lack of knowledge about how to approach and154utilize data has been repeatedly stressed in the literature [10,12–14,35,36]. This includes the155identification of data most suitable for analytics, understanding which data is useful and which is156useless, experience with relevant technologies, ideas about what to do with the available data,157knowledge on how to transform data into information for decision-making, and how to drive the158supply chain with data. Oliveira et al. [36] emphasize organizations’ lack of roadmaps on using data159and information even after systems for data collection have been set up. A further hindrance in this160context is the inability to clearly point out how value in terms of the business objectives is generated161from analytics, and thus to sell analytics initiatives to key stakeholders [13].162 163Logistics 2020, 4, 5164 1654 of 27166 167To adopt and employ analytics, resources are needed. Labor resources comprise the first form168of resource identified in the literature as creating challenges for organizations in LSCM. One study169reports time constraints as a primary barrier [12], indicating the lack of time to get familiar with170analytics and relevant technologies as well as to execute initiatives in addition to daily business. Zhu171et al. [37] emphasize the time-consuming effort of getting analytics solutions into production172(meaning implemented and used in the value creating process), which might consume time from a173variety of employees and hinder parallel initiatives. In addition, some labor requirements for174analytics present a challenge for LSCM organizations in themselves. The literature emphasizes a lack175of employees in LSCM organizations to handle and understand data, analytics software, and IT176systems, as well as being able to interpret the analytics results [38]. However, Kache and Seuring [13]177report that LSCM managers lack an understanding of what skills the required employees need to178possess, leading to problems in recruiting these employees.179The second form of resources needed is monetary investment. The cost of solutions available on180the market, the cost of getting the necessary raw data, and systems integration (of the focal181organization and its partners) are named as barriers to adopting analytics [10,12,14,39]. In addition,182the returns from the investments and the benefits are hard to predict and unclear in advance. Thus,183it is hard to estimate the breakeven point [10,14,36]. Furthermore, scholars highlight the problem of184attribution of any performance increase to the investment in analytics. They present issues of185quantifying benefits, attributing benefits to analytics or process changes, and time lags between186implementation and identifiable performance increase (ramp-up) [36,38,40]. In this regard, scholars187have referenced the “IT productivity paradox,” as discussed by Brynjolfsson [41], which presents the188paradoxical situation of researchers not being able to measure a productivity increase from IT while189organizations were increasingly investing in it (e.g., because performance increase is not measurable190with their usual performance measurement or because of time lags).191To employ analytics, several data sources have to be (continuously) combined to solve complex192business question. However, the issue of integrating systems in a diverse and incompatible IT193landscape impedes the integration of data sources. LSCM managers reported a lack of integration of194existing systems to scholars due to evolved (rather than designed) environments with specialized195systems for the various business units or due to legacy systems not intended for data exchange [12–19614,38]. Scholars have emphasized that the issue is particularly present when supply chain partners197with unaligned systems (fragmented systems with varying maturity) are expected to combine data198sources [13]. Mistakes in this integration of systems can lead to inaccurate results and wrong199decisions [36].200The supply of data is another major barrier to the adoption and employment of analytics in201LSCM. First of all, a lack of data or a lack of timeliness in providing the data has been reported as202hindering [10,12,42]. Hazen et al.[42] have particularly stressed the issue of insufficient data quality203in general, which other scholars have confirmed, expanding the issue to having different quality204levels controlled by different employees [7,14,39]. Hazen et al. [42] further stressed lack of means for205measuring data quality, controlling data quality, and the persistence of bad quality data as input to206analytics until they are actively removed (since they are not consumed as opposed to other inputs).207Further scholars added inexperience in assuring data quality to the list [43].208Having access to data comes with the issue of responsibility for that data, which can limit its209usability for analytics. Handling customer data alongside the organization’s own data brings the210responsibility for organizations in LSCM to ensure data privacy and data security [13]. Moreover,211access does not equal ownership. Data ownership and the associated rights to data can be missing or212unclear [14,38].213To gain access to relevant data from supply chain partners or the right to use the data from214partners for analytics, their collaboration on data and analytics is needed. However, several scholars215present the issue of unwillingness of supply chain partners to participate in data sharing or its216dependency on the provision of incentives and security [13,43]. Partners may want to, and should,217keep certain data confidential (e.g., for legal reasons), but they also keep their data to themselves218because of lack of trust and fear of losing control over their data [13,14].219 220Logistics 2020, 4, 5221 2225 of 27223 224If all the barriers above are somehow overcome, the use and subsequent benefits of analytics225solutions in LSCM organizations are not assured. Scholars report issue with the mindset of employees226regarding analytics solutions. The employees may not want to use new, analytics-based systems, do227not show openness toward them, and are not eager to change their culture [9,12,38]. Richey et al. [14]228specifically discuss how employees in LSCM value relationships and trust-based business generation,229which they do not want to sacrifice for data driven solutions.230Another barrier to the adoption and employment of analytics in LSCM can be the physical231process assumed to be supported by the analytics solution. Scholars discuss the reduced impact of232analytics if the process has a low level of uncertainty (or volatility) [37,39]. While this takes only a233certain range of application areas into account and ignores benefits for complex information inputs234or benefits for faster decision-making, it presents the potential flaw of employing analytics to235processes that may offer little value in terms of return on the investment. Reversing the perspective,236Oliveira et al. [36] emphasize the need for a certain level of process maturity before employing237analytics and gaining value from it, and Srinivasan and Swink [39] present the need for flexibility in238the physical process such that reactions to analytical results can be applied. If there is no flexibility,239the opportunities to gain value from analytics are foregone since they cannot be executed. As a result,240the value and benefits from analytics depend on the status of the physical process, whose weaknesses241can become barriers to utilizing the full potential of analytics.242Assuming that all the above barriers are overcome, the absence of governance of analytics can243nevertheless hinder organizations from gaining the expected value from analytics. A lack of top244management commitment and understanding of advantages from analytics will of course hinder the245adoption of analytics in the first place [11]. When employing analytics, the lack of governance can246result in a lack of focus and missing relevance of the results produced (“measurement minutiae”),247reducing the effect of analytics [10]. However, scholars have stressed that the result of the absence of248strategically and managerially controlled analytics is isolated and fragmented analytics efforts,249resulting in only small benefits (even in disadvantages for other functions) and unaligned analytics250maturity along business functions [10,38]. To conclude, analytics efforts need to be guided by an251organizational understanding of the purpose of analytics and it should be incorporated into the252business strategy. Otherwise, gaining the hoped for value and benefits can be impeded [13].2532.3. Measures to Fully Utilize the Benefits of Analytics254As opposed to barriers, measures are rarely addressed in the LSCM literature. Even in the field255of analytics, countering barriers is usually a side note. In particular, no study could be found256systemizing measures, deriving core categories, or identifying effects and