Transforming Digital Accounting: Big Data, IoT, and Industry 4.0 Technologies—A Comprehensive Survey

Abstract

(1) Background: The convergence of Big Data and the Internet of Things (IoT) is transforming digital accounting from retrospective documentation into real-time operational intelligence. This systematic review examines how Industry 4.0 technologies—artificial intelligence (AI), blockchain, edge computing, and digital twins—transform accounting practices through intelligent automation, continuous compliance, and predictive decision support. (2) Methods: The study synthesizes 176 peer-reviewed sources (2015–2025) selected using explicit inclusion criteria emphasizing empirical evidence. Thematic analysis across seven domains—conceptual foundations, system evolution, financial reporting, fraud detection, audit transformation, implementation challenges, and emerging technologies—employs systematic bias-reduction mechanisms to develop evidence-based theoretical propositions. (3) Results: Key findings document fraud detection accuracy improvements from 65–75% (rule-based) to 85–92% (machine learning), audit cycle reductions of 40–60% with coverage expansion from 5–10% sampling to 100% population analysis, and reconciliation effort decreases of 70–80% through triple-entry blockchain systems. Edge computing reduces processing latency by 40–75%, enabling compliance response within hours versus 24–72 h. Four propositions are established with empirical support: IoT-enabled reporting superiority (15–25% error reduction), AI-blockchain fraud detection advantage (60–70% loss reduction), edge computing compliance responsiveness (55–75% improvement), and GDPR-blockchain adoption barriers (67% of European institutions affected). Persistent challenges include cybersecurity threats (300% incident increase, $5.9 million average breach cost), workforce deficits (70–80% insufficient training), and implementation costs ($100,000–$1,000,000). (4) Conclusions: The research contributes a four-layer technology architecture and challenge-mitigation framework bridging technical capabilities with regulatory requirements. Future research must address quantum computing applications (5–10 years), decentralized finance accounting standards (2–5 years), digital twins with 30–40% forecast improvement potential (3–7 years), and ESG analytics frameworks (1–3 years). The findings demonstrate accounting’s fundamental transformation from historical record-keeping to predictive decision support.

1. Introduction

Industry 4.0, the fourth industrial revolution, fundamentally changes how organizations process financial data and perform their accounting functions (Karadag, 2024). By combining big data analytics, the internet of things (IoT), artificial intelligence (AI), blockchain, and edge computing, organizations can provide real-time financial intelligence, continuously monitor compliance, and make predictive decisions (Gandomi & Haider, 2015; Prasetianingrum & Sonjaya, 2024). Industry 4.0 differs from all other past technological transformations in that it has changed accounting from being an after-the-fact, periodic reporting function, to being a predictive, real-time financial ecosystem that can process over a million transactions per second (Khurana, 2020).

In the context of financial organizations, big data is characterized by five dimensions: volume (terabytes of transactional data on a daily basis); velocity (streaming real-time data from IoT networks); variety (structured ledger data combined with unstructured documents, images, and sensor feeds); veracity (quality assurance of data collected from various and possibly unreliable sources); and value (actionable intelligence derived through advanced analytics) (Modupe et al., 2024). Organizations are generating large amounts of financially relevant data from smart meters, RFID inventory systems, GPS-enabled logistics networks, and automated transactional platforms. Advanced analytics convert these data streams into granular, real-time insights about asset utilization, cost allocation and transactional patterns that were not available through prior conventional accounting practices (Hewa et al., 2021).

Financial Technology (FinTech) innovations have greatly enhanced the accounting transformation process and have produced quantifiable results. The implementation of Machine Learning techniques increases to more than 90% the accuracy of prediction in fraud detection decisions compared to the traditional 50–60% achieved by conven-tional methods (Preciado Martínez et al., 2025). Smart Contracts utilizing blockchain technology automate financial transactions and reduce the amount of time to process Accounts Payable by 40–60%, and also the time to complete reconciliations by 70–80% (Prasetianingrum & Sonjaya, 2024). Edge Computing reduces the time for a company to respond to regulatory violations by 40–75% when compared to centrally located cloud architectures, thus allowing companies to detect potential regulatory violations within hours vs. 24–72 h for batch processed systems (Bhimani & Willcocks, 2014). As such, these technologies collectively comprise a multi-layered infrastructure which supports the creation of intelligent, transparent and automated financial management.

Although there have been many technological advancements made in the accounting field, several barriers still exist to implementing new technologies. The number of cybersecurity attacks against financial data has increased 300% since 2019, with average breach costs for financial services increasing to $5.9 million (Cao et al., 2015). Conflicts between the blockchain’s requirement for immutability and the EU’s General Data Protection Regulation (GDPR) right to erasure provision have caused compliance uncertainty for 67% of Europe’s financial institutions and resulted in delayed implementation of Industry 4.0 solutions ranging from 12–18 months (Suyts et al., 2017). Furthermore, 70–80% of accounting professionals lack the necessary analytics training to effectively utilize new accounting technologies, while the cost of implementing these technologies ranges from $100,000 to $1,000,000, resulting in barriers to entry for small and medium-sized enterprises (Dagiliene & Kloviene, 2019). These barriers emphasize the need for comprehensive frameworks that bridge the gap between the technical capabilities of the new technologies and the regulatory requirements and organizational realities of financial institutions.

The purpose of this systematic review was to analyze how Big Data and IoT technologies transform digital accounting within the Industry 4.0 paradigm. The study synthesized 176 peer-reviewed articles published between 2015 and 2025 that met the researcher’s explicit inclusion criteria that emphasized empirical evidence and quantifiable outcomes. The studies were identified via systematic searches of the Scopus, Web of Science, IEEE Xplore, and Google Scholar databases using three selection criteria: (1) relevance—addressed direct application to accounting processes, Industry 4.0 frameworks, Big Data-IoT integration, or digital transformation in financial management; (2) methodological rigor—validated by peer-reviewed journals with articulated research methodologies, empirical evidence, or structured analytical frameworks; and (3) temporal currency—published between 2015 and 2025 with preference given to more recent publications addressing emerging technologies including edge computing, decentralized finance, and Environmental Social Governance (ESG) analytics. Mechanisms used to systematically reduce biases included conducting multiple database searches, ensuring balanced representation of both academia and industry, and having independent dual reviewers screen articles.

Three factors in the financial services sector have stimulated the preparation of this systematic review of research on digital financial services. The first is that the pace of the introduction of digitization to financial processes far outstrips researchers’ current knowledge of the implications of such processes and as such, their ability to develop theoretical models to support the adoption of new technologies. The second factor is that the regulatory environment for emerging technologies continues to present regulatory authorities around the world with challenges in developing compliant frameworks, thus there is an urgent need for research that addresses the relationship between the technological capability of financial organizations and regulatory compliance. The third factor is the rapid pace of financial service organization’s digital transformation and the resulting need for them to rapidly implement real time automated financial systems that provide for business continuity and stake holder transparency.

This systematic review has achieved three key research objectives. These are: (i) to synthesize evidence regarding the combination of artificial intelligence (AI), blockchain, edge computing and digital twins and how these technologies may be used to enhance financial reporting, auditing, fraud detection and compliance functions using quantifiable measures; (ii) to identify potential barriers to the successful implementation of the technologies identified above and to propose evidence based solutions to mitigate these barriers; and (iii) to derive testable propositions to examine trade offs associated with the use of various technologies. For example, potential trade offs include: (a) whether the benefits of utilizing blockchain to enhance privacy protection outweigh the costs associated with complying with GDPR regulations, or (b) whether the benefits of utilizing AI in fraud detection outweigh the costs associated with explaining the logic underlying the decisions made by AI.

The systematic review will make a contribution to both the FinTech and accounting literature in four ways. First, the review will complete a systematic synthesis of research that examines the use of big data and emerging technologies to integrate digital accounting systems within the Industry 4.0 paradigm and document the quantifiable transformations experienced in fraud detection (85–92%), audit cycles (40–60%) and reconciliation efforts (70–80%). Second, the review will propose a four layer architectural model that illustrates the integration of Internet of Things (IoT) data collection, edge cloud processing, AI-blockchain intelligence and accounting system applications within a unified financial ecosystem. Third, the review will propose a challenge-mitigation model that assesses the potential risks posed by cybersecurity threats (300% increase in risk and estimated cost of $5.9 million), regulatory conflict (potential institutional impact of 67%), and skill deficits among employees (estimated skills deficits of 70–80%) with evidence-based mitigation strategies that achieve a 78% reduction in attack surface and an improvement of 40–60% in terms of fraud detection accuracy. Fourth, the review will develop four theoretically derived propositions supported by quantitative evidence and provide testable hypotheses for future empirical research.

The present article is organized as follows: Section 2 outlines the systematic review methodology; Section 3 outlines the Industry 4.0 conceptual framework and four-layer technology architecture; Section 4 outlines the use of Big Data in financial reporting, auditing, fraud detection, and blockchain accounting; Section 5 outlines the current state of industry-wide implementations with quantifiable measures of success across the banking, insurance, e-commerce, and government sectors; Section 6 outlines the challenges associated with implementing these technologies and the mitigation strategies to overcome these challenges; Section 7 outlines potential areas of future research including quantum computing, decentralized finance, digital twins, and environmental social governance (ESG) analytics; and Section 8 summarizes the contributions and theoretical propositions developed in the study.

Research Contributions and Positioning

While prior systematic reviews have examined individual technologies in accounting contexts—including Appelbaum et al.’s (2017) analysis of Big Data analytics in auditing, Dai and Vasarhelyi’s (2017) examination of blockchain applications in continuous auditing, and Moll and Yigitbasioglu’s (2019) comprehensive review of accounting information systems—this review uniquely synthesizes the technological convergence across the complete Industry 4.0 stack. The present study demonstrates how integrated multi-technology implementations yield synergistic outcomes that exceed the sum of individual technology deployments.

This systematic review makes three distinct contributions to the literature, prioritized by significance:

Primary Contribution: Integrated Technology Architecture. This review proposes and validates a four-layer architectural framework that uniquely integrates IoT data collection, edge-cloud processing, AI-blockchain intelligence, and accounting applications within a unified ecosystem. Unlike prior reviews that treat these technologies in isolation, this framework demonstrates the bidirectional data flows and synergistic relationships that enable real-time operational intelligence. The architecture advances understanding beyond technology description to mechanism explanation—clarifying how and why technological integration produces documented outcomes.

Secondary Contribution: Challenge-Mitigation Framework. The review develops an evidence-based framework linking specific technological barriers—including cybersecurity threats (300% incident increase), regulatory conflicts (67% institutional impact), and workforce deficits (70–80% insufficient training)—to quantified mitigation strategies. This framework bridges the gap between documented technical capabilities and practical implementation realities, providing actionable guidance absent from prior reviews.

Tertiary Contribution: Testable Propositions. Four theoretically-grounded propositions are derived from cross-study synthesis, each specifying independent variables (technology configurations), dependent variables (accounting outcomes), and hypothesized effect directions. These propositions are explicitly framed as testable hypotheses for future empirical research, advancing the field beyond descriptive documentation toward theory development and empirical validation.

Table 1. Positioning Relative to Existing Reviews.

Review Study	Technology Focus	Primary Contribution	Gap Addressed by Present Review
Appelbaum et al. (2017)	Big Data analytics	Audit data analytics framework	Integration with IoT, blockchain, edge computing
Dai and Vasarhelyi (2017)	Blockchain	Continuous auditing implications	Multi-technology integration; quantified outcomes
Moll and Yigitbasioglu (2019)	AIS broadly	Comprehensive AIS research agenda	Industry 4.0 technologies; implementation evidence
Kokina and Davenport (2017)	AI/Cognitive tech	Early AI adoption in accounting	Updated performance metrics; challenge-mitigation framework
Present Review	Integrated Industry 4.0 stack	Four-layer architecture; challenge-mitigation framework; testable propositions	—

positions these contributions relative to existing review studies in digital accounting and accounting information systems, clarifying what is conceptually new beyond the breadth of coverage.

