Codifying construction supply chain responses under COVID-19 into a decision tree: structured decision-making under extreme disruptions

The construction industry's performance depends on tightly coordinated deliveries of critical, often long-lead equipment and materials, where recent Black Swan Events (BSEs) exposed how fragile these flows can be, producing shortages, schedule slippage and cost escalation across projects (Arcadis International Construction, 2019; Deloitte, 2020a, b; IBISWorld, 2023; Regional Economic Conditions Conference - Interview to David Mortenson, 2024). Within lean-based practice, Just-in-Time (JIT) reduces onsite inventory and space needs by timing deliveries to installation, improving cost control under stable conditions (Tommelein and Li, 1999; Tommelein and Weissenberger, 1999; Vokurka and Davis, 1996). Yet JIT's low buffers increase exposure to upstream failures and variability, as widely documented in implementation challenges and supply chain risk literature (Im et al., 1994; Zsidisin et al., 2005). To hedge extreme uncertainty, organizations pivoted toward Just-in-Case (JIC), strategically maintaining buffers, advancing purchasing and adding redundancy, marking a resilience-oriented shift from pure Lean efficiency (Brakman et al., 2020; Suri and De Treville, 1986). Industry cases reflect this balance, even leaders of JIT recalibrated after major shocks (e.g. stockpiles for critical components), reinforcing the need to combine Lean with preparedness under BSE conditions (Choi et al., 2023). In construction, practical mitigation includes early procurement of long-lead items and planned storage, with the cost-of-carry weighed against schedule risk and escalation (Akintoye, 1995).

Recent BSEs showed the fragility of global supply chains and their impact on construction. The COVID-19 pandemic disrupted material availability, logistics and workforce mobility, leading to widespread project delays and cost escalation. The Suez Canal blockage in 2021 further demonstrated systemic vulnerability by halting nearly 12% of global trade, stranding critical cargo and causing severe schedule delays for time-sensitive components (Ebrahim, 2021; Shoulberg, 2021). The Russia–Ukraine war has since added new constraints by limiting access to essential commodities and energy inputs, escalating costs across international construction markets (Kim et al., 2022; Liadze et al., 2023). Together, these crises highlight how global interdependencies amplify local shocks and reveal the acute vulnerability of construction supply chains, which often rely on long-lead, highly customized and nonsubstitutable equipment (Ivanov, 2020).

Although resilience strategies such as diversification and inventory buffers are increasingly discussed, decision-making during such crises is often reactive and lacks systematic evaluation. Many contractors relied on intuition, ad hoc negotiations with suppliers or reactive redesigns, which created inconsistency in project outcomes (Soto Ortiz, 2025). Literature has offered valuable insights into supply chain risk management, but the absence of structured tools to guide procurement decisions under BSE uncertainty remains a critical gap (Dolgui and Ivanov, 2020). Decision trees, applied successfully in other complex systems, provide a means to structure expert reasoning into clear, comparable alternatives and outcomes.

While previous crises such as the COVID-19 pandemic, the Suez Canal blockage and the Russia–Ukraine conflict have revealed the vulnerability of JIT systems and prompted shifts toward JIC strategies, systematic tools for evaluating procurement decisions under extreme uncertainty remain limited. To bridge this gap, this research develops a decision tree framework for procurement under BSE conditions. Drawing on semi-structured interviews with industry experts, the study identifies how professionals navigated procurement decisions before, during and after the COVID-19 pandemic and structures these experiences into a decision tree model. The tree was validated using classification metrics to assess predictive performance. The findings demonstrate that decision trees can codify practitioner knowledge, highlight trade-offs across procurement strategies and provide a structured basis for decision-making under extreme uncertainty. By translating experiential knowledge into a structured, data-driven format, the study offers both a practical tool for industry decision-making and a methodological contribution to resilience-oriented procurement planning and construction supply chain research.

BSEs are defined by their extreme rarity, unpredictability and significant impact, making them distinct from conventional risks (Taleb, 2007, 2008). Although they lie outside normal expectations, their systemic consequences are felt across industries and sectors, with construction particularly exposed due to its dependence on long-lead, customized equipment and fragmented supply chains (Craighead et al., 2007; Sodhi et al., 2012). For construction firms, vulnerabilities were intensified by project-based structures, tight cash flows and limited inventory buffers. Research has highlighted that disruptions can trigger contractual disputes, reputational damage and even bankruptcy when continuity cannot be maintained (Louch and Cooper, 2020). These consequences reveal that the challenge is not only logistical but also financial and organizational.

Another critical dimension is organizational response, where Prieto (2015) observed that extreme events shift firms from risk minimization to resilience-seeking, while Bieńkowska and Tworek showed that resilience depends as much on dynamic capabilities and workforce adaptability as on material supply chains (Bieńkowska and Tworek, 2020). As disruptions escalate and uncertainty compounds, firms must navigate conditions that surpass routine risk events and demand coordinated responses across project teams, contractors and supply networks. Given these vulnerabilities, recent scholarship has turned toward identifying how construction supply chains face extreme disruptions. These discussions form the foundation for understanding the range of mitigation tools available to construction professionals and how such strategies can be applied under BSE conditions. The next section builds on these insights by examining existing approaches to construction supply chain risk management, highlighting operational, contractual and organizational strategies proposed in the literature to strengthen resilience under deep uncertainty.

Building on the outlined vulnerabilities, some mitigation strategies were proposed by authors showing that effective risk management must combine operational fixes with governance and learning mechanisms (El-Sayegh et al., 2020; Kendrick, 2015; Muneeswaran et al., 2020; Osipova and Eriksson, 2009). For supply chain disruption, studies recommend preparedness built from prior events and formalized learning, as well as workforce-safety protocols and site-level compliance systems to sustain operations during health or logistics shocks (Amoah and Simpeh, 2021; Boztemur, 2018).

A persistent weakness during disruptions is information delay and communication breakdowns across project actors. Recommended countermeasures include establishing deliberate information-sharing routines, escalation paths and onsite awareness/training programs to keep safety and logistics decisions synchronized under stress (Adafin et al., 2020; Amoah and Simpeh, 2021; Bing et al., 2005; Tembo Silungwe and Khatleli, 2020). Technological capacity gaps at suppliers, when firms cannot meet evolving specifications or data requirements, are mitigated through early capability vetting, phased qualification and collaborative problem-solving with key vendors (Kendrick, 2015; Osipova and Eriksson, 2009). At the network level, insufficient multisourcing and reactive outreach heighten exposure; thus, relationship depth and collaborative routines with suppliers, coupled with tailored strategies for globally expanding chains, are repeatedly emphasized.