relationships. In this257section, practices and measures are explored considering literature on analytics in general and the258few studies from the LSCM domain.259Schoenherr and Speier-Pero [12] investigated several contemporary aspects related to the260adoption level of analytics in LSCM organizations and created a list of circumstances that lead to a261higher motivation for organizations to adopt analytics. Their list does not intend to show how262organizations created the circumstances intentionally or unintentionally but explores the status quo263in these organizations. Thus, the list does not represent executable practices. The motivators include264the existence of aspects the absence of which was noted, in the discussion above, as forming a barrier,265such as senior leadership promoting analytics and displaying commitment. Further, the use of266analytics is encouraged by competitors and colleagues, and to a lesser degree by customers as well.267The strongest motivator for using analytics has been reported as the user’s conviction about the value268of analytics, implying the need to experience the value from analytics, probably in support of the269user’s own processes.270Scholars have presented several practices to make the use of analytics more appealing, which271could lead to increased conviction concerning the value of analytics and motivate its continued use.272This includes visually appealing software interfaces, which would also help to make the results easier273to understand and comprehend; mobile availability via tablet computers and smartphones, which274have been observed to increase the frequency of using analytics; and user-engaging approaches that275 276Logistics 2020, 4, 5277 2786 of 27279 280strive to stimulate interaction, like gamification or self-service [44]. Further, showing the outcome of281decisions on individual scorecards or assessments after (minor) decisions is supposed to build trust282in analytics and show its value, while scholars report this application in a supportive way as opposed283to a blaming or judging way [25,45]. Additionally, for stakeholders, who are needed to provide284funding, analytics can be made more appealing by articulating relevant business cases [4].285In contrast, scholars have also reported more instructive practices. Recommended supportive286change management efforts include promotional activities, training courses, and coaching287concerning analytics and its goals and benefits [45–47]. Further practices include underlining the288necessity to discontinue non-data driven methods, the demand for data-based explanations in289decision-making, and incentives for data-driven decision making [46,48]. One recommendation even290suggests managers allowing themselves be overruled by analytics to promote their trust in it [48].291However, the replacement of employees who are unwilling to change has also been suggested [46].292For advancing analytics in the organization, several practices have been presented. To create293more relevant business cases requires managers to build a certain understanding of analytics—not294for application, but for understanding execution and requirements [25]. Another recommendation295for managers is to ask second order questions, which are, roughly described, questions not about296how to improve the solution to a problem, but questions around finding another solution to the297problem [49]. However, efforts are also suggested from the analysts’ side, either by getting298“translators” to explain analytics in business terms or requiring analytics experts to build business299understanding [4,25,50]. Overall, the LSCM-specific and general analytics literature suggests an300incremental and evolving approach, potentially starting in high impact areas [10,50].301One relatively specific practice, as compared to the other practices presented above, is the302collaborative approach between analytics experts and business experts, which helps to build analytics303expertise on the business side, and vice versa, and supposedly results in better use cases and304outcomes. This has been suggested by several scholars, named a “field and forum” approach,305“shadowing,” or simply described as the analytics expert joining the business expert in his real-work306value creation and decision-making processes, including the related data usage, to learn and discuss307opportunities for improvements afterwards [44,49,51]. Another related practice is the use of “data308labs”, in which experts from the different areas are co-located to work collaboratively free from the309distractions of daily business [44,49]. These practices are complemented by agile development310methods, in which the progress of shorter periods (“sprints”) is discussed, e.g., by assessing311prototypes [44].312In contrast to these practices oriented toward individuals, some practices to align and integrate313the analytics efforts of an organization are suggested. Scholars suggest setting up an enterprise-wide314information agenda and strategic directions for data and analytics [46,52]. A cross-functional315integration of several business units and collaboration in the organization (e.g., in a newly created316center of excellence for analytics) and with key partners has been suggested specifically for LSCM,317since analytics initiatives quickly influence partners [10,13,46]. Additionally, researchers have318suggested creating a “single-source-of-truth,” which demands data exchange between all business319units, as an organizational measure [45], although this also touches on technological issues.320Concerning technological aspects, one issue addressed in the literature is data. Scholars have321recommended creating data standards and automating data onboarding, integration, and quality322processes as much as possible to accelerate the creation of insight from data [44]. It has further been323observed that analytically mature organizations employ centralized groups responsible for ensuring324data quality and availability [4]. In the context of LSCM, the structuring of data in a manner described325as “scrubbing” has been recommended as a first step in maturing in analytics [10]. The highlighted326importance here is that the errors which would occur during data generation are eradicated such that327data is already clean, structured, and organized for insight creation.328Another technological aspect comprises the IT systems. By analyzing the challenges and329opportunities of analytics in LSCM, scholars stress the status of existing IT systems and their role in330making the use of analytics challenging [13]. Thus, they recommend the prioritization of continuous331IT investments. Concerning data integration issues in and between organizations (e.g., with supply332 333Logistics 2020, 4, 5334 3357 of 27336 337chain partners), researchers have presented the example of digital platforms, which act as a hub-andspoke system with interfaces to several partners and customers instead of creating a point-to-point338connection [10]. Such a platform would level different formats and standards.3393. Methodology340For this study, a mixed-methods approach has been chosen to support the sub-tasks of the two341formulated research objectives, identification and systematization. The approach is set out in the342following sections. Data collection and data analysis are explained. This section concludes with343reliability considerations.3443.1. Research Design345A mixed-methods approach has been chosen, combining grounded theory [18] and the Qmethodology [19,20,53] to exploit synergies between them. Explicitly, after open coding in which346phenomena are identified, Grounded Theory intends the scholars to use axial coding to relate the347identified phenomena to each other and selective coding to extract core categories [18]. This search348for core categories is executed via the Q-methodology, which intends to perform a perception based349sorting, creating categorizations and additionally unveiling patterns of perceptions forming the350categorizations and providing insights on them [19,20]. This mixed-methods approach provides a351more robust framework for the relationships compared to either individual approach. Each of the352methods has been used in the context of LSCM to identify barriers and measures.353Researchers use Grounded Theory with the intention of creating deeper knowledge on a354research phenomenon and have repeatedly done so to identify barriers and measures in LSCM. For355example, studies have identified and categorized barriers and coping mechanisms on implementing356sustainable practices in LSCM [54], the implementation of supply chain technologies in emerging357markets [55], and the transition towards a supply chain orientation [56]. These studies investigated358barriers to transition to a rationally more beneficial position, usually strongly influenced by human359behavior, and the systematization of measures and their effects. They emphasize the capability of360Grounded Theory to identify and aggregate