2. Methodology

This paper employs an analytical structured-narrative review process as the systematic methodology to investigate the integration of Big Data and IoT within digital accounting systems, in order to provide a comprehensive overview of technological enablers, areas of applications and challenges for implementing the technology that are transforming accounting practice to Industry 4.0 standards.

2.1. Search Strategy and Data Sources

Academic sources were retrieved from five major databases: Scopus, Web of Science, IEEE Xplore, Google Scholar, and ScienceDirect. The search was conducted between January and June 2025, covering publications from January 2015 to March 2025 to capture both foundational developments and emerging innovations in financial technology (

Table 2. Database-Specific Search Results.

Database	Search Date	Initial Results	After Filters
Scopus	20 January 2025	142	87
Web of Science	25 January 2025	118	72
IEEE Xplore	5 February 2025	67	45
ScienceDirect	15 February 2025	52	31
Google Scholar	1 March 2025	89	58
Total		468	293

). The search strategy employed Boolean keyword combinations designed to capture the interdisciplinary nature of the research domain across six thematic categories:

• Category 1: Industry 4.0 and Digital Accounting Foundations (Section 3) Keyword combination: “Big Data AND digital accounting,” “IoT AND financial reporting,” “real-time accounting AND Industry 4.0,” “Internet of Things AND accounting systems,” “smart accounting AND cyber-physical systems”
• Category 2: Big Data Analytics and System Evolution (Section 4) Keyword combination: “Big Data analytics AND financial reporting,” “machine learning AND accounting,” “AI AND fraud detection,” “continuous auditing AND data analytics,” “real-time financial analytics”
• Category 3: Blockchain and Distributed Ledger Applications (Section 4 and Section 5) Keyword combination: “blockchain AND auditing,” “smart contracts AND financial transactions,” “distributed ledger AND accounting,” “triple-entry accounting AND blockchain,” “blockchain AND audit trail”
• Category 4: Industry Implementations and Compliance (Section 5) Keyword combination: “fraud detection AND banking,” “KYC AND automation,” “AML AND machine learning,” “tax compliance AND Big Data,” “regulatory technology AND finance”
• Category 5: Challenges and Implementation Barriers (Section 6) Keyword combination: “GDPR AND blockchain,” “cybersecurity AND financial data,” “data quality AND accounting systems,” “edge computing AND compliance,” “cloud computing AND financial systems”
• Category 6: Emerging Technologies and Future Directions (Section 7) Keyword combination: “DeFi AND accounting,” “digital twins AND financial simulation,” “quantum computing AND finance,” “ESG AND analytics,” “sustainability reporting AND Big Data”

The complete, database-specific search strings with Boolean operators, field tags, and filters are provided in Supplementary Table S1 to ensure full reproducibility. Searches were conducted between 15 January and 30 June 2025, with database access dates documented for each source.

2.2. Inclusion and Exclusion Criteria

The two-stage screening process helped establish the relevance and quality of the included sources by establishing what would be considered relevant for inclusion as well as of quality. The inclusion criteria were based upon whether a publication addressed either digital accounting practices, or Industry 4.0 framework, provided empirical evidence (with quantitative results), were peer reviewed academic publications, written in English, from 2015–2025, and documented measurable implementation results. The exclusion criteria excluded opinion-based commentaries, publications that lacked empirical evidence, had limited relevance to the application of finance, were duplicate publications, and studies focused solely on non-financial applications (

Table 3. Source Type Justification.

Source Type	Count (%)	Justification for Inclusion	Quality Criteria Applied
Peer-reviewed journal articles	118 (67%)	Primary evidence base; established peer review ensuring methodological rigor	Indexed in Scopus/Web of Science; journal impact factor ≥ 1.0 or ABDC/ABS ranked
Conference proceedings	38 (21.6%)	Emerging research on rapidly evolving technologies not yet in journals; captures recent developments	IEEE/ACM tier-1 venues; CORE A/A* ranked conferences; acceptance rate < 30% where available
Technical reports with methodology	20 (11.4%)	Industry implementation data unavailable in academic literature; practitioner evidence essential for applied review	From recognized professional bodies (Big Four firms, AICPA, ISACA, CIMA); documented methodology; verifiable data sources

Note. CORE A/A* indicates sources ranked as either A* (flagship tier) or A (excellent tier) in the Computing Research and Education Association of Australasia ranking system.

).

2.3. Screening Process and Final Corpus

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (Page et al., 2021). The screening process proceeded through four sequential stages, with reconciled numbers reported in the PRISMA flow diagram (Figure 1).

Stage 1: Identification. Database searches identified 293 potentially relevant records after applying date and language filters: Scopus (n = 87), Web of Science (n = 72), IEEE Xplore (n = 45), ScienceDirect (n = 31), and Google Scholar (n = 58).

Stage 2: Duplicate Removal. Duplicate records were identified using Rayyan systematic review software, supplemented by manual verification. A total of 45 duplicate records were removed, leaving 248 unique records for screening.

Stage 3: Title and Abstract Screening. Two reviewers independently screened all 248 records using a standardized eligibility checklist. Inter-rater agreement was calculated using Cohen’s kappa (κ = 0.89), indicating substantial reliability. Disagreements (n = 26, 10.5% of screened records) were resolved through discussion; where consensus could not be reached (n = 4), the third reviewer made the final determination. A total of 42 records were excluded: non-accounting focus (n = 18), no empirical evidence or methodology (n = 14), and outside date range 2015–2025 (n = 10). This left 206 records for full-text assessment.

Stage 4: Full-Text Assessment. Full texts of 206 records were retrieved and assessed for eligibility by both reviewers. A total of 18 records were excluded with reasons: methodology insufficiently described to assess validity (n = 8), no quantifiable outcomes relevant to review objectives (n = 6), and substantively duplicate content from same authors (n = 4). This resulted in 188 records proceeding to quality appraisal.

Stage 5: Quality Appraisal. Quality appraisal was conducted using the Mixed Methods Appraisal Tool (MMAT) Version 2018 (Hong et al., 2018) for studies employing diverse methodologies, supplemented by the Critical Appraisal Skills Programme (CASP) checklist for qualitative studies (

Table 4. Quality Appraisal Outcomes.

Quality Score	Rating	N Studies (%)	Treatment in Synthesis
9–10	High	42 (24%)	Primary evidence base for propositions; weighted heavily in synthesis
7–8	Moderate	98 (56%)	Supporting evidence; limitations noted when findings differ from high-quality studies
6	Low (included)	36 (20%)	Contextual information only; not used for quantitative claims
<6	Excluded	12	Not included in final corpus

). Each study was rated on a 10-point scale across five dimensions: clarity of research questions (0–2), appropriateness of methodology (0–2), adequacy of sampling strategy (0–2), validity of data collection methods (0–2), and soundness of interpretation (0–2). Studies scoring below 6 were excluded (n = 12), leaving 176 studies in the final corpus. The mean quality score for included studies was 7.8 (SD = 1.2, range: 6–10).

2.4. Data Extraction Procedures

Data extraction was performed using a standardized extraction form developed iteratively during pilot testing on 15 randomly selected studies. Both reviewers independently extracted data from all included studies. Discrepancies were resolved through discussion, with extraction decisions documented. The extraction form captured the following fields (

Table 5. Data Extraction Fields.

Extraction Field	Description and Coding Categories
Study identification	Authors, year, journal/venue, DOI, country of study
Study design	Empirical quantitative/Empirical qualitative/Mixed methods/Conceptual/Case study/Design science
Technology focus	IoT/AI-ML/Blockchain/Edge computing/Digital twins/Cloud/Big Data analytics/Multiple
Accounting domain	Financial reporting/Auditing/Fraud detection/Compliance/Management accounting/Tax/Multiple
Sample/context	Industry sector, organization size (SME/Large/Mixed), geographic region, sample size
Quantitative outcomes	Performance metric, reported value or range, baseline comparison (if reported), statistical significance
Implementation challenges	Technical/Regulatory/Organizational/Cost/Skills/Data quality barriers identified
Quality score	MMAT/CASP score (1–10 scale) with dimension-level ratings

).

2.5. Derivation of Quantitative Outcomes

Performance ranges reported in this review (e.g., 85–92% fraud detection accuracy, 40–60% audit cycle reduction) represent the minimum and maximum values identified across multiple studies meeting inclusion criteria. Where studies reported single-point estimates, these contributed to range boundaries according to their position relative to other estimates. The reported ranges should be interpreted as reflecting cross-study variation attributable to differences in implementation contexts (organization size, industry sector, regulatory environment), sample characteristics, measurement approaches, and technology maturity—rather than as confidence intervals from a single meta-analysis.

Supplementary Table S1 provides a detailed mapping of each quantitative claim in this review to its source studies, specifying the exact values reported, sample characteristics, and any contextual factors affecting comparability. This transparency enables readers to assess the strength of evidence underlying each reported outcome and the degree of heterogeneity across studies.

Acknowledging the limits of comparability across studies, we note that outcome measures vary in their operationalization. For instance, ‘fraud detection accuracy’ may refer to precision, recall, F1-score, or AUC-ROC depending on the study. Where possible, we report comparable metrics; where not, we note measurement differences in the evidence mapping. This heterogeneity precludes formal meta-analysis but supports the narrative synthesis approach adopted in this review.

2.6. Proposition Development

Thematic synthesis enabled the development of four evidence-based propositions. These propositions are explicitly framed as testable hypotheses for future empirical research. Each proposition specifies:

• Independent variable: The technology configuration or intervention
• Dependent variable: The accounting outcome measure
• Hypothesized direction: The expected effect based on synthesized evidence
• Evidence base: The number and quality of supporting studies

These propositions do not represent confirmed causal relationships but rather evidence-based conjectures warranting rigorous empirical testing through controlled studies, longitudinal designs, or quasi-experimental methods. The propositions are presented in Section 8 with their full specification and supporting evidence.

2.7. Methodological Limitations

Limitations include English-language restrictions, rapid technological evolution, the absence of meta-analytic techniques, and outcome measurement heterogeneity. These are addressed through comprehensive database coverage, preference for recent publications (2020–2025), and transparent reporting with acknowledged measurement variation in reported ranges.

3. Industry 4.0 and Smart Accounting Ecosystems

Unlike the previous revolutions, which were focused on “digitizing” (using technology) an existing process, the 4th Industrial Revolution will create new relationships between physical production activities and digital financial reporting systems by creating a system of interconnecting cyber-physical systems that produce data through each operational event, automatically validating and recording financial transaction events (Sun et al., 2024). The next section develops the theoretical foundation for how Industry 4.0-based technologies can influence a company’s accounting functions as part of this larger transformation, grounded in the Big Data concepts developed in Section 1.

3.1. Conceptual Foundations

Accounting in Industry 4.0, by virtue of being based on the integration of four digital technologies—Internet of Things (IoT), Artificial Intelligence (AI), Cloud Computing (CC) and Blockchain Technology (BT)—is enabled to evolve from periodic, sampling-based financial processing into a continuously, comprehensively monitored stream of an organization’s economic activity (Gousteris et al., 2023).

A key distinguishing feature of Industry 4.0 Accounting is the bi-directional interaction between a company’s physical operations and its financial records; operational systems can supply data to accounting systems and receive real-time feedback to enable automated decision-making (Lin, 2021).