Inventory management strategies emerged as a central lever but require differentiation by item criticality and complexity. For commodity/low-complexity items, price-focused tactics may suffice; for high-impact, complex items, total cost-of-ownership analysis and deliberate buffer positioning are advised while balancing cash-flow and obsolescence risks (Dixit, 2022; Sakthivel et al., 2021). In parallel, the field has shifted from a narrow, list-and-monitor conception of risk to resilience-oriented management. Core practices include risk mapping, buffers and multisourcing, supported by network-level coordination and systems thinking (Amoah and Simpeh, 2021; Chopra and Sodhi, 2004). To expose hidden vulnerabilities ahead of a shock, anticipatory methods such as failure determination and red teaming are proposed as structured “pre-mortems” to test plans and suppliers under adversarial scenarios (Masys, 2012).

Emerging work also points to technology enablement, such as data integration, monitoring and analytics, as an enabler of faster sensing and response across partners (Raja Santhi and Muthuswamy, 2022). Previous literature synthesis conducted by the authors further discussed how construction supply chains have increasingly incorporated digitalization, modularization, prefabrication and resilience-oriented procurement strategies to improve adaptability under disruptive conditions associated with BSEs (Soto Ortiz et al., 2026a). Gharaibeh et al. (2024) have similarly mapped the growing role of Industry 4.0 technologies, including digital traceability, real-time data sharing and integrated supply chain platforms, in improving construction supply chain performance and transparency, while identifying persistent gaps between theoretical frameworks and actual industry implementation.

Traditional risk management approaches are criticized for underestimating low-probability, high-impact disruptions and for treating risks as isolated rather than cascading. Recent scholarship emphasized the need for holistic frameworks that integrate uncertainty, adaptability and complex interdependencies within supply networks (Aven, 2015; Bridger, 2019; Dindarian, 2023; Lindaas and Pettersen, 2016; Pettit et al., 2010; Van Der Vegt et al., 2015). Recent literature has further confirmed that procurement and supply chain risks, including supplier unreliability and limited material availability, remain among the most pressing challenges in construction, highlighting the persistent need for integrated risk management frameworks (Ogunmakinde et al., 2025). In sum, the literature converges on a portfolio approach: governance and information discipline; supplier diversification and collaboration; differentiated inventory policies; anticipatory stress-testing; and technology-enabled visibility. This strategic toolkit motivates the structured, criteria-based decision support developed in this paper's decision tree, bridging from general prescriptions to project-specific, transparent procurement choices.

However, despite these advancements in theory and practice, critical gaps remain that justify the need for the current research. Existing studies confirm that construction supply chains under BSEs suffer from systemic interdependencies and reactive responses (Chopra and Sodhi, 2014; Ivanov, 2020). Addressing this requires structured tools that reflect deep uncertainty and project-specific decision conditions. Although literature increasingly addresses supply chain risk in construction, it offers limited methodological guidance for procurement decisions under the deep uncertainty of BSEs. Most strategies remain reactive, context-dependent and poorly documented, leaving contractors without a consistent method to evaluate alternatives during crises (Soto Ortiz et al., 2026b). Studies emphasize resilience but lack the link between project-specific variables, such as the timing of disruption or project phase and structured mitigation choices, where decision-making tools remain underutilized in construction supply chain contexts. Motivated by these gaps, this study develops a decision tree framework that operationalizes resilience principles into structured, criteria-based decision support, bridging theoretical prescriptions with practical procurement choices under deep uncertainty. The proposed model addresses the absence of structured procurement decision-support tools in the existing construction supply chain literature by organizing mitigation strategies by disruption timing, project phase, thematic classification and implementation context using empirical decision pathways derived from industry professionals.

This study employed a mixed-methods research design focused on expert interviews and the development of a structured decision tree model to support procurement decision-making under BSEs conditions. The method was designed to explore how construction professionals adapted procurement decisions under the uncertainty and unpredictability of BSE conditions through qualitative interpretation, thematic classification and structured decision mapping.

The study explored how construction professionals adapted supply chain strategies during and after the COVID-19 pandemic. A semi-structured interview approach was selected to capture practitioner knowledge, context-dependent decisions and evolving mitigation practices. Purposeful sampling was used to select participants with direct experience managing supply chain disruptions before, during and after COVID-19. Exempt approval was obtained from Purdue University's Institutional Review Board under protocol IRB 2024-927. All participants were informed of the study's purpose, data handling procedures and their rights, including the right to withdraw voluntarily and the confidentiality of their data. Consent was obtained prior to recording interviews. Interview protocols included predefined open-ended questions focused on procurement decisions, mitigation strategies, supplier management, schedule impacts and adaptation measures under disruption conditions. Interviews were manually reviewed, sorted, transcribed, segmented and coded to identify recurring themes, parent categories and child decision alternatives. Thematic patterns identified from the interviews were then organized into a structured decision tree framework based on project conditions, disruption timing and procurement response strategies.

Following the data processing of the interview data, a decision tree was developed through hierarchical branching and model validation. Figure 1 presents a structured flowchart outlining each methodological phase, which summarizes the progression from research design to model validation, beginning with participant selection and interview procedures, followed by transcript structuring and thematic coding and culminating in the development and evaluation of the decision tree model.

Validation of the decision tree model utilized a randomized 70/30 training and testing split and evaluated model performance using accuracy, precision, recall and F1-score metrics. Quantitative statistical procedures were limited to model validation and classification performance evaluation, while qualitative analysis focused on thematic interpretation and decision categorization derived from expert interview data.

For the interview part of the study, industry experts were purposively selected from general contractors, EPC firms and subcontractors within the network of the researchers and invited to the study via email. All participants had more than eight years of professional experience and had been directly involved in supply chain decision-making before, during and/or after the pandemic. Selection criteria of the participants emphasized diversity across project types (e.g. industrial, life sciences, aviation and federal projects) and representation from both mid-size firms and Engineering News Record – ENR Top 400 contractors. Interviews followed a structured guide with four sections, (1) introduction and informed consent, (2) demographics and professional role, (3) supply chain practices before, during and after COVID-19 and (4) post-pandemic reflections on process changes and resilience strategies. Preliminary informal conversations with industry professionals during Purdue University Bowen School of Construction career fairs, industry advisory council meetings and job-site visits informed the script design and helped identify early decision-making themes.

