Life cycle cost of communication towers: identification and hierarchical classification of influencing factors

Abstract

Communication towers are essential infrastructure in modern society, require effective life cycle cost (LCC) control for long-term sustainability. While existing research has focused on structural optimization and technological advancements, few studies address cost-related issues across various stages, perspectives, and stakeholders. A multi-source factor identification approach was employed, combining literature review, field surveys, expert interviews, brainstorming, and questionnaires. An integrated Grey-SNA-ISM model was then developed. The method identified 49 factors influencing LCC, integrated Grey-SNA centrality indicators, and established four network hierarchies, which include source-driven, intermediary conduction, collaborative central control, and outcome response. A nine-layer causal hierarchy categorizes factors into fundamental, transitional, and direct influences, revealing a multi-level transmission path from institutional drivers to execution feedback. The results indicate that inadequate policy standards, insufficient management experience, and environmental constraints serve as top-level drivers. These factors flow through intermediary stages like procurement errors and design deviations, leading to downstream cost risks, such as construction delays and operational failures. This creates a dynamic, three-tier cascading structure at the strategic, organizational, and execution levels. The integrated model proposed demonstrates significant adaptability in LCC modeling for communication towers, offering methodological support for factor classification and path identification. Additionally, it shows potential for application in other complex cost systems, facilitating precise interventions and cost control throughout the life cycle.

Introduction

Since the early twentieth century, communication towers and masts have provided essential support for global communication, fueled by the rapid development of radio communication and broadcasting^1,2,3. Communication towers are tall steel structures, as shown in Fig. 1, that have become more diverse in form and design with the advancement of modern communication technology, while their quantity and global demand continue to rise^4,5,6. As critical infrastructure for deploying 5G/6G technologies, many countries and regions have invested significantly in constructing communication towers. Effective allocation and management of these funds are essential for maintaining cost control throughout the communication tower planning, procurement, design, construction, and operation and maintenance (O&M) phases.

Fig. 1

Communication tower: (a) Quadrangular lattice tower; (b) Triangular tower; (c) Monopole tower.

LCC was introduced by the U.S. Department of Defense in 1960⁷ and refers to the total cost of a system over a specified period, encompassing design, construction, operation, maintenance, and disposal costs^8,9. Initially, system cost analysis focused primarily on design costs, intending to optimize overall costs through design modifications^10,11. However, subsequent studies revealed that costs incurred during other life cycle phases also significantly influence the total cost¹². Today, LCC control is widely applied across various engineering fields^13,14, including green buildings, railroads, reinforced concrete structures, residential buildings, bridges, and other infrastructure^{15,16,17,18,19}. Research on cost control in communication towers has predominantly concentrated on base station location deployment, equipment upgrades, component design optimization, and related aspects^20,21, with limited studies addressing the LCC of communication towers.

Although current research on communication towers has made progress in areas such as structural design, equipment layout, and localized cost optimization, most studies primarily focus on optimization analysis. These studies typically address a single stage or technical path. A multidimensional, multi-stage factor identification and integrated modeling framework with a life cycle perspective is lacking. Especially regarding the interaction mechanisms between cost-influencing factors, there is currently no mature theoretical framework or systematic model available. The absence of such structural research limits the scientific and forward-looking nature of LCC control strategies in practical engineering. There is an urgent need to develop an analytical system that balances factor complexity, structural logic, and decision applicability. This system is crucial for achieving refined management of the LCC of communication towers.

In recent years, intelligent modeling methods have been widely used in cost analysis and project risk management. Techniques such as Sugeno fuzzy logic, grey system theory (Grey), hybrid models, neural networks, and machine learning have been explored to improve cost estimation and decision-making. For example, Sugeno fuzzy logic has been used to predict project delays and contract amounts in infrastructure projects²². Grey relational analysis combined with neural networks has been applied in cost estimation for residential projects²³. Machine learning algorithms have shown strong data fitting capabilities in bridge construction²⁴ and building cost forecasting²⁵. Grey fuzzy hybrid models have also been used to predict costs in uncertain environments²⁶. While these methods enhance cost prediction and address nonlinear problems, they remain focused on predictive modeling. Although efforts have been made to identify and rank cost-influencing factors, they lack a systematic approach to understanding the causal relationships between these factors. In complex systems like communication towers, which involve multiple stages and factors, a single black-box predictive model is insufficient. It does not provide the transparency and path identification necessary for effective cost control.

The primary objective of this study is to propose a novel integrated Grey-SNA-ISM modeling framework. It aims to systematically identify and analyze the key influencing factors of LCC in communication towers. This model highlights the importance, causal relationships, and hierarchical structure among the influencing factors. It offers both theoretical support and practical insights for cost control across the entire life cycle of communication tower projects. Based on LCC theory, a multi-source approach is used to identify LCC factors. The integrated Grey-SNA-ISM model combines three components, each with a specific function. Grey reduces subjective bias in expert scoring. Social network analysis (SNA) uncovers the interrelationships and impact strength among factors. The interpretive structural model (ISM) analyzes the hierarchical structure of cost-influencing factors and clarifies causal relationships. This approach integrates the strengths of multiple models, offering strong structural interpretability, incorporating expert knowledge, and handling incomplete information. It is especially suited to address the complexity and interactions in LCC modeling of communication towers. This study aims to identify and analyze key LCC factors in communication towers, along with their hierarchical relationships, supporting both theoretical development and practical decision-making optimization in LCC management.

The LCC of communication towers is influenced by multiple factors, making systematic identification of all cost-influencing factors essential. These factors interact and transmit in complex ways. It is crucial to quantify their importance and uncover their causal relationships and structural characteristics. This will support scientific decision-making for full-process cost control. This paper addresses the following three core research questions:

• 1. What are the main factors influencing the LCC of communication towers?

• 2. How can the significance of these factors be quantified?

• 3. What are the interrelationships and transmission mechanisms among these factors?

To address these three questions, this paper proposes a research framework. First, a bibliometric analysis was performed to identify research gaps. Then, 49 cost-influencing factors were identified, modified, and screened using literature research, field investigation, expert interviews, brainstorming, and questionnaire surveys. Grey was combined with the SNA method to analyze the factors influencing the LCC of communication towers. The Grey effectively reduces the subjective bias of expert ratings, while the SNA method helps identify the causal relationships between factors. Finally, by integrating the ISM to organize the hierarchical relationships between cost-influencing factors, this paper provides both theoretical support and empirical evidence for optimizing the LCC of communication towers. The core contributions of this paper are as follows:

• 1. The life cycle of communication towers is divided into five phases: planning, service procurement, design, construction, and O&M. Procurement typically occurs before the design phase, distinguishing it from traditional engineering projects.

• 2. The factors influencing the LCC of communication towers were identified using literature research, field investigation, brainstorming, expert interviews, and questionnaire surveys. It fills a gap in project management within existing research.

• 3. The Grey-SNA-ISM combination model is proposed for identifying cost-influencing factors in the life cycle of communication towers. It analyzes the network relationships between these factors and reveals their causal and hierarchical relationships.

Literature review

Before beginning the literature review, this section provides a brief introduction to the thematic structure framework formed through keyword co-occurrence analysis and word cloud visualization. The specific analysis methods will be detailed in Section “Bibliometric and word cloud analysis”. This section focuses on presenting the graphical results and serves as the basis for the subsequent structure of the review.

As shown in the bibliometric results in Fig. 2, current research on communication towers can be categorized into 4 areas: structural design optimization, maintenance and inspection, electromagnetic radiation and environmental impact, and intelligent applications.A deeper analysis of these results will provide insight into the current state of communication tower research, allowing for the identification of research gaps and the initiation of further studies.

Fig. 2

Keyword cloud.