knowledge as well as to build a consensus [18].361Specifically focusing on extracting consistent aggregation and groups of concepts, the Qmethodology has been used as the method of choice in LSCM research. The method has been used to362form a mutually exclusive and exhaustive taxonomy of biased behavior in supplier selection [57],363conceptualization of moderators and sources of Supply Chain Volatility [32], and to aggregate364attitudes and levels of acceptance towards different innovations and practices in low-input food365supply chains across multiple cultures and multiple stages of the supply chain [58]. These articles366present, by example, the utility of the Q-Methodology for aggregating measures and extracting367conceptualizations in LSCM research and resultingly support the systematization of identified368knowledge.3693.2. Data Collection370The data collection method was based on established practices. To sample relevant experts as371interviewees, boundaries have to be defined [59]. First, to gain a broad overview of barriers and372measures in LSCM, the boundary of relevant organizations for data collection were set to include373various Supply Chain actors, as introduced in Section 2.1: manufacturing (OEM and suppliers), retail,374logistics service providers, and, in addition, analytics providers with a focus on LSCM. Second,375experts had to have relevant job functions related to analytics. Third, experts had to have experience376with analytics in the context of LSCM, and thus experience with SCA. Potential interviewees were377identified based on the first two criteria and invited to the study. After an invitation to participate,378potential interviewees with a positive response expressing interest were sent an overview of the379study [60]. Based on the overview, the potential interviewees were asked to evaluate their experience380with SCA—the third boundary criterion—and, in case of negative evaluation, asked to help to contact381the most knowledgeable informants on the subject matter. If their experience satisfied the criteria, the382 383Logistics 2020, 4, 5384 3858 of 27386 387experts were included in the study. After 12 interviews with 13 interviewees, the data collection was388concluded. At this point, the identification of new barriers and measures through additional389interviews had attenuated, while the effort to recruit further interviewees had increased390substantially. This was evaluated as fulfilling the condition of saturation [18] and is comparable to391similar studies in LSCM [61–63]. Table 1 lists the interviewees.392The interviews were conducted in a semi-structured format using an interview protocol for393reliability, which was initiated with a grand tour open-end question [64,65]. Interviewees were asked394about barriers regarding analytics as applied to LSCM that they were currently experiencing or had395experienced in the past. Based on their responses, they were subsequently asked about two aspects:396first, to provide detailed information on the barriers for increased understanding of the barriers’397effect, and second, how these barriers were being/had been addressed, and how the measures used398affected the barriers. Subsequently, interviewees were asked about their experience with the barriers399presented in Section 2.2 and, depending on the response, asked about measures taken to overcome400these barriers. The second part of the interview on measures was again initiated with a grand tour401open-end question on best practices used to increase the success rate of analytics initiatives and402interviewees were asked to give justifications for their use. These justifications were requested403interrogated to understand whether these practices were used to solve or overcome particular404problems, and whether these practices qualify as measures. The interviews were concluded with405questions on the use of the measures presented in Section 2.3 and, depending on the response, the406justification for their use.407Table 1. Interview Participants.408 409Participant410A411B412C413D414E415F416G417H418I419J420K421L422M423 424Position (Anonymized)425Manager (functional) Analytics426Data Scientist427Data Scientist428Head of Analytics429Manager Analytics430Director Analytics431Manager Analytics432Head of (functional) Analytics433Manager Analytics434Sen. Data Scientist435Data Scientist436Data Scientist437Head of Analytics438 439Actor440OEM441LSP442Supplier443OEM444Retail445OEM446Supplier447LSP448Retail449LSP450analytics Provider451analytics Provider452LSP453 454analytics Exp. [yrs]4558456124573458194593460846124626463144644465346634673468 469Since none of the interviewees was located near the researchers, most interviews were conducted470by phone. Previous studies have used telephone interviews for interview research in LSCM as a471measure to overcome the restrictions from different locations without any reported bias [61–63,66–47268]. Subject to the approval of the interviewees, the interviews were audio-recorded and transcribed473verbatim, providing the qualitative data for analysis. The interviews lasted between 45 and 105474minutes, with an average of 65 min. Interviewers took handwritten notes during the interviews for475the purpose of recording and guiding the interview. The Analysis was initiated after the first476interviews to allow preliminary interpretations and insights, which were expanded in subsequent477interviews. All data were documented in a structured database for further reliability. Audio records478were deleted after transcription as committed to the interviewees.479 480Logistics 2020, 4, 5481 4829 of 27483 4843.3. Data Analysis485In the first step, the data was rigorously analyzed according to Grounded Theory guidelines486[18]. The analysis was initiated after the third interview to allow continuous contrasting of487developing theory and data collected from the ongoing interviews [18,69]. The analysis steps, starting488with open coding, were executed using ATLAS.ti (Version 8.4). During open coding, the interviews489were coded to identify recurring themes. The coding was focused on identifying the underlying490nature of challenging conditions during the application of SCA in the organization and extracting the491actions taken to cope with those challenging conditions, their original intent, and their actual impact.492Due to the intent of this research, the codes created were mainly abstractions and vivo codes of the493interviewees. The interviews were continuously read and reread to establish similarities and494disparities. As a result, codes and the categories they represent were restructured.495After concluding the data collection and finalizing the open coding for the data, axial coding496was conducted to lift the level of abstraction as well as further restructure and aggregate existing497categories [18]. During this process, short descriptions of categories of barriers and measures were498created as input for the Q-Methodology based selective coding process.499In this regard, the Q-methodology is used as a supplement to the selective coding steps and,500thus, to the Grounded Theory approach. Scholars have emphasized the adaptability of the Qmethodology to the interest of researchers, such as using differences of sorting as input for501discussions [70]. As a result, the Q-methodology was adapted for this study to identify different502patterns of thought about sorting the aggregated codes that were the output result from the axial503coding step. These sorting results were used as input for consensus-building on the core categories,504which the selective coding intends to derive. In this regard, participants developed their individual505sorting and contributed their underlying patterns of thought to the consensus building in an unbiased506manner by performing the steps of the Q-methodology and commenting on their results without the507other participants criticizing the sorting. The value of the participants’ comments on their sorting508process has been highlighted by Ellingsen et al. [53].509In detail, the Q-methodology starts with two steps that set up the sorting process. First, a510concourse is created, which is the collection of statements pertaining to a research area of interest511derived from empirical and secondary data sources [19,20]. This step is equivalent to the open coding512step that identified 154 codes—in this case on barriers in applying SCA and measures to overcome513them – from the semi-structured interviews. The subsequent step is the creation of a Q-sample, which514represents a comprehensive, balanced, and representative subset from the concourse [19,20]. This515step is intended to limit the number of statements by grouping similar statements and refining516statements that are too specific or too general [20,53]. This step is equivalent to the axial coding step517that related similar codes in the data, 30 barriers and 59 measures.518The sorting procedure, named Q-sort, is performed on the Q-sample as a third step [53]. As519participants, three