Interviews were conducted face-to-face or via Microsoft Teams, recorded with participant consent and transcribed using Microsoft's AI transcription tool. Transcripts were manually reviewed and edited by the researchers for accuracy and for removing all identifying information (e.g. names, company identifiers, project locations) to ensure confidentiality. A total of 11 experts were interviewed in the study based on data saturation, a key principle in qualitative research where saturation is reached when no new themes or decision-making strategies emerged from additional interviews. Saturation was evaluated iteratively throughout the coding process by comparing newly identified procurement strategies and thematic classifications against previously coded interviews. After the later interviews, recurring parent themes, mitigation approaches and procurement decision patterns stabilized with minimal emergence of new strategic categories, indicating sufficient thematic coverage for framework development. This aligns with Guest et al. (2006) and Marathe et al. (2017), who demonstrated that saturation often occurs within 10–12 interviews when the participant group shares domain-specific expertise. In construction-focused studies, saturation has similarly been observed within this range (Marathe et al., 2017).

While outreach was extended to additional experts, several qualified individuals were unable to participate due to confidentiality obligations and non-disclosure agreements, particularly regarding the pandemic-era decisions. This limitation reflects the sensitive and high-stakes nature of the topic, as similar studies on crisis management and supply chain resilience have reported comparable recruitment challenges when participants hold proprietary or commercially sensitive information (Craighead et al., 2020; Ivanov and Dolgui, 2021). Nevertheless, qualitative inquiry emphasizes depth over breadth, and a smaller number of information-rich cases can yield substantial analytical insight when participants possess extensive experience and decision authority (Guest et al., 2020; Patton, 2010). The selected participants represented diverse organizational roles, management responsibilities, project delivery environments and construction sectors, allowing the study to capture a broad range of mitigation strategies and procurement decision pathways despite the intentionally focused sample size. In this study, the selected experts provided diverse and detailed perspectives spanning organizational roles and project types, ensuring that the findings capture a representative range of procurement decision-making practices under BSE conditions in the construction industry.

Transcripts were imported into Microsoft Excel and structured using a standardized tabular format that included five columns: line number, interview ID, topic tag (e.g. “Pre-Pandemic,” “Post-Pandemic”), speaker role (Interviewer or Interviewee) and raw transcript text. This structure ensured traceability of each response, supported iterative analysis and enabled contextual cross-referencing across interviews.

Manual coding was conducted line-by-line following the principles of qualitative content analysis, where relevant excerpts were identified and grouped according to recurring decision-making patterns, mitigation strategies and criteria cited by participants. The process replicated NVivo's systematic coding framework manually using Excel filters, comments and pivot tables, an approach shown to be effective in small datasets when executed with rigor and consistency (Auld et al., 2007).

Codes were categorized into a hierarchical structure of parent themes and child themes, consistent with node-based classification methods. Parent themes represented broad categories such as Procurement Optimization or Supplier Management, while child themes captured specific strategies like Early Procurement, Dual Sourcing or Redesign to Fit Availability. Each strategy was coded not only for thematic content but also tagged with contextual variables (e.g. disruption timing, project phase, procurement phase) to support downstream structuring in the decision tree model (Figure 2).

Coding was performed iteratively, where themes were first identified from a subset of transcripts, then tested and refined against additional interviews to confirm recurrence and consistency. Formal inter-coder reliability testing was not conducted because coding and framework development were performed manually by the researchers using a structured thematic classification approach. However, coding consistency was maintained through repeated cross-checking of coded segments, iterative comparison of thematic classifications across interviews and continuous refinement of parent and child themes throughout the analysis process.

Guest et al. (2006) suggested that themes that emerge early and appear frequently across a range of participants serve as strong indicators of data sufficiency. This approach was used to validate saturation at the parent theme level, particularly where decision strategies appeared across multiple interviews with minimal variation in meaning. To ensure traceability and integrity, each coded segment remained anchored to its original line and speaker. Supporting notes were maintained in adjacent cells for interpretive consistency. This method follows Schreier’s (2018) recommendation for transparent data handling and rigorous theme development in qualitative studies. Classification consistency was validated by repeatedly cross-checking early codes and saturation markers. Additional verification was performed by revisiting previously coded transcripts during later coding stages to confirm consistency in thematic interpretation, strategy categorization and alignment between coded responses and decision tree pathways.

The manual method allowed for detailed, context-rich interpretation of responses, while maintaining structured data suitable for frequency analysis and integration into the decision tree. Instead of software-based coding, the process followed established principles of thematic analysis, using a transparent, traceable approach to theme development and categorization. Coding was conducted rigorously and iteratively, ensuring consistency and replicability through structured procedures that enabled cross-case comparisons and analytical depth. This approach reflects recommendations for manual qualitative analysis in small but information-rich datasets, as outlined by Auld et al. (2007) and Bazeley and Jackson (2013).

Using the structured dataset, a decision tree model was developed to map conditional decision-making during and after BSE disruptions. The root node reflected disruption timing, indicating whether the impact and mitigation action occurred during the peak of the BSE or post-BSE. The initial classification established the temporal context and helped distinguish between short-term reactive decisions and longer-term adjustments. From each disruption timing branch, the tree expanded to reflect the project phase, categorized into: (1) Early Construction/Bidding, where procurement flexibility allowed design and sourcing adjustments; (2) In Construction, where awarded contracts and ongoing fieldwork imposed constraints; and (3) Commissioning/Close-Out, where limited time and scope restricted available strategies.

Each project phase then branched into the same parent themes derived from interview data analysis, which represent overarching mitigation categories grounded in the coded responses. These themes include Procurement Optimization, Supplier Management, Contractual Flexibility, Resource Mobilization and Redesign and Adaptation. To maintain analytical traceability, these parent themes were directly aligned with the coding structure shown in Figure 2. Each parent theme was further subdivided into corresponding child themes, such as Early Procurement, Strategic Storage, Local Sourcing, Pre-Buyout of Long-Lead Items, Redesign to Fit Available Stock, Force Majeure Clauses, Dual Sourcing and Supplier Retention. Finally, leaf nodes captured specific mitigation actions drawn directly from interview data. Examples of these actions included: “Replaced 20KVA panel with two smaller units,” “Used emergency letters to prioritize delivery,” “Redesigned spec to accommodate alternate vendor” and “Procured and stored switchgear before award.” These terminal nodes contextualized practitioner responses to disruption and grounded the tree in project-specific realities.