Review of research on communication towers

In recent years, several incidents of communication towers collapsing due to structural inadequacies have occurred domestically and internationally^27,28,29. As a critical infrastructure failure, communication tower collapses have significant economic and social impacts. Therefore, improving their structural performance is of great importance. New materials can enhance structural safety, reduce maintenance costs, and improve productivity^30,31,32,33. Effective parameters affecting lightning performance are analyzed, a 3D numerical model is developed based on the results, and the simulation results are experimentally validated, improving the lightning shielding area of the rod^34,35. A full-wave electromagnetic model is used to determine the optimal grounding configuration for specific situations based on numerous simulation experiments, offering new insights into improving lightning protection in communication tower grounding systems^36,37. Evaluating the lateral strength and stiffness of ultra-high fiber performance reinforced concrete circular flange bolted joints is essential. Finite element analysis can be used to analyze their force transfer mechanism, load-bearing capacity, and failure behavior under transverse cyclic loading. This approach can ultimately enhance the load-bearing capacity of communication towers^38,39. A finite element model of the communication tower was created, and a collapse simulation was performed to identify the tower’s weak points and collapse mechanisms, providing a solid foundation for research on feasible reinforcement methods^40,41.

In maintenance and inspection, communication towers’ performance is influenced by extreme environmental conditions, making timely and effective structural health monitoring essential for their regular operation^42,43. In practice, some damage is difficult to detect. Optimization algorithms can determine the relationship between unhealthy frequencies of communication towers and damage conditions, enabling timely detection of structural damage^44,45. Combined with frequency domain data, a hybrid Adaboost, Bagging, and RUSBoost algorithm can identify damage in communication towers and help develop a health monitoring system based on frequency domain response and innovative diagnostic techniques^46,47. A deformation monitoring method based on multi-source data fusion has been proposed to measure dynamic displacement accurately. By combining numerical simulation with model experiments, the method can accurately reconstruct low-frequency and high-frequency displacements of a single-tube communication tower^48,49.

With increasing demand, deploying mobile base stations is becoming more intensive, resulting in higher levels of electromagnetic radiation^50,51,52. Radiation pollution is recognized as the most significant environmental hazard, with growing concern surrounding it. Assessing the electromagnetic radiation generated by communication towers and developing scientific deployment plans that minimize impacts on human health and the environment is essential. The safety distances for occupational exposure of mobile communication base station workers were evaluated to ensure that RF exposure does not exceed fundamental safety limits under any scenario^53,54. Electromagnetic field strength measurements were conducted for indoor and outdoor masts, as well as communication base stations at building locations. The results showed that the maximum electromagnetic field strength of some indoor base stations exceeded the reference levels. These levels were set by the International Commission on Non-Ionizing Radiation Protection^55,56,57. Tandem measurements in both time and frequency domains, combined with experimental weighting, were used to obtain the real field strength, effectively representing the actual exposure value and laying the foundation for RF exposure measurement and prediction^58,59. A green deployment method is proposed to minimize average electromagnetic radiation while maintaining the user’s quality of service. This method employs an improved dolphin swarm algorithm combined with an enhanced two-species swarm strategy. The effectiveness of the method is demonstrated across various communication scenarios^60,61.

In intelligent applications, as traditional tower building methods mature, unmanned aerial vehicles, IoT technology, and other advanced technologies are integrated with communication towers^62,63,64 to maximize resource utilization and enhance the value of communication towers. The deployment methods of antennas are explored, and algorithms are optimized. Antenna beams are also adjusted to achieve significantly better performance when radar and communication are shared, compared to when they are separate. This approach will better integrate radar and communication functions, improving spectrum utilization and system performance^65,66,67. To address issues such as users being too far from the base station or excessive base station load, a new hybrid network architecture is proposed. This architecture uses drones as mobile base stations to offload data traffic from mobile terminals^68,69,70. A hybrid power supply system for mobile base stations is designed based on independent photovoltaics, wind turbines, and adiabatic compressed air energy storage, utilizing natural resources like sunlight and wind. Simulation results show that the power loss rate of this system is meagre, confirming its feasibility. This system improves resource utilization while reducing the impact of other factors on performance^71,72,73. Installing sensors on infrastructure and transmitting data to base stations via IoT technology ensures real-time information control and secure infrastructure failure warnings^74,75.

Review of research on LCC in communication towers

Current research on communication towers mainly focuses on equipment, energy, and deployment methods. Cost studies primarily focus on using new materials, performance improvements, optimizing the layout of communication towers, and indirectly reducing construction and maintenance costs. Given that communication base stations require air conditioning systems during transitional seasons, a gravity heat pipe heat exchanger is proposed to cool the towers. Various indicators have been measured, confirming that the system reduces air conditioning operating costs^76,77. To prevent service interruptions due to sudden natural disasters, battery resource allocation is optimized by considering the demand transfer and sleep mechanism of base stations. A stochastic programming model and optimization method are proposed, which have been shown to reduce total costs^78,79. Coverage requirements for communication towers vary by location. By combining genetic algorithms with antenna location data from geographic information systems, the distribution and layout of communication towers can be optimized to reduce costs⁸⁰. Recommendations for constructing communication towers based on antenna capacity requirements are provided. Low-capacity antennas are sufficient for rural areas, and constructing such base stations can reduce power consumption, reducing costs^81,82. A genetic algorithm-assisted controller is proposed to replace the power system of communication towers with proton exchange membrane fuel cells, reducing O&M costs^83,84. A study was conducted on the application of a new Glass Fiber Reinforced Polyurethane composite material in communication towers. Bending tests were performed to measure the horizontal displacement of communication towers during typhoon weather, confirming that using such materials reduces construction costs⁸⁵. Case studies confirm that the structural form of communication towers should consider the external environment to maximize performance and lifespan and optimize costs⁸⁶.

Research on LCC for communication towers is limited. However, LCC has been widely applied in infrastructure projects like transmission towers, bridges, and roads. Given the similarities between communication and transmission towers in structure and life cycle management, LCC analysis in communication towers can benefit from transmission tower studies. In these projects, LCC analysis evaluates costs at various life cycle phases^87,88, including support design, material selection, and maintenance strategies. By considering factors like support size, steel grade, corrosion, and vortex loss, it enables a comprehensive assessment of construction, operations, and long-term maintenance costs^89,90. Studies on recycling wind turbine blades show that using discarded blades as transmission tower supports can reduce steel use, lower carbon emissions, and cut LCC⁹¹. Similarly, LCC analysis of high-voltage transmission poles highlights the importance of material selection and its environmental impact on LCC^92,93. Technological advancements and equipment upgrades influence LCC by increasing initial costs but reducing operational and maintenance costs⁹⁴. LCC also helps assess the cost-effectiveness of design choices, technology selection, and future updates. In power distribution poles, evaluating technological updates and maintenance strategies aids in technology selection for transmission towers⁹⁵. These studies underscore the importance of LCC analysis in cost control, offering a solid foundation for communication tower projects.

Although communication towers can draw on research approaches used for other infrastructure in LCC analysis, they still face unique challenges in practical application. Communication towers are individual structures with scattered distribution⁹⁶, and their deployment locations are determined based on demand⁹⁷. These towers operate in diverse environments⁹⁸ and are highly affected by extreme conditions^99,100, which makes the acquisition, aggregation, and standardization of life cycle data difficult. The heterogeneity of communication system equipment is also significant, with frequent updates and iterations¹⁰¹. Given the rapid evolution of current communication technologies, cost analysis during the O&M phases is highly uncertain, which impacts the stability and predictability of LCC models. Moreover, many stages within the project life cycle are interrelated¹⁰². In communication tower construction projects, multiple sites are typically developed concurrently, creating a multi-point, task-overlapping structure. Activities such as procurement, transportation, and site construction occur simultaneously, blurring stage boundaries and complicating cost attribution. To improve cost assessments, the existing LCC analysis framework needs to be adjusted to fit the specific characteristics of communication tower projects.

Research gap

Significant progress has been made in several technical areas of communication tower research, including structural safety optimization, intelligent operation and maintenance, electromagnetic radiation control, and health monitoring. Additionally, some studies have started to address cost control issues in communication tower projects. However, a thorough review of the existing literature highlights several critical research gaps in LCC management that require further exploration.