researchers with high expertise on LSCM with different viewpoints on analytics520have been chosen. In accordance to Ellingsen et al. [53], the participants received a sorting instruction,521were reminded about the non-existence of a “right” sorting, and were given the task to perform the522Q-sample freely. The sorting instruction was explained as to sort the items based on the participant’s523perception of similarity. The free mode of the Q-sort was chosen such that the participants had to524create their own groups. This approach intended to promote the strength of the Q-methodology, as525repeatedly emphasized, of uncovering different patterns of thought, perception, and opinions, which526can be used to asses areas of consensus and friction [19,20,70]. Here, consensus on the core categories527of barriers to SCA and measures to overcome these barriers was assessed.5283.4. Reliability529The reliability of this study was judged based on criteria regularly used by scholars evaluating530the reliability of similar studies [54,56,71,72]. The criteria of credibility, transferability, dependability,531confirmability, and integrity in these studies emerged from the recommendations of Hirschmann [73]532as well as Lincoln and Guba [74]. The criteria of fit, understanding, generality, and control emerged533from Strauss and Corbin [18].534 535Logistics 2020, 4, 5536 53710 of 27538 539Credibility describes the extent to which the results are acceptable representations of the data. It540was addressed by the timeframe of four months in which the interviews were conducted, the review541of the research summary by the interviewees, and the Q-methodology that led to a review of the542codes by three researchers [56,71,72].543The extent to which the findings can be applied from one context to another is covered in the544criterion of transferability. This was addressed by the theoretical sampling, which led to a diverse set545of organizations, business sizes, and roles in the supply chain [56,71,72].546The criterion of dependability refers to the dependence of the results on time and place, and as547a result the extent of the results’ stability and consistency. To ensure dependability, interviewees were548asked to reflect on the decisions, actions, and changes over time that were related to the phenomenon549and, thus, their past and present experience [56,71,72].550Confirmability measures the extent to which interpretations are the result of interviewees’551experience as opposed to researcher bias. To create confirmability, the research protocol was used552before and during the interviews to ensure that interviewees could prepare and answer without553researcher bias [60,65] as well as by providing the results to the interviewees for review [56,71,72].554Integrity of this study, representing the extent of influence by misinformation and evasion by555participants, was ensured by conducting non-threatening, professional, and confidential interviews556[56,72].557The extent to which the findings were a fit to the area under investigation, inherent in the criteria558of fit, was addressed by the credibility, dependability, and confirmability criteria [56,72].559The results’ reflection of the “own world of the interviewees, which is covered in the criterion of560understanding, was addressed by providing the results to the interviewees with a request for561comment and review [18,54,72].562Generality describes the extent to which the findings discover multiple aspects of the563phenomenon under investigation. It was addressed by applying the Q-methodology to integrate564different patterns of thought into the study [20] and by conducting interviews of a sufficient length565(45–105 min), with extensive openness by the interviewees, strongly focused on the area under566investigation and covering multiple facets related to that area [18,54,56,72].567Finally, control describes the extent to which organizations can influence aspects of the emerging568theory. By focusing on measures, the study strongly focuses on aspects over which organizations569have control, which was discussed with the interviewees during the study to establish the existence570of control [18,54,72].5714. Results and Discussion572Resulting from the analysis of the data, one framework regarding the impact of barriers on SCA573initiatives and another framework regarding the measures to counter these barriers have been574derived. Due to the analysis, core categories have been extracted and abstracted, which assemble the575frameworks. This section presents and explains the derived frameworks.576A major inference made from the analysis of the collected evidence has been that barriers and577measures affect different steps along the cycle of applying analytics to a LSCM organization. During578the Q-Methodology, the researchers built consensus on distinguishing four phases:579(1) “Orientation about Analytics” describes actions, circumstances, and events before specific580analytics initiatives are planned. During the orientation, the necessary conditions for applying581analytics are created and employees are motivated to invent analytics initiatives to be executed.582(2) “Planning of analytics initiative” describes actions, circumstances, and events during the set-up583of a specific analytics initiative, in which the addressed business problem/ business case is584specified, the approach designed, resources and budget committed, and relevant people are585invited to participate.586(3) “Execution of analytics initiative” describes actions, circumstances, and events during the587development and creation of analytics solutions in specific analytics initiatives. For example, this588includes interactions with data, applications of analytical methods, the use of technology, and589interaction of analysts with business experts.590 591Logistics 2020, 4, 5592 59311 of 27594 595(4) “Use of analytics solution” refers to actions, circumstances, and events after the solution596development, which include the deployment of the analytics solution to users and their597interaction with the solution in the short and long term.598Allocating the effects to project models such as CRISP-DM was rejected as unsuitable. Several599barriers and measures occur outside the usual scope of such models. Further, barriers and measures600show similar effects in most solution development phases, which are differentiated in too much detail601in these project models.602In addition, as extracted from the evidence, a distinction in the impact of barriers and measures603is made into capabilities and culture. Capabilities refers to the capabilities of the organizations, the604organizations’ processual and technological infrastructure and standards, as well as the knowledge605and skills of organizational members. On the other hand, culture in the context of this study refers to606the attitudes and behaviors of single individuals or groups of individuals in the organization, such607as their motivation, solution-orientation, feelings, openness, and willingness, as well as their critical608reflection on their attitude and behavior. In this sense, culture does not have to reflect the general609culture of the organization, but only groups of individuals of the organization.6104.1. Barriers611This article has identified and systematized barriers that occur when companies initiate,612perform, or deploy SCA initiatives within the organization. These Barriers can be allocated with high613confidence to either capability issues or cultural issues. Ambiguous cases, which researchers had to614discuss extensively, were solely found in actors who omitted critical reflection on their decisions,615creating problems in subsequent process steps. Eventually, a consensus was reached about such616behavior corresponding to the attitudes of the actors and has, thus, been allocated to the cultural side.617Below, the identified categories of barriers and their impacts are described, with examples. The618framework of barriers on applying analytics to LSCM is presented in Figure 1.619From an aggregated point of view, the evidence shows that cultural barriers occur more620frequently during the phase of orientation and use, whereas capability barriers are more frequent621during the phases of planning and execution of analytics initiatives. This result is not surprising, since622the phases of orientation and use demand the engagement and motivation of, as well as interactions623among, individuals who have their primary focus on tasks and activities unrelated to analytics.624Unfamiliarity with and distance from the topic create a different attitude and behavior towards625analytics as opposed to that of analysts. Interviewees repeatedly emphasized that this unfamiliarity626and distance must be overcome by clearly showing these individuals the benefits of analytics in their627specific areas of work. Capability barriers in these phases can either not be determined specifically,628since the specific problem for an analytics initiative is not determined in