The complete decision tree structure mirrored the conditional logic expressed by industry professionals, reflecting how procurement decisions evolved based on timing, phase and available mitigation strategies. This structured classification process allowed qualitative interview data to be systematically translated into hierarchical decision pathways suitable for scenario-based evaluation under highly uncertain BSE conditions. To assist readers in visualizing this multi-tiered structure, a simplified representation of the decision tree development process is included in Figure 3. This model served both as a descriptive framework and as the basis for validation using classification metrics described in the next section.

To evaluate the predictive accuracy of the decision tree model, a randomized 70-30 train-test split was applied to the full coded dataset (Japkowicz and Boukouvalas, 2024; Roḳaḥ and Maimon, 2015). The dataset consisted of structured rows extracted from interview transcripts, each representing a mitigation strategy with its corresponding parent theme, child theme, project phase and disruption timing (During or Post BSE). The split was executed in Microsoft Excel using a random number generator to assign each case to either the training or testing subset. Stratified sampling ensured proportional representation of the various strategy categories in both subsets, preserving thematic diversity while minimizing sampling bias (Japkowicz and Boukouvalas, 2024; Roḳaḥ and Maimon, 2015). Because the dataset was qualitative and the sample size was limited, preserving proportional thematic representation across both subsets was critical to avoid overrepresentation or exclusion of recurring mitigation categories during validation.

The training set (70%) was used to construct the tree through hierarchical decision branching based on conditions observed in the interviews. The model followed the sequential structure outlined in Section 3.4. Disruption timing → project phase → parent theme → child theme → specific mitigation action.

The test set (30%) was reserved exclusively for validation to assess how accurately the trained model could classify new cases not seen during training. The validation process focused on evaluating whether the framework consistently reproduced thematic decision pathways and procurement classifications identified during the qualitative coding process rather than attempting to predict outcomes. To measure model performance, four standard classification metrics were applied:

1. Accuracy = TP/(TP + FP + FN) (1)

2. Precision = TP/(TP + FP) (2)

3. Recall = TP/(TP + FN) (3)

4. F1-Score = 2 × (Precision × Recall)/(Precision + Recall) (4)

where:

TP (True Positives): Cases where the model correctly predicted the actual mitigation theme.

FP (False Positives): Cases where the model predicted a theme that did not match the actual coded response.

FN (False Negatives): Cases where the model failed to correctly predict an existing strategy.

Because the validation process focused on thematic classification rather than binary prediction, True Negatives (TN) were not incorporated into the evaluation structure, consistent with the exploratory and qualitative nature of the framework.

These metrics provide a comprehensive evaluation of classification performance, especially when dealing with multi-class scenarios and imbalanced representation across categories (Japkowicz and Boukouvalas, 2024; Roḳaḥ and Maimon, 2015). The validation process was carried out manually by comparing the model's predicted classification at each decision node to the actual theme assigned during the coding process. Each case was traced from root to leaf in the decision tree and matched against the manually coded classification in the test dataset. This procedure allowed researchers to systematically flag matches (True Positives), incorrect assignments (False Positives) and missing predictions (False Negatives). Validation began at the parent theme level and continued through child themes and leaf node strategy descriptions. This granular verification provided both performance metrics and insight into which levels of the model held predictive strength, particularly relevant in exploratory decision-support applications, where structure and transparency are as important as output accuracy (Roḳaḥ and Maimon, 2015). The classification metrics were then used to evaluate the consistency in thematic decision mapping and framework structure rather than evaluating the predictive forecasting capability under inherently unpredictable BSE conditions of the model.

The study included interviews with 11 professionals actively engaged in procurement and supply chain decisions during the pandemic. Participants' experience ranged from 8 to over 30 years, with most operating in mid-to senior-level roles within construction firms that handle complex, long-lead equipment. Interviewees represented diverse market sectors, including advanced technology, life sciences, water infrastructure, industrial megaprojects, aviation and federal construction. Participants were affiliated with firms across a range of sizes and operational models, including several listed in the ENR Top 400 (2023) and ENR 500 (2024) rankings, alongside professionals from specialized mid-sized firms, as presented in Table 1.

The composition of the sample reflects both the strategic depth and operational diversity of procurement decision-making in the construction sector. Participants operated in varied procurement environments, ranging from tightly integrated EPC projects with early material lock-in to owner-furnished or design-build contexts that required continuous adaptation. Exposure to different contract structures, client pressures and supply chain risks allowed the study to capture a wide spectrum of decision responses during and after the pandemic. This contextual diversity strengthens the external validity of the findings and ensures that the resulting decision tree model reflects not only specific case knowledge but also patterns applicable across construction delivery models and organizational scales.

Interview analysis revealed a diverse range of mitigation strategies deployed by construction professionals to navigate supply chain disruptions caused by BSEs. Coded responses were structured into eleven parent themes and further refined into child themes, representing distinct procurement tactics. Strategies were temporally categorized by whether they were implemented during the active phase of the disruption (e.g. COVID-19) or post-BSE as part of longer-term organizational adaptation.

Table 2 summarizes the frequency of parent themes across both timeframes. During the BSE, strategies such as Redesign and Substitution (n = 10), Alternative Sourcing (n = 10) and Procurement Optimization (n = 8) dominated procurement decisions. In the post-BSE period, Procurement Optimization (n = 15) and Supplier Management (n = 11) remained central, while new emphases emerged in Risk Management (n = 8) and Technological Integration (n = 2). To improve interpretability, the table also includes examples of child themes and mitigation actions associated with each parent category, illustrating how high-level procurement strategies translated into project-specific implementation actions during and after BSE conditions.

Table 2 shows how broad procurement response categories were translated into actionable mitigation strategies during different disruption conditions. Parent themes represent the strategic intent of the response, while child themes and mitigation examples illustrate the operational tactics implemented by construction professionals. This hierarchical organization later served as the structural foundation for the decision tree model by linking disruption timing, project phase, strategic classification and specific procurement actions into traceable decision pathways.

During the BSE, reactive strategies focused on maintaining project momentum amid volatility. Contractors described how unpredictable lead times forced rapid, improvised responses. “Well, during that period … materials that used to take 20 weeks went to 30 weeks, and nobody really knew why …” (Interviewee F). Such uncertainty prompted structured planning for long-lead materials and equipment tracking (Interviewee G). These findings align with existing BSE literature that describes how construction organizations shifted from efficiency-oriented procurement approaches toward resilience-oriented adaptation strategies under conditions of extreme uncertainty and disruption (Dolgui and Ivanov, 2020; Ivanov, 2020).