Current research on communication towers lacks a systematic perspective. Existing literature primarily focuses on cost optimization during the construction phase, material selection, or economic analysis of single technical paths. There is a lack of research frameworks that systematically identify and integrate multi-stage cost-influencing factors from a life cycle perspective, which makes it difficult to fully support the practical needs of comprehensive cost management. The coverage of influencing factors and stages is limited. Existing studies often concentrate on a specific stage or type of cost composition, and do not encompass the entire life cycle of communication towers. Furthermore, no integrated identification method for multi-dimensional influencing factors has been established, making it challenging to reveal the cost linkage mechanisms between different stages. The modeling of factor relationships is insufficient. Current research generally remains at the level of listing or ranking the influencing factors by weight. There is a lack of systematic modeling of causal paths, logical hierarchies, and structural relationships between factors. It is difficult to identify key driving factors and their role mechanisms in the cost transmission network. Research on method integration is relatively lagging. A comparative analysis with LCC research outcomes in the transmission tower engineering field reveals notable differences. In terms of theoretical modeling depth, influencing factor identification accuracy, and structural logical expression capabilities, communication tower studies still lag significantly. Currently, related research on communication towers has limited practical guidance in engineering management and cost control. By learning from and adapting the LCC analysis framework used in transmission towers, we recognize key differences. Although communication towers and transmission towers share some structural and life cycle management similarities, communication towers face unique challenges in their specific application. Therefore, it is necessary to build a more refined and adaptable LCC analysis model.

In summary, existing research has proposed effective technical solutions for cost control in communication towers. However, communication tower construction involves multiple stages, including design, service procurement, transportation, construction, and O&M. Most studies, however, focus on single-stage technical optimization and lack systematic evaluations from a LCC analysis perspective. Research on LCC management for communication towers still lacks systematic, structural, and methodological breakthroughs. There is an urgent need to develop an analytical framework that integrates multiple methods. This framework should ensure both logical clarity and engineering adaptability. Its goal is to improve the scientific and refined management of LCC in communication tower projects.

Methodology

As shown in the framework flowchart of Fig. 3, the research framework in this paper consists of four sequential components, forming a clear and systematic process. The first step is bibliometric analysis and review, which synthesizes existing literature to identify gaps in LCC research for communication towers and to define the research subjects and dimensions. The second step is factor identification and extraction, which uses literature research, field surveys, and expert interviews to refine the influencing factors. The third step is factor modification and screening, where multiple rounds of refinement and screening are done using brainstorming and questionnaires to ensure the validity and adaptability of the indicator system. The fourth step is combined model analysis, where expert scoring data is processed for consistency using Grey, leading to the construction of an influence relationship matrix. SNA and ISM methods are then used to identify the network structure of factors and their causal and hierarchical paths. This process reflects a closed-loop logic from data collection, method selection, to model output, ensuring logical coherence and providing a theoretical foundation for subsequent cost optimization strategies.

Fig. 3

Framework process.

Bibliometric and word cloud analysis

Bibliometric analysis is an approach that synthesizes information from abstracts, keywords, and citations in published materials such as books and journals¹⁰³. The main research clusters and future research directions were identified using bibliometric tools¹⁰⁴. A search for literature topics in the Web of Science database was conducted using the following formula: TS = (“Communication tower” OR “Shared tower” OR “Mobile base station” OR “Communication base station” OR “Base station” OR “Base transceiver station” OR “NodeB” OR “eNodeB” OR “SingleRAN” OR “gNB” OR “Telecommunication tower” OR “Cell tower” OR “Radio tower” OR “Lattice tower”). A total of 25,287 documents were retrieved. To ensure accuracy in identifying the research gap, the initial search results were filtered using the following criteria: (1) Keywords such as “Communication tower”, “shared tower”, “mobile base station”, and “communication base station”. Additionally, only abstracts were included. (2) The period for the literature search was set from 2002 to 2024. (3) The field of study was set to engineering. (4) The language was set to Chinese and English. Finally, 15,682 documents were screened, exported as tab-delimited files, and imported into VOS Viewer software to generate the keyword word cloud, as shown in Fig. 2. The results of this analysis form the basis for structuring Section “Literature review” of the literature review and provide a reference direction for the subsequent identification of influencing factors.

Identification of cost-influencing factors

Literature research method

Databases such as CNKI and Web of Science were used to conduct multiple rounds of searches and screenings with keywords like "cost-influencing factors," "key influencing factors," and "SNA." The literature was limited to the engineering and construction fields. Preference was given to studies closely related to the communication tower implementation process, with empirical foundations and high impact factors. A total of 60 highly relevant studies were included, from which initial cost-influencing factors were identified. These 60 papers were ranked based on their impact factors, citation counts, and download volumes. The top 20% of these papers (12 papers) were selected as representative studies. These studies spanned nine types of engineering projects, including roads, housing, prefabricated buildings, and high-rise buildings. The distribution of these influencing factors is shown in Fig. 4.

Fig. 4

Distribution of literature on influencing factors.

As shown in Fig. 4, the first column lists the 12 sources of literature. The second column shows the number of influencing factors identified in each article: 35, 34, 8, 38, 27, 12, 17, 21, 26, 46, 12, and 20, totaling 296. The influencing factors were classified according to the life cycle of communication tower construction projects, with 23 factors in the planning phase, 29 in the service procurement phase, 31 in the design phase, 208 in the construction phase, and 5 in the reducing O&M phase. Similar influences with different expressions were integrated, resulting in 178 distinct factors. These include 16 factors in the planning phase, 21 in the service procurement phase, 14 in the design phase, 122 in the construction phase, and 5 in the O&M phase, totaling 178.

To identify the critical analysis factors, the frequency of the integrated influencing factors across the 12 articles was calculated, along with their total frequency and cumulative proportion. The ABC classification method divided all factors into three categories: A, B, and C. The specific steps are as follows:

• 1. Grouping the factors. After integration, the 178 factors with the same frequency of occurrence were grouped.

• 2. Calculation of total frequency.

  $$d_{t} = a_{t} \times c_{t}$$

  (1)

  where a_t denotes the frequency of occurrence, c_t denotes the number of factors in each group, and d_t denotes the total frequency of occurrence.

• 3. Calculation of proportion and cumulative proportion using Eqs. (2) and (3).

  $$e_{t} = d_{t} /296$$

  (2)
  $$f_{t} = \sum\limits_{t = 1}^{t} {e_{t} }$$

  (3)

  where e_t denotes the total frequency proportion to the number of influencing factors before integration, while f_t denotes the cumulative proportion.

• 4. Calculation ABC classification. According to the Pareto principle, the factors were categorized using the ABC classification method based on their cumulative frequency percentage. The factors were divided into three categories: A (0–20%), B (20–50%), and C (50–100%), with the impact degree decreasing sequentially. The classification results are shown in Fig. 5, with detailed data provided in Table 1.

Fig. 5

ABC classification.

Table 1 ABC classification calculation.

The ABC classification results show that the 178 identified influencing factors have a frequency distribution ranging from 1 to 10, and were ultimately divided into 9 groups. The specific frequencies, quantity proportions, and cumulative percentages for each group are presented in Fig. 5 and Table 1. Among these, Category A consists of primary influencing factors, Category B includes secondary influencing factors, and Category C represents general influencing factors. For deeper analysis, this study selected all Category A factors and half of the Category B factors, totaling 18 factors, as key research subjects, as detailed in Table 2.

Table 2 List of cost-influencing factors based on the literature research method.

Field investigation method

A field survey was conducted on 158 communication towers in Jiamusi City, Heilongjiang Province, China, covering various structural types and geographic conditions. The distribution of the sites is illustrated in Fig. 6. The surveyed areas spanned across 10 administrative districts. Data on tower parameters, wind force, geology, and other factors were collected on-site, and interviews were conducted with construction and maintenance personnel. In addition to the factors identified with lower frequency in the literature review, a total of 23 influencing factors were identified, as presented in Table 3.

Fig. 6

Heat map of communication tower distribution.

Table 3 List of cost-influencing factors based on field investigation method.

Expert interview method

To complement the gaps in the literature and field survey, expert interviews¹⁰⁵ were conducted to further identify cost-influencing factors. A total of 10 experts, scholars, and project managers were invited, as listed in Table 4. The interviews were conducted through phone calls and face-to-face meetings, focusing on project cost overrun cases and the characteristics of communication towers. As a result, 10 additional influencing factors were identified, as shown in Table 5.

Table 4 Expert interviews with relevant personnel for background information.

Table 5 List of cost-influencing factors based on expert interview method.