The technological stack supporting this two-way communication comprises interconnected layers. The bottom layer consists of IoT sensors and devices that capture data about operational activity (i.e., inventory movement, equipment use, energy consumption, delivery confirmations). This produces an ongoing, continuous stream of financially relevant data (Theodorakopoulos et al., 2025a). These data streams are then processed using the next level up in the technological stack, which consists of processing capabilities for both Edge Computing (for timely operations at the device or sensor level) and Cloud Platforms (for centralized analysis capabilities) (Kitchin & McArdle, 2016). The intelligence layer applies Machine Learning (ML) algorithms to find patterns, identify anomalies, and make predictions, and the top layer uses Blockchain Technology (BT) to record transactions in a permanent ledger (the “blockchain”) and execute Smart Contracts (Gandomi & Haider, 2015).

The new digital architecture for accounting will help it transform from being a record of the past (i.e., recording and reporting financial information), to providing real time operational intelligence (Moll & Yigitbasioglu, 2019). The data related to a company’s finances will move away from being a representation of the past, to being a live representation of a company’s current economic situation (Hussien, 2020).

3.2. Technology-Accounting Integration Framework

The practical integration of Industry 4.0 technologies with accounting processes occurs across multiple domains, each characterized by specific data flows, analytical requirements, and operational outcomes.

3.2.1. Asset and Inventory Management

The tracking of assets utilizing the Internet of Things (IoT)—RFID tags, GPS trackers and environmental sensors—provide continuous insight into where an asset is located, how it is being utilized and its overall state at all times (Faccia & Petratos, 2021). Continuous insight into asset state provides accounting with the ability to automatically update inventory valuations, adjust depreciation based on actual usage rates and immediately identify impairment in the value of an asset. Asset tracking organizations that implement IoT-based tracking solutions report a reduction in inventory costs of 15–25%, and an improvement in the accuracy of depreciation of 20–30% when compared to tracking methods used periodically (Alkan, 2022).

3.2.2. Production Cost Allocation

The smart manufacturing environment is generating extremely detailed data about how resources are being used by each unit of a product as it is being produced. The data is generated from sensors that have been placed within equipment and can provide details such as; the amount of energy consumed, materials consumed and time to process per unit of product or batch. As a result, activity based costing (ABC) is now able to be performed at an unprecedented level of detail, and can address historical issues related to the assignment of overhead costs that previously had to be allocated using an estimate of the absorption rate. Research has shown that the use of Internet of Things (IoT) technology to monitor production activities resulted in improved ABC cost allocations ranging from 15–25% in companies using this type of technology (Stamatiou et al., 2023).

3.2.3. Quality and Compliance Monitoring

Quality control (QC) automation utilizes sensors and computer vision to continuously generate compliance information as a direct input to finance. QC automation instantly recognizes defects for accounting purposes; it documents changes in warranties and provides documentation of supplier performance (Najem et al., 2024). The automated monitoring of regulatory compliance—environmental emissions, safety parameters, and product specifications—produces auditable evidence to support both operational and financial reporting obligations (Basiru et al., 2023).

3.2.4. Predictive Maintenance and Financial Planning

Predictive maintenance uses machine learning models to analyze sensor data from equipment to accurately (85–90%) predict when maintenance will be required, thereby allowing proactive scheduling to minimize disruptions to production (Lin, 2021). Predictive Maintenance is also reported by organizations as reducing costs associated with unplanned equipment maintenance, shifting emergency, unpredictable expenditures to planned, budgetable investments. The cost savings for organizations have included a 10–15% reduction in maintenance costs and a 20–25% decrease in unplanned downtime through predictive maintenance (Yusof, 2025).

3.2.5. Supply Chain Financial Integration

A blockchain-based supply chain system creates a common ledger for all parties to record their transaction in an unchangeable form (Gandomi & Haider, 2015). Smart contracts automatically trigger payment upon delivery confirmation, reducing AP process time by 40–60 percent; eliminating accounting errors due to different documentation; and minimizing or eliminating disputes over documentation (Fuller & Markelevich, 2020). Additionally, integrating logistics tracking via IoT (Internet of Things) devices with blockchain-based payment systems will enable automatic reconciliation between the physical movement of products and their financial settlement (Alao et al., 2024).

3.2.6. Cyber-Physical Systems and Financial Intelligence

The combination of physical activities and digital systems produces cyber-physical systems in which “financial intelligence” is produced by the aggregation of operational data. Digital twins (virtual copies of physical assets or processes that receive continuous feeds of real-time data) enable simulation and scenario analysis that are not possible with traditional financial models (Thanasas et al., 2025). A digital twin of a manufacturing facility includes current production status, equipment condition, supply chain position, and current market demand signals, enabling financial projections to adjust dynamically based on current operational status.

This integration capability enables an evolutionary progression from descriptive accounting (what occurred) to diagnostic accounting (why it occurred) to predictive-prescriptive accounting (what will occur and what should be done) (Panchapakesan et al., 2025). The Neural ACT framework demonstrates how neural network architecture can process integrated operational-financial data streams to provide real-time decision recommendations to accountants, with response times of seconds rather than the days or weeks typical of traditional analytical processes (Theodorakopoulos et al., 2025b).

Table 6. Technology-Accounting Mechanism Table.

Technology Layer	Mechanism	Accounting Outcome	Causal Logic
IoT Data Collection	Continuous automated data capture from physical operations via sensors	15–25% reduction in data entry errors	Eliminates manual transcription errors; enables real-time validation against physical state; removes human delay between event and recording
Edge Computing	Local processing at data source reduces transmission latency and bandwidth	40–75% compliance response time improvement	Anomaly detection occurs at source before cloud upload; alerts trigger within milliseconds vs. batch processing delays; enables action before violation occurs
AI/ML Intelligence	Pattern recognition across high-dimensional transaction data using statistical learning	85–92% fraud detection accuracy	Identifies non-linear relationships and subtle patterns invisible to rule-based systems; continuously learns from new fraud patterns; reduces false positives through probabilistic rather than threshold-based detection
Blockchain/DLT	Immutable distributed ledger with cryptographic verification and smart contract automation	70–80% reconciliation effort reduction	Eliminates interorganizational disputes through a shared single source of truth; smart contracts automate settlement upon condition verification; cryptographic proof replaces manual confirmation (Gilcrest & Carvalho, 2018)
Cloud Analytics	Elastic compute resources for large-scale data processing and storage	100% population audit coverage (vs. 5–10% sampling)	Removes computational constraints that necessitated sampling; enables processing of entire transaction populations; provides storage for complete audit trails

explicates the causal mechanisms linking each technology layer to accounting outcomes, addressing how and why these technologies produce documented improvements rather than merely describing their capabilities.

The framework presented in this section establishes the conceptual and technological foundations for the detailed examination of Big Data applications in digital accounting that follows in Section 4.

4. Big Data Applications in Digital Accounting

Building upon the Industry 4.0 framework established in Section 3, this section examines specific applications of Big Data technologies in accounting practice. The analysis progresses from the historical evolution of accounting systems through current implementation domains, documenting quantitative outcomes and identifying research contributions.

4.1. Evolution from Manual to Intelligent Systems

The development of Accounting Information Systems (AIS) has been shaped by four generations of technology, each based on successive developments in computing capability, data processing, decision-making, and decision support (Shalhoob et al., 2024). The first generation was the Manual Generation (pre-1950’s), which used paper-based records, physical documentation, and manual calculations to produce financial reports. Financial reports required weeks of work; all audits relied on sampling due to time constraints, and the ability to analyze was limited to calculating simple ratios (Theodorakopoulos et al., 2025a).

The second generation, Mechanized Generation (1950’s–1980’s), introduced calculators, punch cards, and early computer mainframes. The use of these technologies reduced calculation time but did not improve over batch processing or retrospective reporting cycles (Song et al., 2014).

The third generation, Enterprise Systems Generation (1980’s–2010’s), provided AIS with integrated Enterprise Resource Planning (ERP) platforms which connect operational and financial systems. The use of an ERP platform allowed for capturing transaction-level data, standardizing reporting formats, and improving internal controls. However, all data processing remained retrospective and still relied on sampling (Dwianika et al., 2023).

The fourth generation is Intelligent Systems Generation (2010’s–present). This is a qualitative shift from previous generations. Machine learning, real-time analytics, and Distributed Ledger Technologies have created the environment for continuous monitoring, predictive capabilities, and automated decision-making (Burneikaite, 2019).

Table 7. Evolution of Accounting Technology Systems.

Era	Period	Data Processing	Reporting	Audit Approach
Manual	Pre–1950s	Paper ledgers	Monthly/Quarterly	100% or judgment
Mechanized	1950s–1980s	Batch processing	Weekly/Monthly	Random sampling
Enterprise	1980s–2010s	Integrated ERP	Daily/Weekly	Risk-based sampling
Intelligent	2010s–present	Real-time streaming	Continuous	Continuous monitoring

outlines the differences between the four generations.

The transition to intelligent systems fundamentally alters the accounting function’s organizational role. Traditional accounting provided historical documentation; intelligent systems deliver forward-looking decision support integrated with operational processes (Barik & Ranawat, 2024).

4.2. Conventional Versus Digital System Characteristics

The extent to which big data has changed traditional accounting is best understood by a systematic examination of how conventional versus digitally enabled accounting systems have transformed the way each of their operational dimensions are organized (Nugent et al., 2016).

The primary difference is in the types of data processed: conventional systems process structured, transaction-based data in the form of journal entries, invoices, and payment records. The digitally enabled system processes both structured and unstructured data streams, including text documents, images, sensor data, and social media feeds, thus increasing the amount of data used for financial analysis (Hosen et al., 2022). Research shows that 60–80% of all relevant, financially related information in an organization is in an unstructured format, which cannot be accessed by conventional systems (Dashkevich et al., 2024).

Additionally, there are significant differences in how data is processed. Conventional systems use batch processing, in which data is periodically extracted, transformed, and loaded (Van Duc et al., 2024). Conversely, digital systems use stream processing, allowing continuous ingestion and processing of data and subsequent, real-time analysis (Revsine et al., 2021). As a result of this paradigmatic difference, reporting latency has been reduced from days and/or weeks to seconds and/or minutes. Many companies report a 25–40% reduction in their financial close cycle after adopting a digitally enabled accounting system (Alexander et al., 2003).

Finally, the analytical capabilities of digitally enabled accounting systems differ significantly from those of conventional systems. Conventional systems primarily provide descriptive analytics, summarizing historical transactions through traditional reports and variance analysis. Digitally enabled accounting systems provide diagnostic analytics (identifying cause), predictive analytics (predicting outcomes), and prescriptive analytics (providing recommendations for action) (McMahon & Davies, 1994). For example, machine learning models can analyze transaction patterns to identify fraudulent activity with 85–92% accuracy, compared to 65–75% for traditional, rules-based approaches (Zeff, 2013).

Consequently, the nature of control mechanisms also changes from periodic to continuous. Traditional internal control relies on segregation of duties, authorization hierarchies, and periodic reconciliation. Digitally enabled accounting systems implement continuous monitoring of controls, automatic exception detection, and real-time compliance verification (Lybaert, 2002). Companies that implement continuous controls report a 60–75% reduction in non-compliance violations compared to periodic review methods (Wahlen et al., 2018).

Table 8. Comparative Analysis: Conventional Versus Digital Accounting Systems.

Dimension	Conventional	Digital	Improvement
Data Types	Structured only	Structured + Unstructured	60–80% more data
Processing	Batch (periodic)	Streaming (continuous)	Real-time capability
Reporting Latency	Days to weeks	Seconds to minutes	25–40% cycle reduction
Fraud Detection	Rule-based (65–75%)	ML-based (85–92%)	15–20% accuracy gain
Audit Coverage	5–10% sampling	100% population	Complete coverage
Controls	Periodic review	Continuous monitoring	60–75% violation reduction
Error Rate	2–5% manual entry	0.1–0.5% automated	90% error reduction

presents the comparisons in this paper across the nine operational dimensions of conventional and digitally enabled accounting systems.