Early procurement and stockpiling became critical. One construction executive noted, “We had a customer that wanted everything in the laydown yard three months before it was needed” (Interviewee J). This JIC approach replaced traditional JIT delivery methods, mitigating exposure to market volatility identified in prior literature, where firms intentionally increased inventory buffers and procurement flexibility to mitigate supply chain fragility during disruptive events (Brakman et al., 2020).

Contractors also implemented design modifications to substitute unavailable materials or equipment. “Maybe we can replace one piece of equipment with two smaller pieces … we basically analyze things from a design standpoint with the design team to see if there's a better thing” (Interviewee H). Similarly, electrical subcontractors altered layouts to accommodate alternate panels or wiring systems (Interviewee E).

Alternative sourcing emerged as a key mitigation strategy. “We did end up going to a different supplier and working through some design issues and redesign …” (Interview E). Others pivoted to local or regional suppliers to reduce dependence on international logistics (Interviewees J & I). Still, several professionals warned that changing suppliers mid-project could worsen delays. “If you were going to change suppliers two years into a three-year project … now you're at the back of the line …” (Interviewee K).

Supplier management and long-term partnerships proved equally vital. “Really, that's why it became more important to have trusted partners … accountable and transparent …” (Interviewee K). Interviewee J observed that “some supplier partnerships got really strong, others weakened a bit …,” highlighting how crisis pressure differentiated reliable vendors from fragile ones.

Contractual flexibility also played a major role. Force majeure clauses and renegotiated terms were invoked widely. “To claim force majeure, you had to get letters from every vendor showing their shutdown …” (Interviewee F). Adjusted agreements included fixed pricing, escalation clauses and delivery guarantees to stabilize commitments amid uncertainty.

Expediting and logistics control featured heavily in early responses. “Thousands of other GCs were trying to expedite … there's only so much they can do …” (Interviewee G). Firms paid premiums for manufacturing slots and freight priority, but many found that as global congestion worsened, expediting lost effectiveness.

Other teams mobilized internal fabrication and local labour resources. Interviewee D described direct inspections: “Go to the integrators and do some of the early inspections … so we knew what state the stuff was ‘gonna’ be in when it got there.” This proactive monitoring extended visibility into supplier production chains and minimized late-stage surprises.

In contrast, post-BSE strategies shifted from reactive measures toward long-term resilience building. Early procurement remained prevalent but was institutionalized through structured procurement plans, long-term supplier commitments and advance production slot reservations (Interviewees A, B & J). Companies expanded supplier evaluation systems and prequalification protocols, ensuring continuity and transparency across multi-tier networks (Interviewees E & I).

Supplier relationship management evolved from crisis collaboration to deliberate partnership strengthening. “More and more companies even consider going to work vertically integrated … make more things themselves and buy locally” (Interviewee J). Vertical integration and alliance contracts helped firms reduce dependency on external vendors and secure priority access to materials.

Risk management matured through revised contract models emphasizing shared escalation clauses and balanced exposure between contractors and suppliers. Internal supply-chain management divisions were established to oversee risk and vendor performance (Interviewee I).

Finally, technological integration emerged as a new theme in the post-BSE period. Interviewees referenced the introduction of AI-driven tracking, data-based supplier performance dashboards and improved logistics visibility tools (Interviewee H). Digitalization replaced manual expediting as firms sought proactive sensing of potential disruptions. These findings reflect the broader transition of the construction industry from conventional procurement practices toward increasingly technology-supported decision environments that incorporate digital monitoring, predictive planning and integrated supply chain coordination capabilities. The emergence of digital monitoring systems, AI-driven tracking tools and supplier visibility platforms reflects broader industry trends discussed in recent literature, where digitalization increasingly supports proactive coordination and resilience planning under uncertain environments (Raja Santhi and Muthuswamy, 2022; Gharaibeh et al., 2024).

In contrast to the short-term, reactive practices adopted during the pandemic, post-BSE strategies emphasized proactive monitoring, strategic supplier engagement and long-term capacity planning. Expediting, resequencing and other crisis-response strategies disappeared, replaced by preventive planning and technological support systems. Overall, the thematic comparison revealed an evolution from immediate survival tactics to institutionalized resilience frameworks.

A total of 113 distinct mitigation strategies were extracted from the interview dataset, capturing the full breadth of procurement decisions made in response to BSE disruptions. As described in Section 3.4, these strategies were systematically organized using a two-level thematic framework consisting of the eleven parent themes (presented in Table 2) and corresponding child themes that represented specific procurement tactics. This hierarchical classification, developed through qualitative coding and contextual tagging, reflected both thematic content and implementation conditions. Each of the 113 mitigation strategies was coded according to its implementation timing (during or post-BSE), project phase (early, in construction or commissioning), strategic classification (parent and child theme) and concrete decision action. This multidimensional dataset served as the analytical foundation for constructing the decision tree model introduced in the next section. By aligning mitigation strategies to specific phases and scenarios, the model enables structured, scenario-driven pathways to guide procurement decision-making under future supply chain disruptions.

The decision tree model organizes 113 distinct supply chain mitigation strategies into a hierarchical structure to support decision-making under the uncertainty of BSEs. Based on industry interviews, the model maps mitigation strategies as potential decisions through a five-level pathway: (1) disruption timing, (2) project phase, (3) parent theme, (4) child theme and (5) specific action. This architecture reflects actual decision logic used by construction professionals when responding to supply chain disruptions.

The first node distinguishes strategies implemented during a BSE from those adopted post-BSE. During-event strategies responded to immediate volatility, such as lead time instability, transportation bottlenecks or supplier failure. In contrast, post-BSE decisions focused on building institutional resilience, including structured procurement planning, revised contracts and technological systems for long-term risk mitigation. From each disruption timing branch, the tree separates strategies by project phase: early construction, in construction and commissioning/closeout. As Interviewee A emphasized, “We no longer have that luxury; we have to make decisions pretty early, bring partners on board …”. Later phases emphasized workaround strategies or stopgap substitutions. Interviewee H described mid-project redesign and explained: “Maybe we can replace one piece of equipment with two smaller pieces … just allowing us to be able to have workarounds.”