Modification of cost impact factors

Due to the limited research on communication tower costs, the initial identification of factors relied heavily on factors from other engineering projects. To address this, a brainstorming method was employed for targeted adjustments. Based on the factors identified in Tables 2, 3, and 5, a total of 54 factors were extracted. An online meeting was held with 10 experts, frontline personnel, and project owner representatives, as listed in Table 6, to adjust the factor names. After the meeting, the collected viewpoints were summarized, and 37 factor names were revised. For example, “poor contract management” was revised to “poor contract management at tower companies”, and several similar factors were merged. As a result, 53 cost-influencing factors were revised.

Table 6 Background information of personnel related to brainstorming method.

Based on the four methods mentioned above, the preliminary list of identified cost-influencing factors is provided in Appendix A.

Selection of cost-influencing factors

After developing a preliminary list of influencing factors, a questionnaire was used to gather opinions from professionals and practitioners in relevant fields to ensure the factors’ relevance and representativeness. This process verified the rationality of the identified factors and facilitated the screening and optimization of insignificant or redundant factors. The list was further refined through the questionnaire survey method¹⁰⁶, Derived from social science research methodologies, it is commonly used to collect expert opinions and assess influencing factors¹⁰⁷. The survey gathered expert perspectives on the factors, enhancing the accuracy of factor identification and providing a solid foundation for subsequent model construction.

Questionnaire survey method

Design, distribution, and collection of questionnaires

The questionnaire design and validation rely on multiple theoretical tools to ensure data accuracy and validity. The Likert five-point scale, introduced by Rensis Likert in 1932, is widely used in psychology, sociology, and management to assess subjective opinions^108,109,110. It standardizes evaluations of the probability and impact of each factor, converting subjective assessments into numerical values for further analysis. The comprehensive degree of influence, derived from the Multi-Attribute Decision-Making (MADM) framework¹¹¹, uses the weighted product method to quantify factors, balancing expert judgment with quantitative analysis¹¹². Reliability and consistency are ensured by Cronbach’s α coefficient, while the KMO value and Bartlett’s test confirm the data’s suitability for factor analysis¹¹³. These methods support subsequent factor modeling and hierarchical analysis.

This questionnaire aims to assess the likelihood of the identified cost-influencing factors occurring and their impact on project costs. First, evaluation criteria for cost-influencing factors were established. Then, the preliminarily identified cost-influencing factors were sorted by the comprehensive impact index, with inappropriate factors eliminated, leading to the formation of the final list. Finally, the reliability and validity of the questionnaire were tested using SPSS software to confirm the feasibility of the results.

The questionnaire consists of three parts: the first part introduces the project background and survey objectives, the second part assesses the probability of occurrence and the degree of impact of each cost-influencing factor on project costs using a Likert scale, and the third part collects relevant information from the interviewees.

The questionnaires were distributed online and offline, and N (N = 180) questionnaires were sent. Of these, 100 were electronic questionnaires, and 80 were paper questionnaires. A total of 153 questionnaires were collected, including 93 electronic and 60 paper questionnaires, resulting in an effective response rate of 85%.

Quantification of cost-influencing factors

This questionnaire assesses the likelihood of identified cost-influencing factors occurring and their impact on project costs. Establish evaluation criteria for cost-influencing factors, as outlined in Table 7. Using Eq. (4), calculate the influence degree index, determine the overall impact of each cost factor, and rank the factors. The statistical results are presented in Table 8.

$$R_{i} = \frac{{\sum\nolimits_{j = 1}^{n} {\chi_{ij} \gamma_{ij} } }}{N}$$

(4)

where R_i denotes the comprehensive degree of influence, χ_ij denotes the degree of influence on project costs, and γ_ij denotes the probability of occurrence.

Table 7 Evaluation criteria for importance index.

Table 8 Influencing degree index.

Reliability and validity tests

Reliability and validity tests were conducted using SPSS software, as shown in Table 9. The Cronbach’s alpha coefficients for the probability of occurrence and the degree of influence on costs were 0.932 and 0.877, falling between 0.7 and 1, indicating a high internal consistency and suitability for further analysis. Additionally, the approximate chi-square values of Bartlett’s test were 1378.365 and 1163.682, and the KMO values were 0.760 and 0.724, both greater than 0.7, with a significance level of 0. These results indicated a high correlation among the questionnaire indices, confirming its suitability for factor analysis and validating the questionnaire data.

Table 9 Reliability and validity.

Factor coding

Based on the data in Table 8, a threshold value of 0.3636, derived from the difference between the mean and standard deviation of the influence index, was used to screen the influencing factors. As presented in Table 10, 49 cost-influencing factors were identified and coded.

Table 10 Factor coding.

The identified factors were correlated with survey results from 158 communication tower projects in Jiamusi, Heilongjiang Province. A case mapping was performed to link specific factors with practical engineering applications. The analysis was also extended to other regions, highlighting both universal and regional variations of these factors. This enhances the practical foundation and applicability of the identified LCC influencing factors.

In the design phase, F12 was prominent in several Jiamusi projects, where incomplete site data caused conflicts with underground facilities, leading to delays and cost overruns. Similar issues were observed in Guizhou, where inaccurate layout drawings, due to complex terrain and low surveying accuracy, caused frequent design changes. While the impact varied by region, this factor consistently interfered with costs during implementation. In the construction phase, F19 was common in Jiamusi, where limited transportation access delayed key components. A similar situation occurred in Yunnan’s Nujiang region, where rugged terrain and transportation issues caused construction disruptions. This underscores the widespread challenges of material supply in remote areas, significantly impacting cost control during construction. In the O&M phase, F45 showed cross-regional relevance. In Jiamusi, quality defects, such as component loosening and corrosion, led to repeated repairs during the warranty period. Similarly, in Jiangsu’s coastal saline environment, structural defects from corrosion caused frequent maintenance tasks. These issues emphasize that inadequate quality control leads to lasting cost implications, with consistent cost consequences across regions.

In conclusion, the LCC influencing factors identified in this study apply not only to Jiamusi but also to other regions. While conditions and manifestations vary, the underlying cost impact logic remains consistent, confirming the adaptability and scalability of the factor system and model.

Analysis of cost-influencing factors based on Grey-SNA-ISM

Existing research on influencing factors commonly uses methods like analytic hierarchy process (AHP), importance-performance analysis (IPA), and structural equation modeling (SEM). AHP is effective for structured problems but relies on expert judgment, making it difficult to address complex interactions between factors¹¹⁴ IPA focuses on comparing factor importance and performance but overlooks factor interactions¹¹⁵. SEM is useful for verifying causal relationships but requires large sample sizes and is less effective with incomplete or subjective data¹¹⁶. As research complexity grows, the limitations of single methods in handling ambiguity and causal complexity are more apparent, prompting the use of integrated approaches for better systematization and adaptability.

For example, Grey-DEMATEL has been used to identify key barriers in automobile recycling and hydrogen energy transitions, demonstrating the Grey method’s ability to handle uncertainty^117,118. The Grey-DEMATEL-ISM framework has been applied in urban flood resilience research, revealing influence chains and hierarchical structures¹¹⁹. Additionally, integrating SNA with ISM has helped analyze key resource factors in urban Water-Energy-Food (WEF) sustainability, highlighting centrality indicators¹²⁰, and has been used to explore driving forces and dependencies among key factors¹²¹. These studies show the broad applicability and effectiveness of multi-method integration in complex systems modeling.

This study on LCC influencing factors for communication towers involves three key characteristics: reliance on expert judgment, complex interactions between factors, and a hierarchical structure. To address these, the study constructs the Grey-SNA-ISM combined model. Each method brings unique strengths in structural analysis, and their integration enhances complementarity. By applying integrated methods from existing literature, this study tailors the approach to the specific features of communication tower cost factors. The combined use of these methods strengthens the model’s explanatory power and practical guidance value.

As shown in the Grey-SNA-ISM model flowchart in Fig. 7, it clearly presents the specific processes involved in data processing, network analysis, hierarchical relationship construction, and comprehensive analysis. The factors influencing the LCC of communication towers exhibit complex internal relationships. This paper proposes the Grey-SNA-ISM combination model to identify the roles and relationship strengths and organize their hierarchical structure and causal relationships. Ultimately, key factors were obtained. The flowchart clearly illustrates the three key steps of the model. First, the degree of influence between cost-influencing factors was quantified based on expert experience, and the scoring matrix was processed using the Grey method. Then, the SNA method and UCINET6 software were used to calculate the network measurement indicators of factors. The intrinsic connections within the complex network were analyzed, and the hierarchical structure of factors was presented using the ISM method. Finally, the key factors are analyzed comprehensively, and the hierarchical relationships between the factors are validated.