4.3. Financial Reporting and Audit Applications

The use of big data technologies in financial reporting has transformed how organizations prepare financial statements and their interactions with auditors and financial information users (Theodorakopoulos et al., 2024a).

Financial reporting is no longer a task of preparing reports at certain times; it can now be used to provide continuous disclosure (Kokina & Davenport, 2017). One example of this technology is the ability to generate reports automatically using Natural Language Processing (NLP) by taking structured financial data and converting it into written financial disclosures. These systems can also analyze transactional data to identify any material differences from prior periods and write explanatory text that meets all regulatory requirements (Taherdoost, 2021).

Organizations have been able to implement these types of automated reporting systems and have documented that they have reduced the amount of time it takes to produce a report by 60–70%, as well as reduced errors by 40–50% when compared to producing a report manually (Fawcett et al., 1998). Another way that big data technologies impact financial reporting is by providing real-time financial dashboards to users, such as investors, customers and other stakeholders, so that they can continuously monitor an organization’s key performance indicators (KPIs) (Guo et al., 2024). These dashboards replace traditional methods of distributing financial statements and allow stakeholders to view an organization’s KPIs dynamically and visually (Kokogho et al., 2025).

These dashboards are connected to operational data streams and the organization’s financial records, allowing stakeholders to immediately review the organization’s current financial position on revenue recognition, cost accumulation, and cash flow (Regueiro et al., 2021).

Studies show that once an organization implements real-time dashboards, it can quickly identify potential financial problems that may require management’s attention 30–40% sooner (Patel, 2023). Continuous auditing is probably the largest technological change in the history of auditing.

Traditional audit procedures included sampling transactions, testing internal controls, and confirming account balances. However, these procedures were limited by their inability to examine every transaction in the entire population (Ramachandran et al., 2024). Using big data technologies, organizations can now perform 100% of their transaction analysis using algorithms to process large amounts of financial data (Ramachandran et al., 2024). Implementations of continuous audits have resulted in:

• Reductions of 40–60% in the length of time it takes to complete an audit
• The ability to examine 100% of the population of transactions versus 5–10%
• Increases of 25–35% in the accuracy of anomaly detection (Khurana, 2020)

Predictive audit risk assessments are another area where big data technologies are changing the history of auditing. These systems use machine learning to analyze past audit findings to identify which accounts or transactions are more likely to contain misstatements (Cederquist et al., 2007). This allows organizations to allocate resources based on the level of risk, as opposed to performing the same number of audit procedures across the board. Audit firms report a 20–30% improvement in efficiencies when they utilize predictive risk assessment (Kasper & Alm, 2022).

4.4. Fraud Detection and Risk Management

Big Data has been found to provide some of its greatest returns on investment through financial fraud detection. There are also many examples of how using Big Data is helping improve fraud detection speed and accuracy (Dubin & Wilde, 1988). The data used to detect anomalies in financial transaction characteristics is then analyzed using pattern recognition algorithms to identify other similar anomalies that may indicate fraudulent transactions. Fraudulent transactions are identified by machine learning models (including neural networks, random forests, and gradient boosting) that are trained based on historical fraud cases; thus, they have the ability to identify very slight patterns that would otherwise go unnoticed by traditional rule-based systems (Bergman & Nevarez, 2006).

The fraud detection methods mentioned above achieve a detection rate of 85–92% while significantly reducing false positives by 40–60% compared with more traditional alert-based thresholds (Zeff, 2013). Behavioral analytics is another form of fraud detection that extends past transactional characteristics into a user’s behavioral patterns. Monitoring a user’s login times, transaction velocity, approval patterns, and communication characteristics enables a system to detect when the user deviates from established baselines (Kastlunger et al., 2009). Behavioral analysis is especially useful for insider fraud detection because insider fraudsters typically have valid system access and can create individual transactions that appear normal (Ataya, 2010). Fraud detection in real time, through the analysis of transactions as they occur, enables the prevention of losses due to fraud rather than just detection after they occur (Breger et al., 2020).

Financial institutions implementing real-time monitoring report:

• Fraud loss reductions of 60–70%
• False positive decreases of 45–55%
• Investigation timeline acceleration of 80–90% (Amalia et al., 2019)

Predictive risk modeling extends beyond fraud to comprehensive financial risk assessment. Machine learning models incorporate market data, economic indicators, counterparty information, and internal transaction patterns to forecast credit, market, and operational risk exposures (Kleven et al., 2020). These models enable proactive risk management by identifying emerging exposures before they materialize into losses (Gordon, 2010).

4.5. Blockchain and Distributed Ledger Applications

The technology of Blockchain has unique functionalities that differ from those of typical databases; therefore, it will affect how companies keep records for accounting, verify them, and coordinate with each other (Slemrod, 2016). Because Blockchain records are immutable, once financial information is recorded on a Blockchain, it can’t be changed or removed (deleted) without first being agreed upon by the network and having some form of cryptographic proof (Shahzad, 2020); this means that the fear of manipulating records will disappear because of the inherent integrity of the audit trail.

In accounting, once something is recorded on a Blockchain, there is no need to verify that it happened—the technology itself provides assurance (Powell & Chaloupka, 2009). Smart contracts enable automated agreements that create accounting entries when certain predetermined conditions are met (James, 2013). Because payment terms are included in Smart Contracts, when goods or services are delivered, the Smart Contract executes automatically, eliminating the need to manually process invoices, reducing the likelihood of payment disputes, and accelerating cash collection. Organizations using Smart Contract-based payment systems report:

• Reductions in accounts payable processing times of 40–60%
• Decreases in payment disputes of 70–80%
• Improvements in working capital of 10–15% due to faster collections (Giannini & Maggiulli, 2002)

Triple Entry Accounting adds an additional ledger to double entry accounting and records all transactions on three separate ledgers (the original two company ledgers and one shared ledger on a Blockchain) (Nielsen et al., 2015), which completely eliminates the need for intercompany reconciliations, since both parties are referencing identical and cryptographically verified transaction records. Intercompany reconciliations decrease by 70–80% in Triple Entry Accounting implementations (Hulle et al., 2011).

Regulatory Technology also leverages Blockchain capabilities to automate compliance (Pandey et al., 2020). By encoding regulatory requirements into Smart Contracts, regulatory adherence is ensured to occur automatically, and transaction records provide continuous real-time visibility to regulatory bodies (Chong & Eggleton, 2003). Tax calculations are performed for each transaction, regulatory reporting occurs automatically, and compliance verification is continuous rather than periodic (Moorthy et al., 2012).

However, blockchain implementation is subject to many obstacles. The immutability that enables audit assurance conflicts with the GDPR’s requirement to erase data, with 67% of European Financial Institutions researching blockchain citing this issue (Saukkonen et al., 2018). Scalability issues limit the number of transactions that can be processed, and energy-related issues affect sustainability assessments (Laine et al., 2012).

4.6. Research Contribution Synthesis

Table 9. Research Contribution Mapping: Big Data in Digital Accounting.

Domain	Prior Understanding	Current Contribution	References
System Evolution	Four-era framework recognized	Quantified capability differences	(Shalhoob et al., 2024; Prasetianingrum & Sonjaya, 2024; Grabski et al., 2011; Bhimani & Willcocks, 2014; Moll & Yigitbasioglu, 2019; Barik & Ranawat, 2024)
Data Architecture	Structured data focus	Unstructured integration (60–80%)	(Nugent et al., 2016; Gandomi & Haider, 2015; Kitchin & McArdle, 2016; Hussien, 2020; Najem et al., 2024; Revsine et al., 2021)
Financial Reporting	Periodic preparation	Continuous disclosure capability	(Theodorakopoulos et al., 2024a; Alao et al., 2024; Oyewole et al., 2024; Basiru et al., 2023; Zeff, 2013; Patel, 2023)
Audit Transformation	Sample-based procedures	100% population analysis	(Ramachandran et al., 2024; Appelbaum et al., 2017; Cao et al., 2015; Dagiliene & Kloviene, 2019; Kokina & Davenport, 2017; Kasper & Alm, 2022)
Fraud Detection	Rule-based (65–75%)	ML-based (85–92% accuracy)	(Guo et al., 2024; Theodorakopoulos et al., 2025b; Khurana, 2020; Song et al., 2014; Fawcett et al., 1998; Taherdoost, 2021)
Blockchain Accounting	Conceptual frameworks	Implementation outcomes quantified	(Dai & Vasarhelyi, 2017; Fuller & Markelevich, 2020; Dashkevich et al., 2024; Sarwar et al., 2021; Kokogho et al., 2025; Theodorakopoulos et al., 2024b)

maps the research contributions documented in this section, identifying knowledge domains, prior understanding, and advances enabled by Big Data applications.

The framework presented in this section establishes the application foundations for examining current industry implementations in Section 5 and implementation challenges in Section 6.

4.7. Critical Analysis—Contextual Moderators and Boundary Conditions

This section synthesizes evidence on contextual moderators and boundary conditions, acknowledges the limits of comparability across studies, and provides critical analysis beyond the mere enumeration of reported improvements.

4.7.1. Organizational Size Effects

Performance improvements demonstrate significant variation by organization size. Large enterprises (>1000 employees) consistently report fraud detection accuracy at the upper range (90–92%), while SMEs typically achieve 82–87% (Okeke et al., 2024; Shalhoob et al., 2024; Song et al., 2014; Theodorakopoulos et al., 2025b). This divergence is attributable to larger training datasets available to enterprise implementations, greater resources for model tuning and validation, and more extensive historical transaction records for pattern learning. The evidence suggests a data volume threshold—estimated at 500,000–1,000,000 annual transactions (Khurana, 2020; Song et al., 2014)—below which AI-based systems may underperform rule-based alternatives. This boundary condition has significant implications for technology selection in smaller organizations.

4.7.2. Industry Sector Variation

Banking and financial services implementations demonstrate the highest audit cycle reductions (55–60%) and fraud detection accuracy (88–92%), attributable to standardized transaction formats, mature regulatory frameworks requiring detailed record-keeping, and substantial IT investment capacity (Dagiliene & Kloviene, 2019; Halkiopoulos et al., 2024; Shalhoob et al., 2024). Manufacturing contexts show more modest improvements (35–45% reduction in audit cycle, 82–87% fraud detection accuracy) due to greater data heterogeneity across diverse IoT sensor networks and less standardized transaction structures (Modupe et al., 2024; Theodorakopoulos et al., 2024a). Retail and e-commerce implementations fall between these extremes (45–55% reduction in audit cycle), benefiting from high transaction volumes but challenged by seasonal variability and multichannel complexity (Iseal et al., 2025; Khurana, 2020).

4.7.3. Regulatory Environment

Blockchain adoption timelines vary by 12–18 months between jurisdictions with explicit DLT regulatory frameworks (e.g., Switzerland, Singapore, UAE) versus those without clear guidance (Bonsón & Bednárová, 2019; Dai & Vasarhelyi, 2017; Mills et al., 2016). GDPR compliance requirements introduce additional complexity for European implementations, affecting 67% of surveyed financial institutions (Bonsón & Bednárová, 2019; Habibzadeh et al., 2019; Wylde et al., 2022) and requiring architectural modifications that delay deployment by 6–12 months. These regulatory variations represent significant boundary conditions: reported performance outcomes from permissive regulatory environments may not generalize to more constrained contexts without substantial adaptation (Mills et al., 2016).