Following the project phase split, the model classifies decisions into eleven parent themes, such as Procurement Optimization, Supplier Management and Relationships, Alternative Sourcing, Redesign and Substitution and Risk Management and Contingency Planning. These represent broad strategic intents, derived from thematic saturation across interviews. For example, decisions grouped under Procurement Optimization included early buyout strategies and timeline buffering. As Interviewee B noted: “Those are automatically day one … one of the first we look at.”

Each parent theme breaks into child themes, offering tactical specificity. Under Alternative Sourcing, for instance, common child themes included “locally sourcing” and “switching to alternate contractors.” Interviewee I described this adjustment: “Developing global supplier networks in other countries for materials.” Similarly, under Redesign and Substitution, Interviewee E recounted altering specifications to accommodate a vendor change: “We did end up going to a different supplier and working through some design issues and redesign …”. The final level captures specific mitigation actions directly cited in interviews. These include responses such as “early identification and ordering of long-lead electrical components” (Interviewee B), “onsite inspections at integrator facilities” (Interviewee D) and “refurbishing salvaged equipment for temporary power needs” (Interviewee G). Such granular nodes preserve the real-world language and application of each strategy.

The model also reflects temporal evolution in strategy type. Some actions appeared exclusively during BSEs, including Expediting and Resource Mobilization tactics such as “paying expediting fees to accelerate delivery” (Interviewee G). Others, like Technological Integration that emerged only post-BSE, reflecting an industry pivot toward predictive planning and digital infrastructure (Interviewee H).

Figure 3 illustrates a simplified branch of the decision tree, from root to leaf. The complete model, including all 113 mitigation actions sorted by disruption timing, project phase, parent theme, child theme and action, is presented in  Appendix Figure A1 for decisions during the pandemic and  Appendix Figure A2 for post-pandemic.

Overall, the decision tree (Figure 4) translates qualitative insights into a structured format for scenario-based procurement decisions, offering construction professionals a framework to navigate future supply chain disruptions caused by BSEs. The framework was intentionally designed as a flexible and modular decision-support structure capable of adapting to different project conditions, procurement environments and disruption scenarios characterized by high uncertainty and evolving constraints, rather than functioning as a rigid predictive algorithm, since BSEs are highly unpredictable and difficult to forecast accurately.

The decision tree model underwent validation to assess its ability to predict mitigation strategies based on disruption timing, project phase and thematic classification. A 70-30 train-test data split was used, a common approach for evaluating supervised classification models, as outlined in Rokach and Maimon's decision-tree framework (Roḳaḥ and Maimon, 2015). The test dataset served as a blind set to evaluate predictive accuracy across different levels of decision abstraction. Each test case was manually verified against the predicted classification, following a structured error classification process: TP for correct predictions, FP for incorrect predictions and FN for missed strategies. The use of Accuracy, Precision, Recall and F1-Score reflects standard practices in classification model evaluation, consistent with guidance summarized by Japkowicz and Boukouvalas in their treatment of model assessment fundamentals (Japkowicz and Boukouvalas, 2024). Model performance was evaluated across three hierarchical levels: parent themes, child themes and detailed mitigation actions (leaf nodes). Table 3 summarizes the validation metrics by classification level.

At the parent theme level, the model demonstrated strong precision (0.83) and reasonable recall (0.69), resulting in an F1-score of 0.75. These values suggest that the model was effective in correctly identifying the broader categories of procurement strategy, validating its usefulness for high-level decision structuring. Performance declined considerably at the child theme and leaf node levels, with both showing low recall (0.26) and accuracy (0.21). Precision at these levels (0.50) indicates that while some predictions were correct, the model missed a large portion of actual strategies. The drop in F1-scores to 0.34–0.35 reflects the challenge of correctly classifying more specific or unique mitigation actions from limited interview data. This reduction in performance is important to interpret within the intended role of the framework. While parent-theme classifications captured recurring strategic procurement patterns across interviews, child themes and leaf nodes frequently represented highly contextual, project-specific actions that appeared only once or twice in the dataset. As a result, lower-level branches were more sensitive to dataset sparsity and strategy fragmentation, limiting the model's ability to generalize uncommon mitigation actions across validation cases.

This performance degradation stems from two main factors. First, limited data on several unique strategies constrained the model's ability to generalize. Many mitigation actions were proposed by only one or two participants, restricting their representation in both the training and test datasets. Second, overfitting occurred at the lower branches of the tree, where the model picked up on strategy differences that were mentioned only once or twice. Such declines in predictive performance at granular classification levels are consistent with known limitations of decision-tree models when facing highly fragmented or unbalanced datasets, as discussed in foundational machine-learning literature (Japkowicz and Boukouvalas, 2024; Roḳaḥ and Maimon, 2015).

The decision tree's strength lies in its ability to represent high-level decision logic and dominant procurement trends. For example, strategies grouped under Procurement Optimization or Supplier Management and Relationships were consistently predicted with accuracy, particularly when supported by recurring references across interviews. However, the model struggled to predict rare adaptations such as specialized redesign tactics, vendor-specific negotiation strategies or dual-function substitutions mentioned only once or twice. Despite these limitations, the validation confirms that the model serves as a valuable exploratory framework. Its high precision at the parent-theme level and ability to trace structured decision pathways from disruption timing to strategy implementation suggest strong utility for scenario-based planning under uncertainty.

The development and validation of the decision tree model in this study provide critical insights into how construction professionals approach procurement decision-making during BSEs. The model not only captures the logic applied under uncertainty in real cases but also offers a structured means of translating practitioner knowledge into a useable framework that supports consistency, transparency and adaptability. Results from interviews, strategy classification and model validation collectively highlight both the strengths of the framework and the areas requiring refinement.

The decision tree organizes 113 mitigation strategies into a hierarchical format reflecting how professionals conceptualize and execute procurement adjustments during BSEs. The model's architecture, beginning with disruption timing, branching through project phase and organizing strategies into parent and child themes, mirrors how supply chain decisions unfold in practice. For example, professionals first assess the temporal context (whether a disruption is currently active or has passed) before considering project-specific constraints such as phase or contractual flexibility. Interviewee A captured this logic, stating, “We no longer have that luxury, that we have to make decisions pretty early, bring partners on board,” pointing to the importance of disruption timing and project phase in early procurement strategies.