Fig. 7

Grey-SNA-ISM process.

Construction of Grey-SNA model

Grey, introduced by Deng Julong in 1982, excels in handling small sample sizes and uncertain information. It improves expert evaluation consistency and reduces biases caused by subjective judgment, enhancing data stability and reliability^122,123. SNA studies the structural relationships and interactions between elements¹²⁴. In this study, Grey addresses the uncertainties in expert scoring, forming the Grey-SNA model. This model strengthens the consistency and reliability of evaluation data, supporting the identification and analysis of key cost-influencing factors.

Establishment of the grey number matrix

First, the 49 identified influencing factors were organized into a set.

$$G = \{ g_{1} ,g_{2} ,g_{3} \ldots g_{n} \}$$

(5)

where G denotes the set of cost-influencing factors, and g_i denotes each.

Then, λ (λ ∈ [1, 4]) experts are invited to evaluate the relationships among cost-influencing factors based on the project’s actual conditions and professional knowledge. This evaluation results in the n-th order initial direct influence matrix P = [p_ij]_n×n, where p_ij denotes the degree of direct influence of factor g_i on g_j. Scoring is conducted on a scale from 0 to 4. Specifically, 0, 1, 2, 3, and 4 represent no direct impact, extremely weak impact, weak impact, strong impact, and extremely strong impact, respectively.

Finally, the rating values in the initial direct influence matrix are transformed into the corresponding grey number intervals defined by the grey semantic scale to construct the grey number matrix C. Subsequently, the grey number interval of importance degree for each expert’s ratings is determined based on their authority. The importance levels of expert ratings range from not important, less important, generally important, more important, to very important. The grey semantic scale for expert ratings and weights are shown in Table 11.

Table 11 Grey semantic variables of expert ratings and weight.

Standardization and clarification of grey number matrix

The obtained grey number matrix C is normalized and clarified, with the clarity value used to reflect the concentration of the expert evaluation results, thus quantifying the uncertainty in their subjective judgments. Equations (6)–(8) are used to calculate the upper and lower limits of the expert scores. Subsequently, Eqs. (9) and (10) are applied to derive the standardized clarity values $Y_{ij}^{\lambda }$ and the clarity direct influence matrix $Z_{ij}^{\lambda }$ for each expert.

The obtained grey number matrix C is normalized and clarified, with the clarity value used to reflect the concentration of the expert evaluation results, thus quantifying the uncertainty in their subjective judgments. Formulas (6)–(8) are used to calculate the upper and lower limits of the expert scores, and formulas (9) and (10) are applied to clarify the obtained standardized grey numbers, yielding the standardized clarity values and the clarity direct influence matrix for each expert.

$$\overline{ \otimes } X_{ij}^{{\overline{\lambda } }} = \frac{{\overline{ \otimes } X_{ij}^{\lambda } - {\text{min}}\overline{ \otimes } X_{ij}^{\lambda } }}{{\Delta_{\min }^{\max } }}$$

(6)

$$\underline{ \otimes } X_{ij}^{{\overline{\lambda } }} = \frac{{\underline{ \otimes } X_{ij}^{\lambda } - {\text{min}}\underline{ \otimes } X_{ij}^{\lambda } }}{{\Delta_{\min }^{\max } }}$$

(7)

$$\Delta_{\min }^{\max } = \max \overline{ \otimes } X_{ij}^{\lambda } - \min \underline{ \otimes } X_{ij}^{\lambda }$$

(8)

$$Y_{ij}^{\lambda } = \frac{{\underline{ \otimes } X_{ij}^{{\overline{\lambda } }} (1 - \underline{ \otimes } X_{ij}^{{\overline{\lambda } }} ) + \overline{ \otimes } X_{ij}^{{\overline{\lambda } }} \times \overline{ \otimes } X_{ij}^{{\overline{\lambda } }} }}{{1 - \underline{ \otimes } X_{ij}^{{\overline{\lambda } }} + \overline{ \otimes } X_{ij}^{{\overline{\lambda } }} }}$$

(9)

$$Z_{ij}^{\lambda } = {\text{min}}\underline{ \otimes } X_{ij}^{\lambda } + Y_{ij}^{\lambda } \times \Delta_{\min }^{\max }$$

(10)

where $\overline{ \otimes } X_{ij}^{\lambda }$ and $\underline{ \otimes } X_{ij}^{\lambda }$ respectively denote the upper and lower limits of the initial rating results of the cost-influencing factors by the λ-th expert, $\overline{ \otimes } X_{ij}^{\lambda }$ and $\underline{ \otimes } X_{ij}^{\lambda }$ also denote the upper and lower limits after standardization. $\Delta_{\min }^{\max }$ denotes the difference between the maximum upper limit and the minimum lower limit of the initial evaluation results provided by all experts for the cost-influencing factors.

Expert weight direct impact matrix

Based on the authority of the experts, the grey number interval of importance is determined for each expert’s rating. Expert weights are clarified and standardized using Eqs. (6)–(10), Eqs. (11) and (12) are then applied to derive the processed direct influence matrix. This step effectively integrates the differences in expert subjective judgments, making the final evaluation matrix more robust and broadly representative.

$$Z_{ij}^{{}} = \mu_{1} Z_{ij}^{1} + \mu_{2} Z_{ij}^{2} + \cdots + \mu_{\lambda } Z_{ij}^{\lambda }$$

(11)

$$\sum\limits_{\lambda = 1}^{\lambda } {\mu_{\lambda } } = {1}$$

(12)

where $Z_{ij}^{\lambda }$ denotes the clarity direct influence matrix for the i-th row and j-th column influence factors of the λ-th expert, and μ_λ denotes the expert weights after clarification and standardization.

Standardized direct impact matrix

The direct impact matrix, after clarity and standardization of the weights, is normalized using Eq. (13).

$$E = \frac{1}{{\mathop {{\text{max}}}\limits_{1 \le i \le n} \sum\limits_{j = 1}^{n} {Z_{ij} } }} \times Z$$

(13)

where E denotes the direct influence matrix.

Calculation of the comprehensive impact matrix

The standardized direct impact matrix was calculated using Eq. (14) to obtain a composite impact matrix.

$$H = E + E^{2} + \cdots + E^{n} = E \times \frac{{I - E^{n - 1} }}{I - E} = E(I - E)^{ - 1}$$

(14)

where H denotes the comprehensive influence matrix, and I denotes the identity matrix.

Construction of the visual network and calculation of metrics

The integrated influence matrix was imported into Gephi software for social network visualization, as presented in Fig. 8. The model consists of 49 nodes and 2127 connections. The network consists of 49 nodes and 2127 edges. The nodes represent the various LCC influencing factors, while the edges denote the relationships between factors whose grey relational strength exceeds a specified threshold. To enhance the interpretability of the graphical structure, the size of each node corresponds to the centrality of the factor. The larger the node, the more central the factor is within the network. Node color is classified based on the weighted out-degree, with darker colors indicating a stronger influence of the factor on other nodes. The edge color is determined by the grey relational weight, with darker colors representing stronger relationships. Due to the complex interdependencies among the factors, further analysis is required.

Fig. 8

Visualization of the cost influence network model.

Using the calculation results of Eq. (14) and UCINET6 software, combined with Eqs. (15), (16), and (17), the degree centrality, betweenness centrality, and closeness centrality of the SNA method were calculated. Tables 12, 13, and 14 present the three types of individual network indicators.

$$C_{d} \left( v \right) = \deg \left( v \right)$$

(15)

$$C_{b} \left( v \right) = \sum\limits_{s \ne t \ne v,s < t} {\frac{{g_{st} \left( v \right)}}{{g_{st} }}}$$

(16)

$$C_{c} \left( v \right) = \left[ {\sum\limits_{v \ne t} {d(v,t)} } \right]^{ - 1}$$

(17)

where s, t, and v are three nodes, C_d(v) is the degree centrality of node v, and deg(v) is the number of edges directly connected to node v. C_b(v) is the midpoint of node v, g_st is the number of shortest paths between nodes s and t, and g_st(v) is the number of shortest paths connecting nodes s and t through node v. C_d(v) is the closeness centrality of node v, representing the shortest path distance between nodes v and t.