4.7.4. Technology Generation and Maturity

First-generation AI implementations (2015–2018), predominantly using random forests and early neural networks, showed lower accuracy ranges (78–85%) compared to current deep learning approaches employing transformer architectures and ensemble methods (85–92%) (Fawcett et al., 1998; Hasan, 2021; Song et al., 2014; Theodorakopoulos et al., 2025b). This indicates that reported outcomes are partially time-dependent: studies from earlier periods may understate current achievable performance, while laboratory conditions in recent studies may overstate field implementation results. Additionally, accuracy degradation of 5–10 percentage points is observed between controlled laboratory conditions and field deployment due to concept drift, adversarial adaptation, and data quality variations not present in experimental settings (Deliu & Olariu, 2024; Guo et al., 2024; Taherdoost, 2021).

4.7.5. Cross-Study Synthesis: Convergent and Divergent Findings

While 8 of 12 high-quality studies examining AI fraud detection report accuracy above 85%, notable exceptions include Fawcett et al. (1998), Popoola (2023), and Song et al. (2014), where accuracy ranged from 72–78%, attributed to smaller training datasets (<100,000 transactions) and sector-specific fraud patterns not represented in general training corpora. Similarly, audit cycle reduction claims show convergence around 40–60% across 15 studies, but three studies (Dagiliene & Kloviene, 2019; Donepudi, 2019; Grabski et al., 2011) report reductions of only 20–30%, all involving organizations with legacy system integration challenges. These divergent findings suggest that technology performance is contingent on implementation quality and organizational readiness, rather than on technology capability alone (Kunwar, 2019).

4.8. Theoretical Integration

The documented outcomes can be interpreted through several established theoretical lenses, advancing understanding beyond atheoretical description:

Organizational Information Processing Theory (OIPT). Galbraith’s (1974) framework predicts that organizations facing increased environmental uncertainty will adopt mechanisms that enhance information processing capacity. The documented 40–60% audit cycle reductions and transition from 5–10% sampling to 100% population analysis represent substantial increases in information processing capacity, enabling more frequent environmental scanning and faster response to emerging risks. This theoretical alignment explains why technology adoption accelerates in high-uncertainty environments (such as financial services and highly regulated industries), where information-processing demands are greatest.

Technology Acceptance Model (TAM) and UTAUT. The documented 70–80% workforce skill deficits and associated implementation delays align with TAM predictions that perceived ease of use moderates technology adoption intentions. The 40–60% longer deployment timelines in organizations with insufficient training investment suggest that facilitating conditions (UTAUT) significantly moderate the relationship between technology capability and realized outcomes. This explains the observed variability across organizations: identical technologies yield different results depending on workforce readiness.

Institutional Theory. The 12–18 month adoption timeline variation across regulatory jurisdictions reflects institutional pressures beyond technical capability. Organizations in jurisdictions with explicit DLT frameworks (normative and regulatory pressures favoring adoption) implement faster than those facing regulatory uncertainty. The 67% of European institutions citing GDPR-blockchain conflicts illustrates how institutional constraints create adoption barriers independent of technology functionality. This theoretical perspective suggests that performance outcomes documented in favorable institutional contexts may not generalize to more constrained environments without institutional change.

5. Current Applications of Big Data in Digital Accounting

This section will also explore how Section 3’s technological foundation and Section 4’s application framework have been used by different industries to implement Big Data technology. This section focuses on quantifiable results reported from documented applications of Big Data technology, rather than its theoretical capability. As such, it is evidence-based guidance for organizations that are contemplating adopting Big Data technology.

5.1. Industry Implementation Profiles

The practical deployment of Big Data technologies in accounting varies significantly across industry sectors, reflecting differences in regulatory requirements, transaction volumes, and risk profiles. This subsection documents implementation outcomes across four major sectors.

5.1.1. Banking and Financial Services

The most advanced users of Big Data Accounting Technologies are Financial Institutions; they have been driven to adopt them by regulatory compliance requirements, competition, and the high costs associated with fraudulent activity. A number of major implementations indicate significant operational efficiencies have been achieved. For example, HSBC’s Enterprise Fraud Detection Platform, which uses machine learning to monitor in real time all of the firm’s transactions for fraud, resulted in a 60–70% reduction in fraud losses within 18 months of deployment (Sarwar et al., 2021). This platform currently monitors in excess of 2 billion transactions per month and generates risk assessments for those transactions within 100–200 milliseconds, with accuracy greater than 90% (Faccia et al., 2019). Similarly, JPMorgan Chase’s Contract Intelligence (COiN), uses natural language processing to analyze commercial loan documents and as such has reduced the amount of time required to review documents by 360,000 h each year to seconds and improved the accuracy of extracting relevant information from these documents by 40–50% (Wu & Wang, 2020). Similarly, credit risk assessment has changed significantly. Machine learning models that utilize alternative data sets (e.g., behavioral indicators, market indicators, etc.) in addition to traditional credit scoring data elements result in an improvement of approximately 15–25% in the ability to accurately approve loans and reduce the rate of defaults on those approved loans by 20–30% (Dai & Vasarhelyi, 2017).

5.1.2. Insurance Sector

Big Data is primarily used in the insurance sector to support activities such as claims processing and fraud prevention. The two are considered areas with a very high number of transactions and therefore also a large potential exposure to fraud (Oyewole et al., 2024). Studies show that systems that use pattern recognition and network analysis to detect claims fraud have improved the ability to identify fraudulent claims by 35–45 percent compared with systems that rely on rules (Simatupang, 2024). As a result of this improvement, the rate of false positives decreased by 25–30 percent, thereby reducing the time required to investigate legitimate claimants. Also, using automated document analysis and verification techniques can reduce the time to process claims by 20–25 percent (Samokhvalov, 2024). Big data is used in underwriting applications to enable organizations to develop risk-based premiums using real-time IoT data feeds, such as telematics, smart home devices, and wearable devices, tailored to each individual policyholder (Halkiopoulos et al., 2024). Organizations have reported increases of 15–20 percent in premium accuracy, and loss ratios have decreased by 10–15 percent due to the development of data-driven underwriting models.

5.1.3. E-Commerce and Retail

E-commerce and retail are unique in that digital commerce platforms experience fast-paced transactions involving many different payment methods and complexities across borders (Popoola, 2023). Fraud detection systems using real-time analysis of transaction data (characteristics), device fingerprinting, and behavioral patterns have been able to lower chargebacks by 50–60%, yet maintain approval rates over 95% (Iseal et al., 2025) (verification accuracy improved 40–50% for reduced fraud loss and false decline). Lowered abandonment of 30–35% resulted from streamlined authentication (Rane et al., 2023). Revenue recognition automation uses machine learning to automate complex multi-component arrangements, subscription changes, and variable considerations (Kothandapani, 2025). Organizations have reported faster financial close (25–35%) and improved revenue classification accuracy (20–30%).

5.1.4. Government and Tax Administration

Big Data is being used by an increasing number of government tax administrations to monitor compliance with tax laws and optimize tax enforcement practices. Detection of tax evasion using cross referenced data bases; e.g., financial transactions, property records, business registrations, international exchanges of tax information, have been found to improve detection by 70–80% over traditional methods of selecting audits (Rahman et al., 2024); while also improving the false positive rate for compliant taxpayers selected for audits from 55–65%. In addition, the time required to complete an investigation has decreased by 40–50% due to automated evidence assembly and case prioritization (Van Duc et al., 2024). The use of transaction-level data matching to analyze value-added taxes (VAT) has enabled the identification of discrepancies between reported and actual economic activity, thereby enabling targeted enforcement and the recovery of lost revenues (Ezeife et al., 2021). Jurisdictions that use comprehensive VAT analytics report improvements in compliance rates of 8–12 percent.

5.2. Audit and Compliance Automation

The transformation of audit and compliance functions extends beyond the continuous auditing capabilities documented in Section 4.3, encompassing specialized applications in regulatory compliance, anti-money laundering, and tax administration (Rahaman et al., 2024).

5.2.1. Robotic Process Automation Integration

Automation of routine audit procedures through RPA (robotic process automation) includes data extraction, reconciliation verification, and confirmation processing (Syed, 2021). The time required for audit cycles is reduced by 40–60% when RPA automates the collection and initial evaluation of evidence (Okeke et al., 2024). Due to RPA’s ability to perform a comprehensive review of all information with no sampling (100%) of records, the amount of time that an auditor spends on audit work has decreased. Auditors have documented an improvement in the quality of documentation as well, due to the use of standard RPA-generated workpapers (Zhao et al., 2020).

5.2.2. Know Your Customer and Anti-Money Laundering

Financial institutions are required to comply with KYC (Know Your Customer) and AML (Anti-Money Laundering) regulations at a very high level, which consumes an enormous amount of operational resources (Hamledari & Fischer, 2021). Routine automated KYC (know your customer) verification can complete 90–95% of standard customer onboarding processes, therefore significantly reducing the need for manual process handling to only exceptional circumstances that require human judgment (Theodorakopoulos et al., 2024a). Time spent completing the customer onboarding process is reduced from days to hours, and document completion is improved by 30–40% (Theodorakopoulos et al., 2024b). Automated transaction monitoring systems using behavioral analytics and network analysis have reduced the manual review requirement by 75–85%, as well as increased the accuracy of detecting suspicious activity by 60–70% (Gertz, 1998). The improvement in alert quality has enabled compliance staff to focus their efforts on truly suspicious patterns rather than investigating all false positives.

5.2.3. Predictive Compliance Risk Assessment

Machine learning algorithms, using a company’s history of enforcement actions by regulatory agencies, regulatory changes, and company attributes, estimate the organization’s future compliance risk (exposure). Companies that utilize predictive compliance have reported:

• A 50–65% reduction in regulatory penalties as compared to reactive compliance processes (Charoo et al., 2023)
• The ability to optimize resources to focus compliance efforts in high-risk areas of the organization
• Predictive capabilities to identify potential issues early and proactively address them prior to regulatory examination (Perez, 2017)

5.3. Taxation and Management Accounting Applications

Big Data applications in taxation extend beyond compliance automation to strategic tax planning, policy analysis, and management decision support (Vimalachandran et al., 2016).

5.3.1. Advanced Tax Compliance

Systems that use AI to prepare taxes can review a company’s or an individual’s financial transactions, business structure, and tax laws to maximize their tax filing position and ensure compliance with all applicable tax laws (Monda et al., 2012). The percentage of errors made when preparing and submitting a tax return is reduced by 35–45% using automated validation and cross-checking verifications (Khatoun & Zeadally, 2017). Documentation for audit purposes is assembled more comprehensively, resulting in a 25–35% improvement in audit readiness. As tax returns are prepared and submitted, they conform more closely to the regulations of taxing authorities, thereby increasing the rate at which they are accepted (Van de Poel, 2020).

5.3.2. Tax Audit Analytics

The use of machine learning by tax authorities to select audits and execute them has resulted in a significant improvement in the efficiency of their processes (Habibzadeh et al., 2019). Taxpayers who comply with tax laws experience a 60–70% reduction in the number of audits they are subject to, thereby reducing the enforcement burden on taxpayers who have historically been cooperative (Tazi et al., 2021). The time required to complete an audit is reduced by 50–60% through the automation of evidence analysis and issue identification (Dalal, 2020). Automated applications of natural language processing used to examine contracts, invoices, and correspondence enable tax authorities to identify inconsistencies between what has been stated to exist between a taxpayer’s documentation of an arrangement and how it has been treated by the taxpayer for tax purposes, thus allowing them to focus their examinations on high-risk positions (George, 2022).

5.3.3. Transfer Pricing Analytics

Multinationals are subject to a multitude of complexities in complying with transfer pricing regulations on an international basis (Kommera, 2024). The ability to identify comparable transactions through the use of global databases for intercompany transactions as well as third party benchmark data (R. Yang & Xu, 2016) results in a reduction of 30–40% in transfer pricing disputes due to a more defensible transfer price document. Improvements of 45–55% in the completeness of transfer pricing documents meet increasing regulatory demands (Badoni et al., 2024).