Structuring the tree around these variables provides clarity in moments when ad hoc decision-making typically dominates (Farooq et al., 2023). The framework improves visibility by organizing a large and diverse set of strategies into logical categories (Soto Ortiz, 2025). Parent themes like Procurement Optimization, Alternative Sourcing and Redesign and Substitution help group decisions by intent, while child themes differentiate specific tactics, such as early procurement or use of salvaged components. For example, under Redesign and Substitution, Interviewee H described the practical implications of supply constraints: “Maybe we can replace one piece of equipment with two smaller pieces … just allowing us to be able to have workarounds.” This quote highlights a strategy that was later formalized as a leaf node within the model. By capturing such contextual decisions and embedding them in the decision tree, the model provides not only a taxonomy but also a roadmap grounded in real project logic.

One of the clearest findings was the temporal shift in mitigation strategies between the active disruption phase and the post-BSE recovery. During disruptions, firms prioritized short-term measures such as expediting, substitution and stopgap solutions to maintain schedule adherence. This environment forced firms to adopt mitigation strategies under rapidly changing information, such as accelerating buyouts, switching suppliers or redesigning systems. These reactive decisions, while necessary, often involved trade-offs in cost, quality or long-term reliability (Smith and Fatorachian, 2023). The tree captures this through branches that represent during-BSE strategies, many of which reside under parent themes such as Expediting and Resource Mobilization or Redesign and Substitution. However, such strategies nearly disappeared from the post-BSE dataset, replaced by more sustainable and deliberate approaches.

Post-BSE strategies reflect institutional learning and a shift toward long-term resilience. Procurement Optimization, Supplier Management and Risk Management emerged as dominant themes in this period. For example, Interviewees referenced strategic supplier partnerships, vertical integration and predictive analytics as emerging pillars of post-BSE procurement resilience. Interviewee J noted, “More and more companies even consider going to work vertically integrated … make more things themselves and buy locally,” reinforcing the forward-looking nature of decisions during recovery phases. The decision tree captures this shift by anchoring each pathway to disruption timing, allowing professionals to tailor mitigation decisions according to whether they are actively managing a disruption or planning for its aftermath. This bifurcation is essential for structured scenario planning, where the decision environment evolves with changing market conditions, contract maturity and material availability (Phadnis et al., 2022).

Model validation showed strong results at the parent theme level, with an F1-score of 0.75 and precision of 0.83. This demonstrates the model's ability to correctly classify high-level strategic intents, such as whether a decision falls under Procurement Optimization or Supplier Management, even when context varies. These strong results indicate that despite the qualitative nature of the data, professional decision-making patterns align around recognizable categories. The decision tree serves as a mechanism to formalize these implicit logics into structured decision flows (Kuntz et al., 2013).

However, performance declined at deeper levels. At the child theme and leaf node levels, accuracy and recall fell to 0.21 and 0.26, respectively. This reduction is not unexpected given the sparsity of unique strategies across the dataset. Many mitigation actions, such as using refurbished equipment (Interviewee G) or performing onsite inspections to confirm production readiness (Interviewee D), were cited by only one or two interviewees. This underrepresentation limited the model's ability to generalize these actions during testing, exposing a known limitation of decision trees when dealing with highly specific or infrequently cited categories (Chaabane et al., 2017).

Nonetheless, this finding reinforces the importance of using the model as a planning and scenario evaluation tool rather than a prediction engine (Cordova-Pozo and Rouwette, 2023). Rather than functioning as a prediction model for specific procurement actions, the decision tree was designed primarily as a structured decision-support framework under disruptions caused by deeply uncertain events. Construction professionals have to interpret higher-level branches, particularly disruption timing, project phase and parent themes, as the most reliable component of the framework for guiding strategic procurement planning. Lower-level child themes and leaf nodes instead serve as structured scenario references that preserve practitioner knowledge and illustrate possible mitigation pathways observed during real BSE conditions. Its ability to represent the full range of options, even if some are rare or highly contextual, gives practitioners a menu of strategies to consider (Falasca, 2008). For instance, strategies like dual sourcing, vendor escalation clauses and retrofitting existing components appear as structured alternatives, enabling project teams to assess feasibility even if not widely used.

The decision tree was developed not only for classification but also for industry application. Interview feedback during validation indicated that the decision tree provides a structured alternative to reactive, intuition-based decision-making. The clarity offered by the root-to-leaf pathways facilitates rapid response planning under high-stakes conditions. Junior staff in particular benefited from having a transparent logic chain, enabling more consistent decision-making across varying levels of experience. This aligns with findings from Interviewee H, who described the benefit of having predefined options during project uncertainty: “It just allows us to be able to have workarounds. Professionals also emphasized the value of the tree for documentation and procurement transparency. By tracing decisions through the structured branches of the model, project teams can explain and justify procurement adjustments to internal stakeholders and clients. This was particularly important in high-risk environments where accountability and traceability were critical. Interviewee F described how invoking force majeure required formal documentation from every vendor: “To claim force majeure, you had to get letters from every vendor showing their shutdown …”.

The decision tree's modular structure allows it to be integrated into existing project management systems or procurement planning workflows. Teams can apply it as a dynamic framework by selecting branches relevant to the current project phase, disruption context and contract type and ignoring irrelevant pathways. As supply chain challenges evolve, the tree can be expanded or pruned to maintain alignment with best practices and emerging strategies. In practical terms, the model can be integrated into procurement management workflows, preconstruction planning sessions and risk workshops, supporting both reactive and proactive decision-making. The framework can also support ERP-based procurement tracking systems by organizing mitigation pathways according to disruption timing, material criticality, supplier conditions and procurement status. Similarly, BIM coordination environments and project management dashboards could incorporate the framework to improve visibility of long-lead risks, alternative sourcing scenarios, escalation triggers and mitigation progress across project phases. This application helps bridge the gap between theoretical resilience concepts and real-world execution by translating practitioner experiences into structured and reusable decision pathways for future emergency conditions. The emergence of Construction 5.0, characterized by the convergence of advanced digital technologies with human-centred and resilience-focused practices, further highlights the potential for structured decision-support frameworks like the one proposed in this study to be embedded within next-generation construction management environments (Yitmen et al., 2026).

During crises, the decision tree helps teams rapidly identify feasible response strategies, such as alternative sourcing, redesign or schedule resequencing, based on real cases validated by industry experts. In more stable conditions, it serves as a resilience planning tool, guiding early procurement, supplier diversification and contract structuring decisions. In addition, the framework supports simulated scenario evaluation exercises, allowing project teams to explore alternative procurement pathways and assess potential mitigation strategies under uncertain conditions before future disruptions materialize, recognizing that BSEs remain inherently unpredictable in timing, magnitude and impact.