Table 12 Statistics of degree centrality.

Table 13 Statistics of betweenness centrality.

Table 14 Statistics of closeness centrality.

ISM model construction

ISM, introduced by Warfield in 1973, identifies and establishes structural relationships among factors using a systematic approach. It reorganizes complex, interdependent factors into a clear, hierarchical model based on expert cognition¹²⁵. In this study, ISM is applied after the Grey-SNA model identifies key influencing factors. It further clarifies their causal chains and hierarchy, providing a solid foundation for optimizing LCC management of communication towers.

Calculation of the adjacency and reachability matrices

The threshold is introduced and adjusted through literature analysis, with the threshold defined as the sum of the mean and standard deviation¹²⁶. To eliminate less significant factors in the comprehensive influence matrix H, the data in matrix H is processed according to Eq. (19), resulting in the adjacency matrix B. This approach enables a clearer hierarchical structure of cost-influencing factors and influence paths. Using Boolean matrix operations, the adjacency matrix B is added to the identity matrix I and iterated according to Boolean matrix rules until the transitivity condition is satisfied, yielding the final reachable matrix M.

$$TH = \alpha + \beta$$

(18)

$$b_{ij} = \left\{ \begin{gathered} 0_{{}} ,h_{ij} < TH \hfill \\ 1_{{}} ,h_{ij} \ge TH \hfill \\ \end{gathered} \right.$$

(19)

$$M = (B + I)^{k + 1} = (B + I)^{k} \ne (B + I)^{k - 1}$$

(20)

where TH is the threshold, α is the average value of all elements in matrix H, β is the standard deviation, h_ij is the element in the i-th row, j-th column of the composite influence matrix H, b_ij is the element in the i-th row and j-th column of the adjacency matrix B.

Identification of the reachable set, antecedent set, and intersection set

After obtaining the reachable matrix, the next step is to identify the causal path relationships between factors. To accomplish this, the concepts of reachable set, antecedent set, and intersection set are introduced. The elements in the reachable matrix are processed using Eqs. (21) and (22) to derive the reachable set R(g_i) and the antecedent set S(g_j) for the cost-influencing factors. Using Eq. (23), the intersection set Q(g_i) is obtained.

$$R(g_{i} ) = \left\{ {\left. {g_{j} } \right|g_{j} \in G,m_{ij} = 1} \right\}$$

(21)

$$S(g_{j} ) = \left\{ {\left. {g_{i} } \right|g_{i} \in G,m_{ji} = 1} \right\}$$

(22)

$$Q(g_{i} ) = \left\{ {\left. {g_{i} } \right|g_{i} \in G,R(g_{i} ) \cap S(g_{j} ) = R(g_{i} )} \right\}$$

(23)

where R(g_i) denotes the set of elements with a value of 1 in the row corresponding to a cost-influencing factor in the reachable matrix M, while S(g_j) denotes the set of elements with a value of 1 in the column corresponding to a factor in M, and Q(g_i) denotes the intersection of R(g_i) and S(g_j).

Construction of a multi-level hierarchical structure

Hierarchical processing is performed using Eq. (23). When the elements in R(g_i) and Q(g_j) are the same, these elements are extracted and placed in the same hierarchy. The rows and columns corresponding to the extracted cost-influencing factor in the reachability matrix are deleted, and the process is repeated until all influencing factors are extracted. Finally, a multi-level hierarchical structure of LCC influencing factors for communication towers is obtained, as presented in Fig. 9. This completes the hierarchical processing of the cost-influencing factors. Based on the integrated influence matrix processed by Grey, this paper combines the structural reconfiguration capabilities of SNA and ISM methods. It achieves network modeling, hierarchical modeling, and causal path identification of the cost-influencing factors in the communication tower life cycle. This approach enhances the interpretability and controllability of key factors in complex systems.

Fig. 9

Multi-level progressive hierarchy of factors influencing the LCC of communications towers.

Grey-SNA-ISM model

The Grey-SNA-ISM model integrates the Grey, SNA, and ISM methods to provide a multi-perspective analysis of influencing factors within complex systems. First, the Grey method was applied to process the scoring data, providing a reliable foundation for subsequent analysis. Then, the SNA method was used to identify relationships and relative importance among factors, revealing key nodes and their influence within the network through network metrics. Finally, the integrated influence matrix derived from the SNA method was used in the ISM method. This further organized the relationships into a clear multi-level hierarchical structure, demonstrating causal relationships and hierarchical pathways among factors. This integrated approach facilitates an in-depth analysis of the system’s key influencing factors and their interactions. It offers a complete perspective from relational networks to hierarchical causation, yielding more comprehensive research outcomes.

Discussion

Network relationship analysis

Degree centrality analysis

Degree centrality indicates the number of connections a node has with other nodes in a network and is generally divided into “in-degree” and “out-degree”. According to Table 12, high out-degree nodes like F42, F39, F2, and F7 are concentrated in the early planning and design phase. These factors form the foundation of the project cost structure and have strong dominance and spillover effects. Mistakes in early-stage decisions can trigger a chain reaction, leading to cost overruns in later stages, such as design errors, delays, and increased maintenance costs. Therefore, these factors should be prioritized in risk prevention efforts. In contrast, high in-degree factors like F45, F35, and F40 are mainly found in the construction and operation phases. These factors reflect the cumulative cost response effects of later stages. While they don’t directly drive other factors, they are influenced by early-stage errors, coordination imbalances, and improper technical paths. These factors should be considered as feedback components within the system, and any anomalies should be integrated into quality control and performance evaluation to monitor the health of system. The transmission path of cost-influencing factors flows from nodes with higher out-degree to those with higher in-degree. Using the data in Table 12, the transmission paths of cost-influencing factors for the life cycle of communication towers are mapped. This study constructs a two-dimensional quadrant diagram for cost-influencing factors using out-degree and in-degree indicators. This approach reveals the interaction mechanisms and transmission directions between factors within the system. As shown in Table 12, the mean values of the out-degree and in-degree are 0.348. Partitioning is performed using the mean as the boundary, as shown in Fig. 10.

Fig. 10

Relational transmission of cost-influencing factors.

As shown in Fig. 10, the factors in Zone A of the conduction diagram exhibit a higher out-degree and a lower in-degree, indicating that these factors are more likely to influence others while being less influenced by themselves. They dominate position in the overall network, belong to spontaneous factors, and serve as root cost-influencing factors. Factors in Zone B of the conduction diagram exhibit relatively small out-degrees and in-degrees, indicating weak abilities to influence or be influenced by others. Compared to other cost-influencing factors, their transitivity and correlation within the network are relatively low. Factors in Zone C of the conduction diagram exhibit relatively large out-degrees and in-degrees, indicating that they influence others and are also highly susceptible to influence, demonstrating both active and passive characteristics within the overall network. Factors in Zone D of the conduction diagram exhibit relatively small out-degrees and larger in-degrees, indicating that they are more likely to be influenced by others and occupy a passive position within the network. It can also be concluded that these factors are the outcomes of others.

As shown in Fig. 10, the factors in Zone A of the conduction diagram exhibit a higher out-degree and a lower in-degree, indicating that these factors are more likely to influence others while being less influenced by themselves. They dominate position in the overall network, belong to spontaneous factors, representing the proactive driving nodes in the system. Managers should prioritize directing resources towards these factors for source management. Factors in Zone B of the conduction diagram exhibit relatively small out-degrees and in-degrees, indicating weak abilities to influence or be influenced by others. Compared to other cost-influencing factors, their transitivity and correlation within the network are relatively low. Although their current impact is limited, their structural position should be dynamically monitored to prevent them from becoming bottlenecks in the system after structural reorganization Factors in Zone C of the conduction diagram exhibit relatively large out-degrees and in-degrees, indicating that they influence others and are also highly susceptible to influence, demonstrating both active and passive characteristics within the overall network. These factors are central to the system, often appearing in high-frequency areas like design changes and resource scheduling. Strengthening cross-departmental collaboration and real-time dynamic scheduling is essential. Factors in Zone D of the conduction diagram exhibit relatively small out-degrees and larger in-degrees, indicating that they are more likely to be influenced by others and occupy a passive position within the network. It can also be concluded that these factors are the outcomes of others. These factors represent the operational outcomes and performance at the operational stage. To trace their causes, they must be analyzed alongside antecedent data.