the orientation phase, or the629potential capability barriers were addressed in the solution development prior to use. The other way630around, in planning and execution individuals familiar with and close to analytics are necessary,631resulting in reduced barriers of attitude or behavior in terms of unwillingness. However,632discrepancies between needed and available skills, technology, and data are impactful since they633hinder the tasks and activities necessary for solution development.634 635Logistics 2020, 4, 5636 63712 of 27638 639Unfitting conditions640• Analysts (talent) not641attracted by organization642Missing responsibility643• Undefined data644ownership leads to645missing appreciation of646data647Missing Knowledge648• Employees don’t649understand value and650possible use cases651 652Capabilities653Missing Knowledge654• Employees miss understanding of functionality and655limits of Analytics656• Users inexpedient requirements on solutions657• Users mis-translate business need to analytical need658• Management is reluctant to invest in uncertain659benefits from Analytics660Unfitting conditions661• Missing budget for foundation for Analytics omits662Initiatives663Missing responsibility664• Missing oversight on activities creates unaligned665and redundant initiatives666 667Orientation about668Analytics669 670Missing Knowledge671• Employees are unable to672apply Analytics673• Users miss understanding674of non-functionality based675performance criteria676 677Unfitting resources678• Shortage of data679• Incorrectness of data680• Heterogeneity of systems681• Business rules not present682in raw data683•684Technical issues omit data685Unfitting conditions686• Privacy and confidentiality sharing with SC Partners687measures complicate688Missing responsibility689unrelated data use690• Undefined data691• Target process has limits692ownership complicates693and restrictions694data access695• Target process is not696standardized or stable697 698Planning of Analytics699Initiative700 701Execution of Analytics702Initiative703 704Unwillingness705• IT department hesitates to create foundation for Analytics706• Employees are committed to established non-Analytics707practices708• Management is not committed to Analytics709 710Missing responsibility711• Undefined data712ownership leads to713missing awareness of714impact of data changes715Missing Knowledge716• Users are unable to use717Analytics solutions718 719Use of Analytics720solution721 722Unwillingness723• IT department doesn’t support724analytical tools725• Users are unwilling the share726data for solution727 728Emotion729• Employees fear to lose jobs and tasks730• SC Partners misbelieve in benefits or fear repercussions731from data sharing732Missing critical thinking733• Employees accept shadow IT as substitute for formal734Analytics735• Management misdirects investments in hyped technologies736 737Missing critical thinking738• Management misdirects739investments in unfitting740Analytics initiatives741 742Missing critical thinking743• Analysts create workarounds744during development that are745not deployable746 747Culture748 749Figure 1. Barriers of Supply Chain analytics.750 751Unwillingness752• Users are unwilling to use753Analytics solution754Emotion755• Users mistrust solution to756capture all domain757particularities758 759Logistics 2020, 4, 5760 76113 of 27762 7634.1.1. Capability Barriers due to Unfitting Conditions764Barriers due to unfitting conditions refer to circumstances in the organization’s processes,765infrastructure, environment, or reputation that strongly impede or prevent the execution or766completion of the development of an analytics solution. The results show that such barriers have been767observed in all phases except for the use phase. An early obstacle for organizations is absence of the768talent needed for analytics initiatives. Such talent does not perceive LSCM as an attractive or769challenging domain, as opposed to technology organizations at the methodological forefront. In the770planning of initiatives, efforts can be halted due to various missing tools, technologies, or data needed771for the initiative. These foundational conditions can be missing since organizations—especially772smaller organizations—lack the budget for continuous investment. Further, such conditions can773become apparent during the execution of the initiative, when activities have already begun. An774example is the realization that the target process cannot be supported by analytics as intended, either775because it is not standardized and stable enough to create a beneficial solution or because it has776process limits and restrictions that prohibit creating improved solutions compared to those that exist.777Such conditions usually demand fundamental changes in the organization outside the realm of778analytics.7794.1.2. Capability Barriers due to Missing Responsibility780Barriers due to missing responsibility occur along all phases and usually result in extra effort,781avoidable with clearly assigned responsibilities. Missing responsibility on ownership of data sources782and the necessary actions and requirements connected to these responsibilities leads to situations in783which knowledge about existing data sources and their location is unevenly distributed and the data784and their flow are not well documented. Thus, individuals in the organization lack the understanding785of data and their importance—an important component for the invention of potential initiatives. In786specific initiatives, this complicates execution due to unnecessary effort to access data owners and787their data, effortful coordination between several owners, and missing traceability of data usage788through the organization. Imprecise responsibilities on data may result in changes to data by789employees unaware of their relevance to analytics solutions in use, compromising the solution790without any communication of changes. In addition, missing oversight of analytics leads to the791development of heterogenous tools, methods, and solution approaches, likely to produce redundant792solutions for similar problems.7934.1.3. Capability Barriers due to Missing Knowledge794Missing knowledge on analytics is a substantial barrier to using analytics and can affect the795application of analytics in LSCM organizations in a variety of ways. Without individuals in business796processes who understand analytics and have experienced how analytics can create value, the search797for potential initiatives in an orientation phase is less likely to result in meaningful initiatives for these798business processes. Since the value is usually generated indirectly from actions taken due to the799analytics solution, this barrier is inherent regarding analytics. During the planning phase, this800indirect value generation and also the inherent uncertainty of the generated benefits, which are highly801dependent on the data and context, can lead to underestimation of the resulting value and reluctance802towards investments. The other way around, missing knowledge can result in high and unworkable803expectations, impractical or absurd demands on solutions abilities, and ignorance of the limitations804of analytics. Combining this with a missing ability to translate the business need into an analytical805need due to missing knowledge, these Barriers will result in initiatives that have no way of achieving806the unrealistic or mis-communicated objectives. In the execution, missing knowledge on functionality807of analytics can result in the inability to let the business experts participate in the solution process, or808missing understanding of performance evaluation in analytics can result in the inability to gain809acceptance for progress. Lastly, missing knowledge can result in the users’ inability to use the810 811Logistics 2020, 4, 5812 81314 of 27814 815developed analytics solutions in their business processes, denying the actual gain from the solution’s816value.8174.1.4. Capability Barriers due to Unfitting Resources818Unfitting resources (data and systems) are barriers that become present during the planning or819execution of analytics initiatives, but usually affect the execution. Unfitting resources can prolong the820development of the solutions, yield subpar success, or even halt the solution. These issues are usually821technical, such as the relevant data is missing, especially for machine learning or time series methods;822the data is incorrect, requiring intensive cleaning and feedback for correct values from the process823experts; or the accessible data does not fully reflect the business rules, requiring additional824preparation. While solving these issues is foremost time intensive but solvable upon identification,825the identification of incorrectness or lack of fit to business rules might occur at a late stage, such that826the effort is doubled, or the solution developed from the data does not gain acknowledgement from827the users. Further, technical issues related to heterogeneity of systems and the