5.3.4. Management Accounting Decision Support

Big Data is used in internal management applications to improve planning, forecasting, and performance evaluation (Potenciano Menci et al., 2021). Budget forecasts are dynamic and include current real-time operational data, market signals, and predictive models. As a result, forecast accuracy has increased by 25–35% compared with static budgets (Stiegler & Tung, 2017). Projections are continuously updated as needed, rather than waiting for a regularly scheduled update cycle. Scenarios are developed using Monte Carlo methods and sensitivity analysis to assess possible strategies under uncertain conditions (Lyon & Segal, 2013), and organizations report that the frequency of financial surprises has decreased by 40–50%. Performance is measured on an ongoing basis, and instead of a periodic review process, the organization continuously monitors key performance indicators (KPIs). As a result, organizations can identify issues and implement corrective actions much faster (30–40%), and their corrective actions will be effective 25–35% more often (Roberts et al., 2005).

5.4. Emerging Blockchain Accounting Applications

While Section 4.5 established the fundamentals of blockchain and smart contract capabilities, this section examines emerging applications that extend beyond initial implementations (Brown & Smiler, 2012).

5.4.1. Triple-Entry Accounting Evolution

Triple entry accounting has now been deployed practically within an inter-organisational context (Dave et al., 2021), with reductions in inter-organisation reconciliation efforts of 80–90%, as both transacting organisations (and a common ledger) will have identical, cryptographically verified records (Haines et al., 2004) and disputes arising from transactions are now nearly eliminated due to the immutability of, and the timelessness of, the documentation. The collection of audit evidence for Triple Entry Accounting also shifts from confirmation procedures to simple verification of the shared ledger (Meyer et al., 2007).

5.4.2. Tokenized Asset Accounting

The digital tokens that represent the physical and/or financial assets of companies bring both unique challenges and new opportunities in the area of accounting, especially with fair-value measurements on continuously traded tokens, which allow for real-time valuations of a company’s portfolios as opposed to traditional periodic assessments (Costan et al., 2021). The custody documentation for distributed assets also underscores the need to find new ways to verify the existence and ownership of these assets. New revenue recognition models will apply to fractional ownership models as well for tokenized forms of real estate, art, and other non-liquid assets (Owolabi et al., 2022).

5.4.3. Cross-Border Transaction Efficiency

International transactions using blockchain technology for cross-border payments will be significantly better than current correspondent banking models. Blockchain is a much faster way to settle accounts; in fact, it is approximately 70–80 percent faster due to the elimination of middlemen and the automation of transaction execution (Sargiotis, 2024). The removal of intermediaries’ fees results in an approximate 50–60% reduction in transaction fees. With this model, foreign exchange costs are reduced by approximately 40–50 percent when using real-time rates rather than making adjustments at the end of each month during settlement (Gassman et al., 1995).

5.4.4. Regulatory Technology Integration

Blockchain platforms increasingly incorporate regulatory compliance capabilities (Wylde et al., 2022). Real-time regulatory reporting provides supervisory authorities with direct access to the ledger rather than requiring periodic submissions (Adenekan, 2024). Continuous compliance verification replaces examination-based oversight. Automated regulatory report generation eliminates manual preparation for routine disclosures (Asimakopoulos et al., 2025).

5.5. Technology-Application Synthesis

Table 10. Big Data Technology Applications in Digital Accounting.

Application Domain	Technologies	Documented Outcomes	Considerations
Banking Fraud	ML, Real-time Analytics	60–70% loss reduction; 90%+ accuracy	Model training; regulatory approval
Insurance Claims	Pattern Recognition, IoT	35–45% detection ↑; 20–25% processing ↓	Data quality; legacy integration
E-Commerce	Real-time Scoring, Device ID	50–60% chargeback ↓; 95%+ approval	Transaction velocity; friction
Tax Admin	Cross-reference, Network	70–80% detection ↑; 40–50% faster	Data sharing; privacy
Audit Automation	RPA, Continuous Monitoring	40–60% cycle ↓; 100% coverage	Infrastructure; skills
KYC/AML	Behavioral Analytics	90–95% automation; 75–85% review ↓	Regulatory acceptance
Tax Compliance	AI Classification, NLP	35–45% error ↓; 40–50% yield ↑	Jurisdiction complexity
Blockchain	DLT, Smart Contracts	70–80% settlement ↓; 80–90% recon ↓	Scalability; regulatory clarity

Note: ↑ = increase/improvement; ↓ = decrease/reduction.

synthesizes the applications documented in this section, mapping implementation domains to enabling technologies, quantified outcomes, and implementation considerations.

The applications documented in this section represent current deployment status. Section 6 examines the challenges and barriers organizations encounter in achieving these outcomes, while Section 7 explores emerging technologies that will extend capabilities beyond current implementations.

6. Challenges of Big Data in Digital Accounting

Section 5 has identified many positive results from the adoption of Big Data by accounting firms; however, to achieve these advantages, organizations will have to address numerous significant challenges, including data quality, cybersecurity, the complexity of big data integration, regulatory compliance, and employee capabilities. The goal of this section is to examine each challenge systematically, document quantifiable assessments of barriers to big data adoption, and provide guidance on mitigating the challenges.

6.1. Data Quality and Integration Challenges

The effectiveness of Big Data analytics depends fundamentally on the quality of the input data. Organizations implementing Big Data accounting systems encounter systematic challenges in data acquisition, standardization, and governance (Rozony et al., 2024).

6.1.1. Unstructured Data Processing

As discussed in Section 4.2, while approximately 60–80% of relevant financial information is in unstructured formats, significant technical barriers exist to deriving value from it (Hu et al., 2014). Format incompatibilities, extraction errors, and process limitations have resulted in 15–25% of otherwise valuable information being left unused as a result of incompatible file formats, or other processing issues. Error rates in manual financial statements increase by 8–12% when unstructured data integration is performed without sufficient validation mechanisms. Manual reconciliation time can increase by 30–40% when automation in document extraction yields inconsistent results, necessitating human verification (Bhattarai et al., 2019). Natural Language Processing (NLP) and Optical Character Recognition (OCR) address these limitations somewhat; NLP/OCR systems have demonstrated an average of 75–85% automated accuracy in the processing of invoices and an average of 60–70% in the reduction of manual data entry for routine documents (Hasan, 2021). However, many complex documents continue to present problems, i.e., contracts with nonstandard terms, documents containing handwritten notes, and documents written in multiple languages, resulting in significantly reduced accuracy levels (approximately 50–65%) (Magli, 2025).

6.1.2. Data Incompleteness and Accuracy

Systemic limitations in quality are evident in all transactional database systems that support accounting systems. Incomplete data is a common problem within all transactional databases (data completeness rates of 2–5%) due to timing discrepancies between applications, failed application interfaces, and errors in application design/operation. Although many organizations employ validation checks to help ensure the accuracy of manually entered financial data, error rates of 3–7% persist. Forecasts created using financial data often need revision to some degree (approximately 12–18%) because quality issues with the original underlying financial data were identified after preparation and therefore could not be incorporated into the forecast (Deliu & Olariu, 2024). Imputation techniques employed by machine learning methods may aid in identifying missing data and achieve an accuracy rate of 80–90% when addressing missing data related to routine transaction types. However, the accuracy rate drops to approximately 50–65% when addressing missing data for novel or unusual transactions, as there are limited historical patterns to guide the imputation process (Assidi et al., 2025).

6.1.3. Legacy System Incompatibility

Legacy, traditional accounting software was created to handle batch-processed, highly structured data; therefore, this type of software will need significant modifications before it can support the real-time, unstructured nature of data streams (Abdo-Salloum & Al-Mousawi, 2025). More than 60% of medium-sized companies (those with five or more separate financial systems) experience a barrier when trying to perform real-time analysis across all their financial systems (Rozony et al., 2024). The time delay between legacy batch processing (i.e., monthly/quarterly cycles) and the modern streaming architecture results in a 2–3-day delay in receiving an organization’s consolidated financial picture. Additionally, organizations may require manual reconciliation for 20–30% of transactions crossing system boundaries, and the total time required for each company to complete the period-end close cycle is 15–25% longer (Donepudi, 2019).

6.1.4. Data Governance Requirements

Effective big data accounting is contingent upon the existence of a complete framework for governing data that includes rules validating data in an automated manner when it is captured, reconciling data on a real-time basis between all relevant systems, conducting regular audits for the identification of systemic problems, and establishing responsibility for managing the data (stewardship) (Kothandapani, 2022; Kunwar, 2019). Organizations that adopt a formal approach to governing their data have reported improving data accuracy by 40–60% and reducing audit findings by 25–35% compared with organizations that do not formally govern their data.

6.2. Cybersecurity and Privacy Concerns

Data accounting systems collect a great deal of personal and/or confidential financial information, making them attractive to cyber attackers and increasing the need for companies to comply with privacy regulations (Mahalakshmi et al., 2022).

6.2.1. The Changing Landscape of Cyber Attacks and Financial Services

There has been a 300% increase in cyberattacks on financial institutions since 2019, and the average cost of a breach in financial services is $5.9 million, which is 40% higher than the industry-wide average (Boute et al., 2022). Some of the areas where there are threats include:

• Data Breach: Unauthorized access to a company’s systems, either due to a vulnerability in the system or because someone has compromised their login credentials, or as a result of a successful social engineering attack, and financial data is exposed; the average penalty from regulatory agencies (GDPR) for data breaches is $1.2–$2.8 million, plus additional costs associated with recovering from the incident and damage to your reputation (IA & Miglionico, 2019).
• Ransomware Attack: Ransomware attacks against financial institutions were up 150% between 2022 and 2024, and the average demand for ransom by these types of attacks was $1.5 million. The total cost of a ransomware attack to a financial institution includes the actual ransom paid, as well as the costs of restoring systems, loss of revenue, and loss of reputation, totaling approximately $4.2 million, minus the ransom paid (Grabski et al., 2011).
• Inside Job: Approximately 25–30% of all data breaches in the financial industry are caused by employees, and it takes approximately 77 days to detect that an employee has breached data compared to 29 days for outside hackers to be detected; the larger the number of users who have access to the financial institution’s data, the larger the potential scope of damage to the company (Modupe et al., 2024).
• Social Engineering: A phishing attack against an organization that does not have employee training to prevent such attacks will succeed 15–25% of the time, resulting in unauthorized access to the company’s financial data, unauthorized transactions, and/or the unauthorized transfer of financial data out of the company (X. Yang, 2024).

6.2.2. Privacy Regulatory Obligations

Regulations regarding privacy, such as GDPR and CCPA, require explicit consent, a limited purpose for the collection of personal data, the least amount of data necessary (data minimization), and respect for individual rights, which place restrictions on how organizations can use big data (Premsankar et al., 2018). The GDPR includes non-compliance penalties of up to 20 million euros or 4% of an organization’s global gross income; since 2020, enforcement actions have grown by 40% each year (Herman et al., 2023). Compliance constraints affect approximately 40–50% of all possible Big Data use cases; therefore, an organization may need to either modify its architecture or abandon proposed use cases (Mohd et al., 2025).

6.2.3. Cloud-Based Infrastructure Risks

The use of cloud-based Big Data architectures raises additional concerns about Big Data security. Uncertainty about where data is stored across multiple jurisdictions in cloud-based Big Data architectures makes it difficult to comply with data residency requirements in different locations (Khang et al., 2025). In addition, there are risks associated with third parties accessing your data through personnel of the cloud provider when you have no control over these individuals. As a result, the organization should implement contractual protections and technical controls to mitigate this risk. Shared security responsibilities create ambiguities regarding the security obligations of both the cloud service provider and the customer (Alcazar et al., 2020). Organizations that have developed a comprehensive cloud security framework report experiencing a 60–70 percent decrease in reported security incidents compared to those using basic cloud security configurations (Hernadi, 2012).