Based on the degree centrality analysis results, it can be concluded that the factors in Zone A are highly influential source factors. During the construction of communication tower projects, the owner should enhance their understanding of relevant policies and standards, rigorously evaluate the qualifications of participating units, and establish a robust management system. A comprehensive survey of the project’s environment and surrounding transportation conditions should be conducted, with detailed planning for future demand changes at the site and optimal routes for transporting materials, components, and equipment between base station sites. Advanced intelligent information systems should be fully utilized to minimize the likelihood of cost deviations at the source. The structural analysis of degree centrality reveals causal relationships and influence chains among factors in the LCC system of communication towers. It provides a theoretical basis for classification control and stage-specific interventions. The system intervention strategy, including source control, central coordination, and end-stage tracking, enhances cost control across each project phase and supports the shift from experience-based to structure-driven project management.

Betweenness centrality analysis

In SNA, betweenness centrality measures a node’s role as an “intermediary” for information or influence. It reflects a node’s position in transmission paths between others. Nodes with high betweenness centrality act as “bridges” or "links," playing a key role in coordination, control, and transmission. In the cost influence network of communication towers, failure or loss of control of these nodes can disrupt transmission or cause systemic amplification. Therefore, they are crucial for identification and management. As shown in Table 13, the arithmetic mean of betweenness centrality is 4.767. Factors with betweenness centrality higher than the mean have a prominent ability to influence other factors and play a significant role in transmitting influence between connected factors. n this study, factors like F20, F12, F19, and F28 show high betweenness centrality, indicating their role as intermediaries in influence paths. These factors span key stages, such as design, construction, and resource coordination, and serve as links between upstream and downstream factors. For instance, F12 is influenced by early planning and impacts construction layout, material organization, and schedule management. Similarly, F28 often forces construction teams to adjust plans, affecting multiple factors. While these factors may not directly influence others, their intermediary role can amplify systemic risks, so they should be prioritized for identification and early warning mechanisms. In contrast, factors like F21, F25, F45, and F48 have lower betweenness centrality, indicating their limited role in the path structure. Though less central, these factors may form bypasses or supplementary risk chains in certain situations, requiring flexible evaluation and real-time monitoring based on context.

Therefore, during the construction of the communication tower project, managers should reasonably plan and allocate the tasks of design units, communication operating companies, and material, component, and equipment suppliers. They must monitor the time nodes required to complete tasks within each site’s life cycle, actively coordinate with suppliers to address issues arising at construction sites promptly and ensure the project progresses according to the optimal plan. In summary, betweenness centrality analysis helps identify control hubs and reveals transmission pathways among stages and factors. For communication tower project management, high and medium betweenness centrality factors should be used to establish coordination mechanisms and node linkage response strategies. This will enhance process stability in construction, planning, and change response. Additionally, a regional risk monitoring mechanism for low-betweenness centrality factors is essential to prevent them from becoming risk nodes under abnormal conditions. By intervening based on betweenness structure, the overall stability and coordination efficiency of the cost system can be improved.

Closeness centrality analysis

Closeness centrality measures the distance between a factor and all others in the network, indicating information accessibility and response speed. In the communication tower LCC influence network, factors with high out-closeness centrality have shorter distances to other nodes, showing stronger transmission efficiency and stability. In contrast, factors with high in-closeness centrality are more influenced by others, indicating they are easier to control but less stable within the system. As shown in Table 14, the mean values of out-closeness centrality and in-closeness centrality are 83.409 and 16.146. Factors with out-closeness centrality above the mean reach other nodes at shorter distances, are less influenced by the overall network, exhibit strong stability, and are more likely to drive changes in other factors, making them relatively important. In contrast, factors with in-closeness centrality above the mean are more influenced by other factors in the network and exhibit weaker stability. Factors like F42, F39, F2, and F4 have significantly higher out-closeness centrality, indicating their strong ability to rapidly influence other system components. These factors act as high-accessibility control nodes, typically concentrated in early project stages or resource allocation. Adjustments to them have a strong impact on the overall system. Therefore, they should be prioritized for system intervention, especially during the initial evaluation and design phases. In contrast, factors like F45, F48, F47, and F38 show higher in-closeness centrality, making them more vulnerable to indirect influences. These factors are mainly found in the operation and maintenance or final delivery phases, heavily shaped by earlier decisions and upstream controls. They should be incorporated into the cost feedback and monitoring system as key indicators of system stability.

A comparison of out-closeness centrality and out-degree rankings in Tables 12 and 14 shows that there is a high consistency between these metrics. Factors with higher out-degree are generally high-accessibility nodes, validating the Grey-SNA model’s effectiveness in identifying cost drivers for communication towers. The coupled analysis of centrality metrics reveals the roles of key factors from different perspectives, ensuring the results are grounded in a solid theoretical framework and stable data. Closeness centrality analysis highlights the propagation efficiency and system volatility of factors, offering insights into structural stability and control mechanisms. In practice, a layered strategy based on out-closeness and in-closeness centrality should be developed, including prioritized interventions, collaborative control, and endpoint early warnings. The goal is dynamic cost risk management across the entire chain, from core sources to responsive endpoints.

Hierarchical causal analysis

ISM, as a hierarchical modeling tool, is capable of revealing the underlying causal relationships and transmission paths among the factors within a system. It aids in identifying control nodes and determining intervention priorities. As illustrated in Fig. 9, the LCC influencing factors of communication towers are divided into nine layers, further simplified into three categories: (i) fundamental influencing factors, (ii) transitional influencing factors, and (iii) direct influencing factors. This results in the construction of a multi-level causal chain of costs, from institutional drivers to execution feedback.

The fundamental influencing factors, including F42, F32, and F39, are located in the eighth and ninth layers of the system. As the deepest-level nodes within the system structure, these core variables define the operational boundaries and fundamental rules of the entire cost system. While these factors do not directly affect the cost of communication towers, they exert a long-term, sustained, and irreversible influence on the execution of various project phases through policy guidance, environmental adaptability, and the management decision-making process. These factors are the primary source of LCC risks for the project. Relevant policies, standards, and technical systems provide the normative foundation and critical support for communication tower construction, shaping component production, transportation, and construction processes. The tower company, as the project owner, can strengthen project control and improve cost management by enhancing its management experience and professional capabilities in technical and managerial aspects. Constraints from natural conditions and social environments affect all stages of communication tower construction, ultimately influencing costs. Addressing these root causes is crucial for controlling and optimizing the LCC of communication towers. In practical management, a high-level governance system should be established. This system should include mechanisms such as institutional pre-research, environmental adaptability assessments, and management capability maturity models. These measures will enhance system stability and ensure alignment with the policy environment from the source. Addressing these root causes is critical for controlling and optimizing the LCC of communication towers.

The 40 factors located between levels 2 and 7 are classified as transitional influencing factors. These factors serve as links between the fundamental and direct influencing factors, forming a pathway between the root causes and the final outcomes. The factors involved include F10, F44, F14, F26, and others. Fundamental influencing factors are transmitted through these transitional factors, triggering direct influencing factors and leading to cost changes. These factors have a complex structure and interwoven transmission paths. They represent both the amplification effects and the concentration of bottleneck nodes within the system. On one hand, they reflect the functionality of upstream institutional frameworks and resource allocation. On the other hand, they serve as triggers for changes at the downstream execution level. Therefore, in practical communication tower project management, it is essential to manage transitional influencing factors, managers must monitor and predict their transmission trends by organizing and analyzing real-time data. Timely measures should be taken to interrupt the propagation of these factors, achieving effective cost control. A three-dimensional dynamic control system should be established around this area, focusing on mid-level response, risk buffering, and early warning linkage. This system should rely on information technology tools for path tracking, trend forecasting, and node pre-control.