resulting828incompatibility of the data with the systems prevent data access or the integration of the data for829more comprehensive analysis, which can either be overcome with additional time and effort, or830development of the solution may ultimately be abandoned.8314.1.5. Culture Barriers due to Unwillingness832As mentioned above, the reference to culture considered in this article does not have to reflect833the culture of the general organization, but can apply to small groups of individuals. In this case,834small groups that behave unwillingly in their areas of responsibility can seriously impact the success835of analytics in an organization, including specific initiatives or the ability to perform analytics at all.836Considering the barriers of unwillingness, they become relevant when individuals motivated for837analytics become dependent on the input and cooperation of others. During orientation, the838dependency on IT departments to create the foundation for analytics in data and analytical software839can block the efforts of motivated individuals, when the IT department is unwilling. Interdependent840with this, employees unwilling to commit to analytics will suppress management’s motivation for841analytics, while missing commitment from management can suppress employees’ motivation (and842necessary resources such as budget) for it. During the planning phase, when ideas are expressed, the843motivated individuals bring together their ideas and business problems but are less dependent on844third parties, assuming this phase is entered after management expressed support. In the subsequent845phases, in particular the unwillingness of potential users to provide their data or to use the solution846can have a strong negative effect on the benefits realized with deployment of the analytics initiative.847This unwillingness can be traced to an unwillingness to change any business customs that are already848established, because of the effort to change (or rather the ease to remain with the familiar) and the849implication of changes, such as different responsibilities (including reduced control), modified850behavior, and altered processes. This is often accompanied by an expressed disbelief in the potential851and value of anything that is not the habit.8524.1.6. Culture Barriers due to Emotion853Emotions, especially of fear and perceived unfair treatment, are strong barriers to employees854accepting the organization’s pursuit of analytics or use of solutions. In the data, they have been855primarily observed as a reaction to information asymmetries of individuals not actively working on856analytics or not involved in solution processes, who fill the asymmetries with assumptions. Involving857employees such as potential users or supply chain partners in the planning and collaborating with858them in the execution can create some level of transparency, such that emotion is a lesser barrier in859these phases. However, in the phase of orientation, during which employees and supply chain860partners might be left without well communicated intentions of the usefulness of analytics or might861not fully believe the communicated intentions due to events in the past (e.g., something that builds862general mistrust unrelated to the individuals eager to use Analytics), they might react emotionally to863 864Logistics 2020, 4, 5865 86615 of 27867 868the information asymmetries. In detail, employees fear losing their jobs and supply chain partners869fear repercussions from sharing data, building resistance against analytics initiatives. Similarly,870developing solutions such that the solution process is not transparent to the potential user can lead871to mistrust and skepticism towards the solution by the users. This emotion of mistrust is a reaction872to the information asymmetries between requirements and the adherence to the requirements by the873developed solution. This is particularly nurtured by analysts who over-sell their solutions.8744.1.7. Culture Barriers due to Missing Critical Thinking875Missing critical thinking in the context of this paper refers to decisions made without the876necessary reflection of these decisions. While this could be a result of missing knowledge and, thus,877be a capability barrier, it can also result from not taking enough time and effort for consideration, or878enthusiasm about a critical problem finally being solved, which is not to be impaired by critical879considerations. This can occur as bandwagon behavior from management directing the budget880towards some hyped technology or some hyped use cases unfitting to the business need, while the881budget would be needed for foundational technologies or uses cases that are boring, but beneficial.882The missing business impact of hyped but failed initiatives can generate mistrust of employees in883new technologies, as discussed above. On a user level, the continued use of business owned884tools/shadow IT, which usually lack adherence to standards, documentation, or may even have885errors, is often not critically reflected on regarding the impact of the creator leaving the organization,886compatibility with other systems, or fit with business strategy and rules. Thus, the development of a887standardized and adhering substitute in an analytics initiative is usually not triggered, while being888highly necessary. During solution development, one recurring issue is the creation of workarounds—889technical debt—by analysts without reflection on the impact on later development stages. The issues890usually surface later and create additional effort or can even prevent the success of the analytics891solution and its benefits.8924.2. Measures893This article has identified measures organization deploy to cope with the barriers above. Overall,894an allocation of measures that are directly assigned to specific barriers could not be derived from the895data, since interviewees expressed that some measures address several barriers, and vice versa, by896intention or not. Further, interviewees explained that measures are adjusted and advanced over time897as a reaction to the specific needs and capabilities of the organization and employees. Thus, measures898on barriers are path and context dependent. However, the measures have been categorized and899systematized to core categories and allocated to barrier categories. While several measures are900presented in the following section, it must be emphasized that the presentation is reduced to these901core categories as derived from the Grounded Theory process. Examples are presented to improve902comprehensibility and to further illustrate the core categories. It cannot be generalized that certain,903specific measures will help all organizations to overcome certain, specific barriers, especially in their904vanilla form.905Reduced to the core categories, measures generally address capability barriers but contribute to906overcoming cultural barriers at the same time. Acceptance is gained by building knowledge and907creating processes for more transparent development and information exchange—information908asymmetries are reduced. Thus, the cause is addressed (missing capabilities), but not the symptoms909(unwillingness). Comparably to the barriers, the identified measures affect the different stages of the910analytics initiative lifecycle differently. While demystification is essential to aid the orientation phase,911the creation of specific capabilities affects the project development phases and communication and912involvement become vital for the actual analytical work in the execution and use phases. The links913are illustrated in Figure 2, with examples for measures in the respective core categories.914 915Logistics 2020, 4, 5916 91716 of 27918 919Demystification920• Diverse formal and informal921communication of benefits922from Analytics923• Expressing needs for Analytics924foundation towards leadership925• Present Analytics value in926tangible form (business needs,927completed initiatives, PoC)928• Pro-active Analytics training929• Guided ideation workshops to930identify use cases931• Substitute Shadow IT by932formal Analytics solution933• Evaluate data for use cases934 935Obtaining Capabilities936Human937• Mentored self-service Analytics938• Pro-active Analytics training939• Recruiting of Analytics savvy people940• Develop employees to citizen Data941Scientists942Data943• Datafication and increase of944automated data collection945• Allocation of data ownership to946business units947• Data quality feedbacks948• Data Lake and open data policy949Technology950• Stepwise convergence of IT system951(consolidation, integration,952modularization)953 954Involvement and communication955 956Organization957• (permanent) cross-functional teams958• Hybrid localization of Analysts959(centralized and decentralized)960• Business units have designated budget961for (meaningful) Analytics962• Transparent Analytics code of conduct963Procedures964• Documentation on used data965preparation and analytics stored with966data