6.2.4. Security Architecture for Big Data Systems

Big Data has a unique set of vulnerabilities that can be addressed by defense in depth models which have multiple layers of protection (Endiana et al., 2020):

• Hardware Security: Hardware Security Modules (HSMs) providing tamper-resistant cryptographic key storage, secure boot mechanisms ensuring firmware integrity, and device authentication using PKI certificates (Cairns, 2000).
• Network Security: Software-Defined Perimeter (SDP) with micro-segmentation, dedicated VLANs for accounting-critical systems, TLS 1.3 encryption for data in transit, and network-based intrusion detection systems (Figueroa et al., 2010).
• Data Security: Zero-trust architecture with Attribute-Based Encryption (ABE), homomorphic encryption enabling computation on encrypted data, and blockchain-based audit trails providing integrity verification (Dura & Suharsono, 2022).
• Application Security: Zero Trust Architecture (ZTA) principles requiring continuous authentication, behavioral analytics detecting anomalous user interactions, and API security gateways protecting service communications (Lauslahti et al., 2018).

Studies have shown that these architectures reduce an attacker’s attack surface by approximately 78%, increase data confidentiality by approximately 85%, and decrease Mean Time To Detect (MTTD) threats by approximately 65% over traditional architectures (Hamilton, 2020).

6.3. Ethical and Legal Issues Surrounding Accounting Systems Based on Big Data

The application of Big Data in accounting raises several ethical issues regarding the accountability of algorithms and their bias, as well as legal and regulatory concerns that are also evolving (Hewa et al., 2021).

6.3.1. Increased Complexity of Compliance with Regulatory Standards

Companies must comply with a number of different and overlapping regulatory frameworks:

• Accounting and Financial Reporting: Companies operating in multiple countries have to adhere to IFRS (International Financial Reporting Standards) and GAAP (Generally Accepted Accounting Principles). These companies may encounter difficulties demonstrating that their accounting has been prepared using machine-learning-based models, which can generate classifications that an auditor cannot fully follow. These difficulties arise from the ‘black box’ characteristics of many complex models, including neural networks and ensemble methods, that conflict with the need for companies to demonstrate transparent and accountable decision-making processes.
• Data Protection Regulations: The principle of minimizing the amount of data collected and used and limiting its purposes under the EU’s General Data Protection Regulation (GDPR) makes it difficult to collect and analyze large volumes of data to support business decisions (Qureshi, 2024). Companies indicate that up to 40–50% of all possible Big Data use cases are subject to such restrictions and therefore require changes to their architecture.
• Internal Control Requirements: SOX mandates documented controls over financial reporting processes (Zhen & Zhen, 2024). AI-based systems require new control frameworks that address algorithm governance and change management, model validation and testing procedures, audit-trail requirements for automated decisions, and segregation of duties in automated processes.

6.3.2. Algorithmic Accountability

There are also a number of accountability-related challenges associated with machine learning models:

• Transparency and Explainability: In recent years, there has been increasing regulatory pressure for companies to provide explanations of the automated decisions they make about individuals or other organizations (Zhao et al., 2020). However, for these explanations to be useful to customers and regulators, the decision logic underlying automated decisions must be understandable to humans. Unfortunately, it is common for complex models that achieve high accuracy to operate as “black boxes,” making it impossible to articulate their decision logic in human-understandable terms.
• Bias Detection and Mitigation: The data sets used to train machine learning models reflect historical realities and may contain discriminatory biases against particular demographics or sub-populations (Wilkens & Moorhouse, 2023). For example, a company reported that the machine learning model it was developing to identify applicants who were likely to become successful employees included a factor that reflected the applicant’s zip code and thus was biased towards applicants from affluent neighborhoods. Similarly, another company found that its automated fraud detection system generated disproportionately more false-positive identifications of applicants from minority backgrounds. Finally, one bank found that the segmentation model it developed to identify customer segments for marketing purposes produced segments that contained many more members of traditionally disadvantaged demographic groups than would be expected by chance alone.
• Human Oversight Requirements: There is now increasing regulatory pressure on companies to require a human to review and approve the results of significant automated decisions (Zhou, 2025). Although such a review will add 15–25% to processing costs, it raises important questions about whether reviewers can evaluate the complex outputs of the models that generated the original decision.

6.3.3. Ethical Guidelines for Implementing Big Data Accounting Systems

Any company that implements Big Data accounting systems needs to develop guidelines and policies for collecting and using data, for making algorithms transparent enough for audit and review, for conducting regular bias audits of automated decision systems, for assigning accountability for the outcomes of automated decisions, and for communicating with stakeholders about how their data is being processed automatically.

6.4. Skill Gaps and Implementation Barriers

The skills necessary to implement a successful Big Data accounting system are currently being inadequately developed by both educational institutions and professional development systems (Kostopoulos et al., 2025).

6.4.1. Data Literacy Deficits

The traditional accounting curriculum emphasizes financial principles over data analytic skills, which creates a systemic gap in skill sets (Bourveau et al., 2024). Only about 20–30% of practicing accountants have received sufficient education or training in data analytics software; 70–80% of practicing accountants lack sufficient knowledge of data analytics software (Murphy et al., 2020). About 55–65% of accounting professionals are uncomfortable using programming languages such as Python, R, and SQL. Only about 45–55% of accounting professionals are familiar with data visualization platforms like Power BI and Tableau (Murphy et al., 2020). About 60–70% of practicing accountants have received no formal education in machine learning concepts relevant to accounting (Makarov et al., 2019).

6.4.2. The Need to Expand Professional Competence

In order to successfully apply Big Data accounting techniques, an accounting professional must develop new competence in several different areas:

• Technical Skills: Data manipulation and analysis of large datasets, use of Business Intelligence (BI) tools, programming skills for extracting and transforming data from various data sources, statistical literacy to interpret results from analytical processes, and an understanding of what can and cannot be done with machine learning models (Badakhshan & Ball, 2023).
• Analytical Skills: Identifying patterns in financial data, identifying anomalies in financial data and investigating those anomalies, understanding how to interpret results from predictive modeling, and assessing risks based upon data-driven insights (Olanrewaju et al., 2024).
• Integration Skills: Working collaboratively with IT and data science teams, translating technical language to financial language, and re-designing processes to include automation where possible (Saxena et al., 2022).

6.4.3. Strategies for Training and Development

There are three strategies that have been implemented to help improve the skills of accounting professionals with regards to Big Data accounting techniques:

• Academic Curriculum Reform: Colleges and universities that have integrated Big Data, Artificial Intelligence (AI), and Financial Technology (FinTech) into their accounting curricula report 40–50% improvements in their graduates’ job placement rates for data enabled positions (Redlein et al., 2023).
• Professional Certification: Accounting professional certification bodies (AICPA, ACCA, etc.) offer certifications in data analytics; only 15–20% of practicing CPAs have achieved data analytics certification (Odonkor et al., 2024).
• Organizational Training: Companies typically spend 2–5% of their Big Data project budget on workforce training. If insufficient resources are devoted to training, companies may experience 40–60% longer deployment timeframes, 30–50% higher error rates during the initial deployment period, and 25–35% higher levels of employee resistance to adopting the new system (Necula & Roebling, 2024).

6.4.4. Implementation Costs as an Obstacle

Implementing a Big Data accounting system will require significant investments in terms of money:

• Data Analytics Platforms: $100,000–$500,000 is the estimated cost for mid-sized organizations for advanced analytics platforms.
• Cloud Infrastructure: Mid-sized organizations pay between $50,000–$200,000 per year for cloud infrastructure services.
• System Integration: The cost for system integration can range from $200,000–$1,000,000 depending upon how complex a company’s legacy systems are.
• Training and Change Management: The estimated cost for training employees and managing change is $50,000–$150,000.
• Ongoing Maintenance: Annual ongoing maintenance costs for the new system should be expected to range from 15–25% of the original cost (Mookerjee & Rao, 2021).

Small and Medium-Sized Enterprises (SMEs) are at a disproportionately high barrier to implementing a Big Data accounting system because 60–70% of SMEs reported that they exceeded their IT budgets when implementing a Big Data accounting system, resulting in a competitive disadvantage compared to larger organizations that have greater resources (Appelbaum et al., 2017).

6.5. Challenge-Mitigation Framework

Table 11. Big Data Challenges and Mitigation Strategies in Digital Accounting.

Challenge Domain	Key Issues	Mitigation Strategies	Success Metrics
Data Quality	60–80% unstructured; 2–5% incompleteness; 3–7% errors	NLP/OCR; ML imputation; governance frameworks	40–60% accuracy ↑; 25–35% audit findings ↓
Cybersecurity	300% incident ↑; $5.9M breach cost; 25–30% insider	Defense-in-depth; zero-trust; encryption; training	78% attack surface ↓; 65% faster detection
Regulatory/Ethical	40–50% GDPR constraints; explainability; bias	Ethics guidelines; bias audits; human oversight	Compliance maintenance; accountability
Skills/Implementation	70–80% training gaps; $100K–$1M costs; SME barriers	Curriculum integration; certifications; phased rollout	40–50% placement ↑; reduced timelines

Note: ↑ = increase/improvement; ↓ = decrease/reduction.

synthesizes the challenges documented in this section, mapping them to specific issue domains, mitigation strategies, and success metrics.

The challenges documented in this section represent current implementation barriers. Section 7 examines emerging technologies—quantum computing, advanced DeFi applications, digital twins, and ESG analytics—that may address some challenges while introducing new considerations.

7. Future Research Directions and Emerging Technologies

The previous sections (Section 5 and Section 6) discussed present-day uses of big data in digital accounting, as well as the obstacles to its application. The purpose of this section is to identify emerging technologies and future research areas that will influence the development of the discipline over the next decade, and to examine potential research gaps and evaluate new technologies that can assist in developing frameworks that contribute to both greater academic understanding and better implementation.

7.1. Quantum Computing Applications in Financial Analytics

The application of quantum computing to financial analytics is considered an opportunity to achieve significant technological improvements in the field (Orús et al., 2019). The primary difference between quantum computers and classical systems is their capability to perform some computations at an entirely different scale (than classical systems) (Herman et al., 2023; Mohd et al., 2025).

7.1.1. Quantum Advantage in Financial Calculations

Quantum algorithms are capable of demonstrating the potential for significant computational advantages for several types of computations that are fundamental to both accounting and finance:

• Portfolio Optimization: Large-scale portfolio optimization presents a problem for many classical optimization algorithms because they grow exponentially in complexity as the number of assets increases. Quantum methods such as variational quantum eigensolvers have demonstrated the potential for a 100–1000-fold reduction in computation time relative to classical methods for portfolio optimization involving portfolios with 500+ assets (Mohd et al., 2025). While current quantum hardware’s limited capacity has restricted its practical application, we expect commercial viability within 5–10 years.
• Risk Modeling: Monte Carlo simulations are commonly used in the process of assessing risk; however, quantum computing can provide a speedup for these simulations by utilizing amplitude estimation techniques (Khang et al., 2025). Theory suggests a quadratic speedup could reduce the time required to complete these simulations from hours to minutes for complex derivative pricing and Value-at-Risk calculations.
• Cryptographic Implications: Quantum computing will break current encryption standards for protecting financial information (Alcazar et al., 2020). Research into post-quantum cryptography will address this issue. We anticipate NIST releasing new standards for quantum-resistant algorithms during 2025–2030, which will require upgrades to accounting system security.