The direct influencing factors, including F18, F23, F24, F38, F41, and F48, are located in the first layer of the entire system. These factors exert the most immediate impact on the LCC of communication towers while also being influenced by other factors. These factors are the specific variables that ultimately influence cost outputs. They typically exhibit strong time sensitivity and high volatility, and they emerge as issues during the construction and operation phases, following early decisions and mid-term strategy implementation. The stability of material, equipment, and labor market prices is fundamental condition in the communication tower construction process. Price volatility has a particularly significant impact, directly affecting the project’s budget and execution plans. Effective distribution route planning, real-time monitoring of routes and on-site conditions, and efficient management of equipment installation are crucial for controlling construction costs. Additionally, during the O&M period, environmental changes, advancements in communication technology, and large-scale renovations or reconstructions of communication towers should not be overlooked. Therefore, a real-time scheduling mechanism and rapid response process should be established, oriented towards feedback. In the event of an anomaly, it should quickly be communicated to the mid-level system for coordinated action. This would enable a closed-loop cost control system that integrates both feedforward and feedback mechanisms.

Comprehensive analysis and identification of key cost-influencing factors

This study analyzes the role mechanisms of different factors in the cost network using the Grey-SNA-ISM model. It provides a multi-level cost management strategy, covering source control to end feedback. Additionally, the study proposes a three-tier governance logic, consisting of strategic guidance, middle-level collaboration, and frontline response. This research provides theoretical foundations and practical guidance for cost control and risk management in communication tower projects.

The model results highlight the importance of factors such as F39, F27, F14, F7, F6, F1, and F9. These factors relate to issues such as the environment, weather, personnel levels, planning, and surveying. They directly impact the structure and safety of communication towers. Existing literature emphasizes structural design optimization^30,33,37. It also highlights the use of health monitoring^46,48 and damage identification technologies^44,45 to ensure structural stability and safety. These measures indirectly reduce the cost of communication towers. This is consistent with the findings of this study. However, this study extends the perspective by emphasizing the profound influence of factors such as policy standards and long-term planning. Specifically, F42, F39, and F32 address issues related to policy standards, external constraints, and owner experience. These factors highlight the importance of non-technical elements and reveal their global role across different stages of project implementation, which differs from existing literature.

The model results also provide a clear mechanism for the transmission of factors. F42, with the highest out-degree and centrality, serves as a fundamental upstream factor. It gradually transmits through intermediary factors such as F14 and F7, ultimately triggering end-stage factors such as F45. Several key transmission paths are further revealed by the ISM hierarchical causal analysis, reflecting the gradual accumulation from institutional deficiencies to uncontrolled construction costs. These path identification results clearly indicate that the weakness of policy and technical standards causes problems in the early decision-making phase. It also amplifies cost risks through multiple stages, including design, construction, and procurement.

Additionally, the impact of F45 on costs was not adequately considered in the initial predictions. A re-examination of the Jiamusi City project revealed that quality issues surfaced during the warranty period, leading to additional repair and resource allocation costs. These issues had a significant impact on cost in the later stages. The project’s quality control should span the entire life cycle, with increased monitoring of quality issues, especially during the warranty period. This further emphasizes the importance of cost factor identification, relationship analysis, and causal path analysis at all stages of the life cycle for dynamic cost management in communication tower projects.

A summary of the Grey-SNA model indicators reveals that individual network metrics emphasize different dimensions and reflect importance only in specific aspects. Therefore, a comprehensive analysis of the three types of network metrics in the Grey-SNA model was conducted, resulting in stratification into four layers, as presented in Table 15.

Table 15 Stratification statistics of factors influencing the LCC of communication towers.

As shown in Table 15, the first layer of cost-influencing factors occupies a central position in the overall network and is likely to trigger other influencing factors. The second layer of cost-influencing factors induces other factors and serves as a critical transmitter of their effects. The third layer of cost-influencing factors is strongly influenced by others and plays a critical role in transmitting their effects. The fourth layer of cost-influencing factors is positioned at the periphery of the network, resulting from the evolution and accumulation of other factors. In summary, the first layer consists of key influencing factors, the second and third layers consist of general influencing factors, and the fourth layer consists of secondary influencing factors.

When combined with the ISM model, fundamental influencing factors are positioned at the first layer of the Grey-SNA model as the most critical factors. These factors exhibit high out-degree, low in-degree, low betweenness centrality, and high out-closeness centrality, indicating strong influence over other factors and resistance to being influenced.

Direct influencing factors are positioned at the fourth layer of the Grey-SNA model and classified as secondary influencing factors. These factors exhibit low out-degree, high in-degree, low betweenness centrality, and high in-closeness centrality, indicating they are primarily influenced by others and represent resultant factors formed through the evolution and accumulation of influences.

Most indirect influencing factors are positioned in the second and third layers of the Grey-SNA model, with fewer distributed in the first and fourth layers. Factors in the first and fourth layers of the Grey-SNA model are closer to the fundamental and direct influencing factors in the ISM hierarchy. These factors exhibit high betweenness centrality, act as bridges between fundamental and direct influencing factors, and play a key role in system transmission.

A comparison of the results from the ISM model and the Grey-SNA model shows consistency in the importance and hierarchy of factors. Integrating the Grey-SNA and ISM models validates the results and refines the hierarchical structure of network composite indices derived from the Grey-SNA model. This refinement highlights each factor’s position and importance within the system, further confirming the applicability and scientific validity of the Grey-SNA-ISM model.

The Grey-SNA-ISM model reveals key factors that occupy central roles in the network, confirming the model’s ability to explain causal mechanisms. Factors F42, F39, and F32 are positioned at higher levels and exhibit strong driving forces. In the Jiamusi project, F42 highlighted delays in the tower approval process, which disrupted construction schedules. Similarly, in a Tianjin base station project, the lack of supporting standards led to repeated reviews, impacting costs and timelines. F39 showed that extreme cold and wind in Jiamusi increased equipment failure rates and maintenance cycles. In Hainan, typhoon-related issues also raised costs for structural reinforcement and repairs, demonstrating regional cost deviations due to environmental factors. F32 reflected poor internal coordination in Jiamusi’s engineering management, causing resource duplication and delays. This issue was also observed in remote areas like Kashgar, Xinjiang. The model’s core factors align with key issues in real-world projects, confirming the theoretical value and strong interpretative power of the structural model. The Grey-SNA-ISM model illustrates the network and hierarchy of LCC factors, validating its practical applicability through multiple case studies.

Conclusions

This study proposes a Grey-SNA-ISM combined modeling approach for communication tower projects, which systematically constructs a hierarchical structural model of the LCC influencing factors. Through a multi-source factor identification approach, including literature review, field surveys, expert interviews, brainstorming, and questionnaires, 49 influencing factors were identified. By combining network relationship analysis with causal path modeling, the hierarchical relationships and transmission mechanisms of cost-driving factors were revealed. Additionally, key influencing factors were identified.

The core innovation of this study lies in the integration of methods and structural modeling. By combining Grey, SNA, and ISM, the study facilitates the representation of factor structures under conditions of expert cognitive uncertainty. Furthermore, a factor modeling framework has been developed to cover the entire life cycle of communication towers. This framework fills the theoretical gap in causal network modeling within LCC research. Compared to traditional listing-based identification methods, the proposed approach enhances the logical explanatory power of the model and provides greater practical guidance.

The model results demonstrate significant practical value for engineering applications. The identified factors reveal considerable cost sensitivity across the planning, service procurement, design, construction, and O&M phases of communication towers. Additionally, the model shows robust adaptability in cross-regional case studies. It offers structured decision support for owners, operators, and managers, enhancing full-process cost control, resource allocation, and risk forecasting. Furthermore, the research provides theoretical support for policy-making and the revision of industry standards, promoting the institutionalization of cost governance systems.

The limitations of this study include potential subjectivity in expert scoring. While Grey reduces bias, it may still impact the accuracy of the results. Although the model has been validated for adaptability, further validation is needed in dynamic environments. Future research could integrate BIM and IoT technologies to collect real-time data for parameter adjustment, thereby enhancing the intelligence and global capability of the model. The integration of big data and machine learning methods would strengthen prediction and decision-making capabilities. Additionally, the model could be extended to other areas, such as transmission towers, wind towers, and intelligent transportation systems, establishing a universal LCC analysis framework to improve cross-scenario transferability and scalability.