source967• Integration of solution into workflow968• Scheduled long-term evaluation of969Analytics solutions970• Pre-checks on data quality and971process standardization972 973• Enhancing Analysts’974collaboration975• Extensive user involvement in976agile development process977• Employees are demanded to978justify decision-making with979data980• Co-location or field-and-forum981approach of Analysts with982potential users983• Extensive trainings and retrainings on solutions984• User-focused deployment of985solutions986 987Capabilities and988Culture989 990Orientation about991Analytics992 993Planning of Analytics994Initiative995 996Execution of Analytics997Initiative998 999Figure 2. Measures to support supply chain analytics.1000 1001Use of Analytics1002solution1003 1004Logistics 2020, 4, 51005 100617 of 271007 10084.2.1. Measures Contributing to Demystification1009The measures contributing to demystification are intended to communicate the value and1010benefits of analytics, fitted to the intended receivers and their context. The most widely used1011measures to reach the mass of employees are formal and informal communications presenting1012achieved value and benefits, such as workshops, presentations, internal conferences, or1013communication material. Interviewees stated that the act of showing achieved value, especially1014accompanied by individuals who benefitted from the executed initiatives, was effectively convincing1015and provided the opportunity to address questions and lack of clarity in live formats. However,1016critics may not attend and, thus, may not be convinced. Successful initiatives are therefore to go on1017tour with roadshows disseminating the evident benefits of analytics across the organization.1018Similarly, pro-active training on analytics creates an understanding of analytics and allow questions1019to be answered. Even when applied to employees less likely to use the methods, this resourceintensive approach creates acceptance and a realistic expectation. Another form of demystification is1020the creation of use cases combined with process experts for their respective domains (e.g., the1021different units involved in the LSCM activities). The created use cases represent a more tangible and,1022thus, comprehensible form of value from analytics. Similarly, this approach is more likely to convince1023supply chain partners to cooperate. Use cases can be identified by guided ideation workshops,1024evaluating existing data sources, as part of pro-active training, or by offering to formalize business1025owned tools/shadow IT. However, time for conferences and training events meant to demystify1026analytics for top management is rarely available. Investments are motivated by the value from1027tangible and clear use cases or a consistently voiced need for investments harmonized across several1028business units.1029While the value created by analytics initiatives is mostly uncertain, different approaches for a1030value estimation exist. Estimations based on comparable solutions (e.g., other organizations or1031business units, previous initiatives) or based on the terminated inefficiencies engaged in the analytics1032initiative are preferable, but such tangibles are frequently missing. Thus, the creation of proofs-ofconcept or prototypes is needed, which demands initial investment to allow a value estimation. If1033these initial investments are too high (i.e., cost intense data collection), a rare approach is the use of1034hypothetical prototypes, for which the data are simulated. However, this kind of prototype only tests1035ideas and requires serious changes to develop a production-ready solution.10364.2.2. Measures Contributing to Obtaining Capabilities—Human1037Enabling humans by building their analytics capabilities involves several forms of training,1038dependent on whether they are expected to contribute to analytics initiatives as domain experts or1039expected to apply analytical methods. To enable their contribution, either a training event for1040employees is carried out or data savviness becomes a recruitment criterion for business roles and top1041management. The approach of enabling contribution accepts that not every employee needs to1042execute analytical methods. However, this portion will decline, while not every employee can be1043upskilled. To improve application capabilities, self-service analytics and citizen data scientists are1044created. In self-service analytics, employees are provided with role specific tools, which allow for1045data-driven decision making and dedicated analytical analysis for their roles. This should be fed back1046by analysts regularly. The citizen data scientist concept acknowledges that certain employees are1047already practically in the role of an analyst for their respective unit. These are identified, provided1048with additional training and the respective title to act as analyst in the business while having strong1049domain knowledge, therefore providing the first contact for analytical inquiries. Both concepts1050improve access to analytics to allow employees—potential users of analytics solutions—easier1051interaction with analytics.1052To get access to analytical talent, organizations may alter their image to appeal to potential1053candidates outside of the company. This includes the portrayal of analytically complex and1054innovative initiatives and the active engagement of the talent in universities and platforms relevant1055for that talent (conventions, internet forums). As talent is understood to be a limited resource1056 1057Logistics 2020, 4, 51058 105918 of 271060 1061necessary for superior performance in competition, organizations must take active steps to attract1062that talent.10634.2.3. Measures Contributing to Obtaining Capabilities—Data1064In regard to data being repeatedly expressed as an inevitable issue in any analytics initiative in1065the discussions on barriers, the multitude of described measures was not surprising. However, these1066measures are often peripheral topics to other core categories. Considering the measures solely1067dedicated to data, centralizing data and establishing open data policies are supposed to create access1068to the data needed for individual analysis and analytics initiatives. Further, measures to design data1069collection processes such that the eventual collection of sensitive personal data is factored in1070beforehand reduce side effects on accessibility of related non-sensitive data. Emphasized for data1071quality and availability are automation, such as automated data collection, validation rules for1072manual data collection, or the automated interpretation of free text with machine learning, as well as1073increasing datafication of processes and products in consultation with analysts. In conclusion, the1074measures emphasize the design of data collection and storage under the consideration of future use1075of the data for analytics.1076Measures peripheral to organizational measures include organizational measures to improve1077data quality and availability. The core of the collected measures is to create awareness on data quality1078at the point of data creation, which is the business unit. This includes the allocation of data ownership1079and responsibility to the business units, providing feedback and training on data quality, but also, in1080return, establishing data quality business units as points of contact for issues and guidance.1081Peripheral measures with technology will be discussed below.10824.2.4. Measures Contributing to Obtaining Capabilities—Technology1083The capability building measures for technology address the creation of an IT ecosystem that1084supports analytics. This means, on the one hand, creating accessibility and consistency of data, and1085on the other, to enrich options for analytics solutions and enhance the deployment of solutions. The1086essence of technology capability measures taken by organizations is to develop the IT ecosystem1087towards a single-source-of-truth. However, this vision might not be achieved—or even aspired to—1088since it displays a level of complexity that is hard to handle and might not be cost efficient. In1089particular, the idea of leapfrogging to this vision was explained by interviewees to be unreasonable.1090Instead, organizations move in the direction of this vision in different forms dependent on their1091individual needs either by consolidating IT systems in smaller numbers and integrating obsolete1092systems into newer systems; by setting up IT platforms as integration layers, which allow interchange1093between systems; or by replacing systems with more modular systems to develop a plug-and-play1094style IT ecosystem. Thereby, the platform solution is the preferred way to overcome technical1095integration issues with supply chain partners. In conclusion, the measures taken that aimed to create1096IT ecosystems for analytics need a few years’ foresight to keep complexity controllable and1097investments reasonable.10984.2.5. Measures