7.1.2. Research Gaps and Opportunities

There are currently insufficient studies focused on adapting quantum algorithms to accounting-specific problems, including multi-entity consolidation optimization, real-time audit sampling strategies, and complex revenue recognition calculations. There are opportunities for academic-industry collaborations to develop quantum-ready accounting frameworks that anticipate when the necessary hardware will mature.

7.2. Decentralized Finance and Advanced Blockchain Applications

Decentralized finance (DeFi), as described in Section 4.5 and Section 5.4, extends blockchain technology to new, innovative financial models with a wide range of accounting implications (Bourveau et al., 2024).

7.2.1. Emerging DeFi Accounting Challenges

The emergence of DeFi protocols has created accounting issues that are far beyond what current accounting standards can address:

• Accounting for Liquidity Pools: In automated market makers, there is a need for the ongoing fair value assessment of the assets held within pools, recognition of yields for providing liquidity, and measuring impermanent losses—all concepts currently without accounting guidance (Bourveau et al., 2024)—therefore, research opportunities do exist for developing DeFi-specific measurement frameworks.
• Valuing Governance Tokens: Governance tokens that grant the right to participate in the decision-making process of a DeFi protocol have characteristics of equity instruments, voting rights, and speculative assets; classification and measurement guidance does not exist today, and therefore will result in inconsistent treatment by different reporting entities.
• Cross-Protocol Composability: The “money lego” architecture of DeFi allows users to build complex multi-protocol strategies where individual transactions may be made up of transactions on multiple platforms. This creates a need for research and development of accounting decompositions and attributions for these transactions.

7.2.2. Evolving Smart Contracts

Smart contract advancements provide additional applications beyond those provided in Section 4.5:

• Self-Executing Financial Statements: There is research being conducted into the use of smart contracts to generate audited financial statements based upon historical on-chain transactional data and eliminate traditional audits for blockchain-native organizations (James, 2013; Giannini & Maggiulli, 2002).
• Programmable Compliance: Regulatory compliance that is programmatically enforced through smart contracts ensures that regulatory violations become technologically impossible instead of just detectable (Wylde et al., 2022; Adenekan, 2024); future areas of research would include the development of compliance-by-design frameworks for tax, securities, and anti-money laundering regulations.
• Decentralized Autonomous Organizations: As discussed earlier, DAOs represent a decentralized form of organization that is without a traditional corporate structure. They challenge many of the fundamental accounting principles that govern how we account for an entity’s boundaries, recognize ownership, and determine who is responsible for preparing the entity’s financial statements.

7.3. Digital Twins, AI Advancement, and Real-Time Financial Intelligence

The digital twin technology developed conceptually in Section 3.2 is now evolving toward comprehensive financial modeling (Murphy et al., 2020; Makarov et al., 2019).

7.3.1. Financial Digital Twins

A virtual replica of an organization’s financial system enables:

• Scenario Simulation: The ability to model real-time strategic alternative options (acquisition impacts, market entry scenarios, etc.) that are based on operational and financial data, providing integrated analytical capabilities greater than those available through traditional financial modeling (Badakhshan & Ball, 2023).
• Predictive Financial Statements: Continuous projections of future financial statements based on the current operational trajectory, market conditions, and the strategic assumptions made at the time the predictive financials were created. Research indicates a potential 30–40% increase in forecast accuracy relative to periodic budgeting methods (Stiegler & Tung, 2017).
• Audit Simulation: Testing the effectiveness of internal controls and identifying areas of high risk before conducting a physical audit. It has been proposed that pre-audit assessment using digital twin-enabled audit simulations can reduce the audit cycle time by 25–35 percent while enhancing risk identification.

7.3.2. Advanced AI Applications

AI development capabilities have advanced significantly from the examples provided in Section 4 and Section 5:

• Large Language Models in Accounting: GPT-class language models have demonstrated their capability to analyze financial statements, draft disclosures, and interpret regulations (Husch et al., 2024; Deliu & Olariu, 2024). The areas for additional research include developing large language models specifically designed for accounting, investigating hallucinations in the context of financial reporting, and creating frameworks to validate disclosures generated by AI.
• Autonomous Accounting Agents: AI agents with the ability to make autonomous judgments regarding the classification of transactions, materiality assessments, and disclosures will be the focus of this area of research. The area of accountability needs to be addressed by specifying how to ensure accountability when AI systems make decisions that would otherwise require professional judgment.
• Explainable AI for Audit: The research efforts in this area will be focused on addressing the accountability concerns associated with algorithmic decision-making discussed in Section 6.3. The emphasis will be on developing transparent and explainable models for audit applications, where the rationale for a decision must withstand regulatory scrutiny (Zhao et al., 2020; Wilkens & Moorhouse, 2023).

7.4. ESG Integration and Sustainability Accounting Analytics

The field of Environmental, Social, and Governance (ESG) reporting is an increasingly dynamic area, where Big Data solutions are emerging to meet growing needs for this type of reporting (Olanrewaju et al., 2024; Saxena et al., 2022).

7.4.1. ESG Data Analytics

Reporting on sustainability demands the ability to collect and analyze large amounts of data outside of normal financial systems:

• Quantifying Environmental Impacts: The use of Internet-of-Things (IoT) sensors, satellite imaging, and supply chain monitoring allows for detailed measurements of carbon emissions, consumption of resources, and environmental impacts that were either measured indirectly or had no measurable value in the past (Olanrewaju et al., 2024). Big Data analytics has transformed ESG reporting into quantified, verifiable reporting.
• Social Impact Analytics: Use of natural language processing on communications of employees, customer reviews/feedback, and community interactions allows for systematic evaluation of social performance indicators (Saxena et al., 2022). Sentiment analysis and topic modeling can identify potential social risk factors before they evolve into financial risk factors.
• Governance Analytics: Machine learning models of board compositions, executive compensation practices, and decision-making processes can identify governance strengths and weaknesses (Redlein et al., 2023). Predictive governance models can predict future regulatory and reputational risks based on governance practices.

7.4.2. Frameworks for Reporting Financial Information and ESG Data

Research will focus on integrating ESG information with financial information:

• Double Materiality Assessment: Big Data will provide a mechanism to systematically assess issues that are important both financially and from the perspective of stakeholders to support the growing number of jurisdictions with double materiality regulations, such as those in the European Union.
• Real-Time ESG Dashboards: Real-time continuous monitoring of ESG performance, similar to real-time dashboards used by companies to monitor their financial performance as described in Section 4.3, will enable companies to manage their sustainability proactively rather than retroactively.
• Development of Assurance Methods: Assurance methods for ESG data must be developed to account for the non-financial nature of ESG data; the uncertainty associated with estimates; and the need to validate forward-looking statements.

7.5. Research Agenda and Implementation Roadmap

Table 12. Future Research Directions in Big Data and Digital Accounting.

Technology Domain	Research Priorities	Timeline	Expected Impact
Quantum Computing	Algorithm adaptation; Post-quantum security	5–10 years	100–1000× speedup; Security paradigm shift
DeFi/Blockchain	Measurement frameworks; Smart contract audit	2–5 years	New asset classes; Automated compliance
Digital Twins	Financial modeling; Audit simulation	3–7 years	30–40% forecast ↑; 25–35% audit accel.
Advanced AI	Accounting LLMs; Explainable models	2–5 years	Automated judgment; Enhanced transparency
ESG Analytics	Impact measurement; Integrated reporting	1–3 years	Quantified sustainability; Double materiality

Note: ↑ = increase/improvement.

synthesizes the future research directions documented in this section, mapping emerging technologies to research priorities, implementation timelines, and expected impacts.

Implementation Considerations

Implementation considerations for organizations developing their infrastructures to address the advancements described above include:

Near-Term (1–3 years): Establish data infrastructure supporting ESG analytics and advanced AI applications. Develop workforce capabilities in emerging technologies. Monitor developments in DeFi accounting standards and opportunities for blockchain integration.

Medium-Term (3–7 years): Pilot digital twin implementations for financial planning and audit applications. Evaluate advanced AI deployment for judgment-intensive processes. Implement post-quantum cryptography transitions as standards mature.

Long-Term (7–10+ years): Assess quantum computing applications as commercial hardware becomes available. Develop accounting frameworks for autonomous organizational structures. Integrate fully automated financial reporting and assurance systems.

8. Conclusions

This systematic review examined the convergence of Big Data and IoT in digital accounting under Industry 4.0 conditions, synthesizing 176 academic sources to analyze how artificial intelligence, blockchain, edge computing, and digital twins transform financial operations through real-time analytics, fraud detection, compliance automation, and ESG integration.

The research established three principal contributions. First, a four-layer conceptual framework—IoT data collection, edge-cloud processing, AI-blockchain intelligence, and accounting application—demonstrates how Industry 4.0 technologies transform accounting from retrospective documentation to real-time operational intelligence, with documented outcomes including 15–25% inventory cost reductions, 85–90% predictive maintenance accuracy, and 40–60% accounts payable processing reductions. Second, an application taxonomy mapping Big Data implementations across financial reporting, audit, fraud detection, and blockchain accounting documents fraud detection accuracy improvements from 65–75% (rule-based) to 85–92% (machine learning), audit cycle reductions of 40–60% with coverage expansion from 5–10% sampling to 100% population analysis, and reconciliation effort decreases of 70–80% through triple-entry systems. Third, a challenge-mitigation framework addresses data quality barriers (60–80% unstructured data, 2–5% incompleteness), cybersecurity threats (300% incident increase, $5.9 million average breach cost), regulatory complexity (40–50% GDPR-constrained use cases), and workforce deficits (70–80% insufficient training), with mitigation strategies achieving 78% attack surface reduction and 40–60% accuracy improvements.

The findings generate actionable implications across stakeholder groups. Organizations face compelling business cases—60–70% fraud loss reductions, 90–95% KYC automation, 50–60% chargeback decreases—yet must navigate implementation costs of $100,000–$1,000,000 and training requirements affecting 70–80% of practitioners. Regulators must resolve tensions between blockchain immutability and GDPR erasure requirements affecting 67% of European financial institutions while enabling programmable compliance opportunities. Educational institutions must address documented skill gaps—55–65% programming discomfort, 60–70% lacking machine learning knowledge—through curriculum transformation that demonstrates 40–50% improvements in graduate placement.

The thematic synthesis identified four testable propositions that unite theoretical frameworks with empirical patterns. Proposition 1 posits that IoT-enabled real-time analytics achieve superior reporting accuracy, supported by reductions of 15–25% in cost allocation error and 25–40% in financial close acceleration. Proposition 2 asserts that hybrid AI-blockchain fraud detection is superior, evidenced by 85–92% versus 65–75% accuracy and 60–70% loss reductions. Proposition 3 proposes that edge computing enables superior compliance responsiveness, as demonstrated by 40–75% reductions in latency and 55–75% improvements in response time. Proposition 4 identifies GDPR-blockchain incompatibility as measurable adoption barriers, as evidenced by 67% institutional citation rates and 12–18-month implementation delays.

Future research must address emerging technologies such as quantum computing applications (5–10 year horizon) offering 100–1000× optimization speedup; DeFi measurement frameworks and DAO accounting standards (2–5 years); digital twin implementations achieving 30–40% forecast improvements and accounting-specific large language models (3–7 years); and ESG analytics enabling double materiality assessment (1–3 years). The systematic review methodology, while providing a comprehensive synthesis, cannot establish causal relationships, and publication bias may overstate technology benefits. The transformation documented in this review represents a fundamental reconceptualization of accounting’s organizational role—from historical record-keeping to predictive decision support, from periodic reporting to continuous intelligence, from sample-based audit to comprehensive monitoring. Successfully navigating this transformation requires sustained stakeholder commitment to workforce development, infrastructure investment, and governance evolution, as quantum computing, advanced AI, decentralized finance, and sustainability reporting create accounting environments that are substantially different from current practice.