For organizations with advanced or maturing digital capabilities, the framework can be embedded within Power BI or Tableau dashboards to provide visual analytics and real-time filtering or integrated into BIM, ERP and project management platforms such as Primavera P6, Autodesk Construction Cloud or Procore to connect procurement decisions with live project and supplier data. It can also interface with digital twin environments or AI-driven analytics platforms to simulate disruption scenarios and optimize decision pathways. These integrations link qualitative decision logic to quantitative risk indicators such as lead times, supplier reliability scores and escalation indices, enabling data-driven scenario testing and real-time visibility of supply chain vulnerabilities. At a broader level, the decision tree supports organizational learning and workforce development by codifying tacit expert knowledge into a replicable, transparent decision framework that promotes consistency and resilience across projects. Furthermore, the framework also provides potential educational and industry benchmarking for teaching and evaluating emergency procurement mitigation strategies, resilience planning and structured decision-making under highly uncertain BSE conditions.

This study addressed the lack of systematic tools for procurement decisions under extreme uncertainty by developing and validating a decision tree grounded in construction practice. Semi-structured interviews with 11 professionals revealed 113 unique mitigation strategies, which were classified and structured into a five-level decision tree framework. The model distinguishes between decisions made during and after disruptions, contextualizes them by project phase and organizes them thematically through parent and child categories. This structure reflects the actual flow of procurement logic under stress, from immediate adjustments to long-term planning. Professionals adapted strategies not only to external disruptions but also to internal constraints, such as contract terms and timeline pressures.

The results confirm that construction professionals adapted their procurement strategies not only to external shocks but also to internal project constraints, such as phase timing, contract type and risk exposure. During BSEs, decisions emphasized short-term continuity by expediting, redesigning and using alternative sourcing to maintain momentum strategies. Post-BSE strategies reflected long-term resilience planning, including early procurement protocols, digital tracking systems and supplier relationship strengthening. The decision tree captured both response logics, preserving the practical sequence and trade-offs involved in high-stakes procurement adjustments.

Validation results confirmed the decision tree's ability to reliably classify broad procurement strategies. At the parent-theme level, the model achieved an F1-score of 0.75, demonstrating strong precision in identifying high-level strategic intent. Performance declined at lower levels due to the limited representation of unique or highly specific strategies. Nonetheless, the model offers substantial value as a decision-support tool. Its structured format enables construction managers, procurement professionals and project teams to map scenarios, compare alternatives and document decision pathways in a transparent, replicable manner. The framework also provides industry organizations with a structured basis for developing resilience protocols, procurement contingency plans and internal supply chain risk-management procedures for future BSEs.

The study also identified specific tactics with strong potential for resilience building. These included early buyout of long-lead items, redesign to fit available stock, vertical integration and the use of refurbished or salvaged components. Contractual tools, such as escalation clauses and force majeure letters, also played a central role in risk allocation. Together, these strategies provide actionable options for future disruptions, supported by a structure that helps tailor them to specific project phases and delivery models. Procurement professionals can adapt the identified mitigation strategies according to project-specific constraints such as contextual conditions like schedule sensitivity, supplier availability, regional sourcing limitations, escalation exposure and client requirements.

While the model proved robust at capturing strategic logic, several limitations warrant discussion. First, validation was limited to the 113 coded strategies from 11 interviews. Although saturation was reached at the parent theme level, rare strategies were underrepresented, limiting prediction reliability at the leaf level. As noted in the performance summary, increased data volume and diversity would improve generalizability. Future research should expand the dataset across different regions, contract structures and project delivery methods to validate or refine existing branches. Second, the model's linear tree structure assumes conditional, rule-based progression from root to leaf. While effective for organizing decisions, this may not fully reflect the fluid, parallel nature of some project decisions, especially when strategies are executed simultaneously across multiple domains (e.g. redesign and sourcing). Additional studies may evaluate how the framework performs under varying regulatory environments, local supply chain structures and international procurement practices beyond the North American context represented in this study.

Future enhancements could explore ensemble methods (e.g. random forests) or hybrid models that integrate parallel pathways and probabilistic weighting. Future work may also evaluate integration of the framework into procurement management systems, digital twins and AI-supported supply chain monitoring tools to strengthen real-time decision support and resilience planning capabilities. Lastly, the decision tree should be used as a flexible guide rather than a rigid algorithm. Some professionals noted that prioritization must shift depending on project size, market volatility or internal capabilities. As an interviewee, I stated, “Some of these strategies we couldn't even consider in that type of project,” pointing to the need for adaptation. The tree is best viewed as a modular foundation that firms can customize using internal data, contractual realities and evolving risk assessments. This flexibility allows organizations to continuously adapt the framework as future BSEs introduce new supply chain conditions, procurement challenges, technological changes or regulatory constraints that are not fully predictable at the time of model development, particularly given the inherently uncertain and unpredictable nature of BSEs.

Expanding the dataset will improve performance at deeper levels of the tree, allowing for better generalization of rare strategies. Integrating parallel decision paths or weighting alternatives by risk profile may also enhance its flexibility and predictive utility. Overall, this study contributes a practical, data-informed framework that translates experiential knowledge into a structured, transparent and adaptable decision-support tool. Beyond procurement decision support, the framework contributes to future resilience-oriented research and industry preparedness efforts by providing a benchmark model for studying emergency mitigation, organizational adaptation and structured response planning under disruptive conditions in construction environments. The decision tree provides construction firms with a replicable foundation for navigating procurement under extreme uncertainty, strengthening supply chain resilience and institutionalizing lessons learned from BSEs into future project planning and organizational risk-management practices. The framework also establishes a foundation for future integration into digital construction management environments, including ERP-supported procurement systems, BIM coordination platforms and project risk-monitoring dashboards focused on resilience-oriented decision-making under uncertain conditions.

Microsoft's AI transcription tool was used exclusively for interview transcription. ChatGPT, based on GPT-4o (OpenAI), was used only for grammar/language editing and rewording assistance during manuscript refinement. No AI tools were used to generate research findings, interview content, analysis, references or conclusions.

Informed consent was obtained from all interview participants prior to data collection, and all responses were anonymized to ensure confidentiality. This study received exempt approval from the Purdue University IRB under protocol number IRB 2024-927.