Contributing to Obtaining Capabilities—Organization1099To clarify, organizational capabilities address changes of the organizational structures and1100processes, while procedural capabilities address changes of the analytical solution development1101process. The organizational capabilities have several focus areas, including structures that reduce the1102organizational distance between analysts and domains, as well as creating a better understanding of1103each other’s tasks and issues. This includes the localization of analysts in a hybrid format with1104centralized analysts in a center of excellence and decentralized analysts in the business units to1105exploit the advantages of both forms. Dependent on the existing structures, a recurring measure is to1106enhance process improvement units with analytics to make use of the existing closeness of these1107established units and the business domains. Further, the initiatives are executed in cross-functional1108 1109Logistics 2020, 4, 51110 111119 of 271112 1113teams. One organization advanced the latter to permanent cross-functional analytics teams with own1114backlog of projects.1115Another focus area is the development of, and commitment to, rules and codes of conduct. This1116can be employee focused to build acceptance and trust in analytics, such as a code of conduct on what1117restrictions exist on accessing and analyzing data as well as committing to restrictions for automation1118solutions, which must benefit the employees instead of replacing them. Transparency and1119communication of these commitments are key. The rules can further be organization focused and1120shift control on analytics related decisions. Examples are the centralization of decisions on sharing1121data with partners or even making data sharing part of contracts with partners, which is becoming1122more frequent due to service level agreements and performance evaluations.1123Finally, designated budgets for analytics as an organizational change in the budgeting process1124are used to stimulate analytics initiatives. However, the appropriateness of the resulting initiatives1125needs to be checked (e.g., by consultation with analysts).11264.2.6. Measures Contributing to Obtaining Capabilities—Procedures1127The collected measures to enhance the development process of analytics solutions are most1128notable for presenting reactions to overcome issues. Some of these formalized procedures address1129information and knowledge gaps, which are time- and resource-consuming to close, while the1130measures are merely cost-efficient fixes. As described by the interviewees, the solution for these gaps1131is to gain experience with analytics, which can be achieved by participating in analytics initiatives1132where the fixes are used to ensure avoiding these issues until the experience gained makes them1133obsolete. In this sense, business experts’ inability to translate business needs to analytical needs can1134be fixed through review and feedback with analysts. Analysts’ misinterpretation of users’ demands1135for analytics solutions, which are as a result unneeded and unused, is fixed by formal agreements1136establishing commitment and willingness. Business experts’ unwillingness to cooperate is fixed by1137analysts assuming responsibility for the focal decision-making processes of the analytics initiative.1138Other formalized measures schedule activities early to avoid impairing issues in later stages. This can1139occur during the stage of project execution, supported by pre-checks of available data, or process1140stability and flexibility, and cloud-hosted containerization of the solution development. This can also1141occur during the phase of using the solutions and facilitating their long-term success using1142documentation, including incorporating solutions into standard operational procedures and1143scheduling long-tern evaluations of the solutions as part of deployment.1144As pointed out by one researcher during the data analysis phase of this article, some measures1145listed under this core category are merely practices for successful project management. However,1146humans might have difficulties in transferring practices from one domain of application to another1147while still collecting experience with the new domain.11484.2.7. Measures Contributing to Involvement and Communication1149The final core category of measures identified represents measures contributing to the creation1150of involvement and communication in order to improve each side’s understanding of the other side’s1151actions and needs in order to improve decision-making. These measures can benefit an executed1152analytics initiative both directly and indirectly. Direct measures address a high level of involvement1153of the intended users in the development process, such as early and constant inclusion in an agile1154solution development format with regular sprint meetings, co-location of analysts and users, or a1155collaborative field and forum approach with analysts directly observing and taking part in the1156activities and decision-making of the users. Further, deployment processes can focus extensively on1157the users’ experience with the solution by presenting the solution in a user-oriented form, by1158gradually introducing the solution into the process, or by providing consumable and user-oriented1159training events. Interviewees emphasized that the focus on the user during all stages of development1160is vital for the initiative’s success. Besides the importance of user involvement, it was further1161acknowledged that the user’s mindset is occupied with their business responsibility, while their1162 1163Logistics 2020, 4, 51164 116520 of 271166 1167obligation to an analytics initiative is the advocation of business interests and not the understanding1168of analytical methods.1169However, to advance in analytics, the organization’s employees need to develop an1170understanding of the interpretation of the results of analytics solutions and their implications and1171become familiar with them. Thus, a repeatedly highlighted indirect measure is constantly requesting1172justification for employees’ decisions in the form of data-based results. Another indirect measure is1173to promote and support exchange of analysts. These indirect measures create an environment that1174enhances collaboration in potential upcoming analytics initiatives.11754.3. Discussion on Applying Measures and Handling Barriers1176As introduced above, the impact of measures on an organization could not be generalized from1177the data and is highly dependent on the context of the organization. Due to this and the dynamic1178nature of measures existing at different stages of development, allocating measures to specific1179barriers was not supported by the collected data. Below, further points regarding the core categories1180derived from the Grounded Theory methodology are discussed.1181Considering the capability barriers, the variety of requirements for more complex analytics1182initiatives are underlined. Scholars have highlighted the need for a technical foundation [52] as well1183as the required talent [10], the co-dependency on technical capacity and skills [75], the need for data1184quality [42], or the required knowledge of employees and managers [76]. This study identified a1185multitude of requirements: analytical talent, the IT systems landscape, the data to be analyzed,1186organizational structures, the knowledge of the people expected to cooperate with analysts, and the1187condition of the unit of analysis. Therefore, this study identified that data interacts with all other1188requirement while decreasing in quality and accessibility, subject to these other requirements. In1189particular, lack of responsibility affects data negatively, which indicates a missing understanding of1190data as an organizational asset—a paradigm scholars have emphasized in the past [75]. However,1191interviewees reported impactful results achieved with analysis on a small scale—small enough to be1192run on personal laptops—demanding lower levels of commitment.1193Discussing cultural barriers, this study uses this label for the behaviors and beliefs of individuals1194in the organization, which may not reflect the general culture of the organization. The barriers1195allocated to this category highlight the cross-functional nature of analytics in LSCM, since they result1196from treating analytics as an isolated technology topic. The need for collaboration and exchange has1197been discussed in the literature [77], but this study emphasizes the information asymmetries resulting1198from lack of cooperation as having the consequence of individual beliefs that negatively impact the1199acceptance of analytics in the organization. While unwillingness on the employee level is the more1200direct effect to observe, this study coincides with Hayes’s [78] argument that unrealistic expectations

Showing the first 1,200 of 1704 lines. Download the file for the rest.