Previous studies have investigated various predictors to fraud detection in accounting firms and organisational contexts, for example, Xu, Wang, & Ye (2024) examined the role of digitalization on fraud detection if fraud occurs in Chinese publicly listed companies. Moreover, Soehaditama (2024) examined the role of organisational culture and internal audit on the fraud detection in the Indonesian organisations. Furthermore, detection procedures and technology anti-fraud awareness and fraud prevention significantly and positively influence the fraud detection skills (Sipayung, Yanti & Setya, 2022). Furthermore, Awosika, Shukla, & Pranggono (2024) examined the effect or influence of explainable artificial intelligence and learning in fraud detection in the financial aspects. Furthermore, research indicates that strong internal control systems and favourable ethical environments enhance the probability of fraud detection (Bonrath & Eulerich, 2024). Research in the Middle East underscores the impact of governance reforms, Shariah-compliant oversight in Islamic finance, and the quality of disclosure on fraud risk mitigation. However, it also reveals deficiencies in investigative techniques and analytics among practitioners (Amin & Motta, 2023; PwC Middle East, 2022).

In Saudi Arabia, growing data indicates advancements in regulatory enforcement and the professionalisation of internal auditing; nonetheless, there is a paucity of empirical research connecting practitioner competencies to detection outcomes within enterprises (Al-Shammari et al., 2018; CMA, 2023). This literature indicates that although organisational and system-level elements are extensively examined, the micro-foundations of fraud detection—specifically the forensic skills and educational preparation of practitioners—necessitate more empirical scrutiny, especially within the Saudi Arabian context. It is because there is a lack of studies conducted in the Middle Eastern countries’ context specifically, the current research on fraud detection among the accountants in Saudi Arabia is limited. There is dearth particularly with empirical studies linking forensic skills and education of practitioners to quantified detection outcomes in accounting firms (Al-Shammari et al., 2018). This gap makes it harder to make policies and develop training based on evidence, especially when rules are changing quickly. Hence, filling this gap in the current literature could enrich current understanding and help practitioners figure out if investing in skills and education leads to real improvements in detection in the Saudi professional world.

With the above gaps, this study proposes job satisfaction as a mediator in connection between forensic skills as well as detection findings for practitioners. Studies (Riyanto, Endri & Herlisha, 2021) indicate that job satisfaction promotes motivation, persistence, and discretionary effort—elements that might improve investigative thoroughness and the successful utilisation of forensic technologies (Vo, Tuliao, & Chen, 2022; Bakker & Demerouti, 2007; Fan, 2025). Research has employed job satisfaction as a mediator in various audit and accounting scenarios, including the correlation between resources and performance or ethics and compliance behaviours (Berhanu, 2023; Promchart & Potipiroon, 2020). However, to our knowledge, it has yet to be examined as a mediator in the direct relationships between forensic skills/education and fraud detection outcomes proposed herein. This signifies a significant gap: if skills and education augment perceived competence and job clarity, subsequently increasing satisfaction and investigative diligence, then satisfaction emerges as a viable conduit through which competences boost detection particularly in high-pressure contexts.

The questions of the study are outlined below:

Empirical literature as well as practitioner reports indicate that containing forensic accounting education and curriculum and professional development enhances fraud detection skills (Kinanti et al., 2023; Michelet et al., 2023; Awwad & Abdelsattar, 2025). Studies (i.e., Kinanti et al., 2023; Awwad & Abdelsattar, 2025) report a market demand for forensic education and demonstrated that familiarity with forensic subjects (fraud schemes, investigative procedures, litigation assistance) prepares graduates for detection responsibilities. Apostolou et al. (2010) recognised forensic and fraud-related subjects (analytical methods, interviews, data mining) as essential competencies for assurance professionals (Labuschagne & Fourie, 2025). Research indicates that auditors with forensic training demonstrate increased professional scepticism and enhanced red flag detection, resulting in superior identification of misstatements and fraud risks (Atika & Rochayatun, 2023). Khaksar, Salehi, & Lari DashtBayaz (2022) contended that specialised fraud training improves audit quality by synchronising procedures with fraud risk indicators. Universities that incorporate forensic courses observe improvements in students’ critical thinking and case-based reasoning pertinent to fraud detection (Nilendu, 2024).

From a Human Capital Theory standpoint, this schooling enhances domain-specific knowledge beyond conventional auditing, whilst the JD–R model conceptualises it as a job resource that elevates efficacy and motivation in high-demand investigative settings (Khaksar et al., 2023; Kinanti et al., 2023; Awwad & Abdelsattar, 2025; Nilendu, 2024). Thus, we propose the following hypothesis.

Hypothesis H1: Forensic accounting education is positively related with fraud detection.

Forensic accounting skills, including digital forensics, data analytics, investigative questioning, link analysis, and litigation support, are closely linked to how well someone can find fraud. Empirical evidence demonstrates that data analytic expertise facilitates the early detection of anomalies within extensive transaction datasets, hence enhancing the sensitivity to hidden schemes (Kinanti et al., 2023). Interviewing and deception detection skills improve the gathering of diagnostic clues and confessions during investigations (Nilendu, 2024). Understanding internal controls and fraud schemes enhances the identification of high-risk procedures and the creation of customised testing (Mohd-Sanusi, Mohamed, Omar, & Mohd-Nassir, 2013; Ziorklui, 2024). Additionally, proficiency in electronic evidence collecting maintains the chain of people who are associated with frauds and is essential for effective case settlement (Dkhar, Lyngdoh, & Kumar, 2025).

Experimental study indicates that forensic professionals surpass common auditors in fraud risk assessments and red flag identification, particularly in high-pressure or intricate situations (Campa, Quagli & Ramassa, 2023). Companies that use forensic analytics say they find more problems and fix them faster (Holmes & Douglass, 2021; Veledar, Bašić, Demirović, & Beširević, 2024). According to HCT, these skills are a type of human capital that makes it easier to find things. In the JD–R model, skill depth serves as a job resource that mitigates cognitive demands and enhances outcomes, in part by boosting confidence and perseverance. These arguments affirm a relationship of forensic accounting skills with fraud detection. Hence, we hypothesise that:

Hypothesis H2: Forensic accounting skills is positively related with fraud detection.

Job satisfaction of employees is called primary attitudinal outcome in the JD–R paradigm, occurring when job resources meet or surpass job expectations, and when employees perceive competence, autonomy, and effect (Bakker & Demerouti, 2007; Schaufeli & De Witte, 2017; Veledar et al., 2024). Greater job satisfaction is associated with enhanced task performance and discretionary effort, especially perseverance in fraud detection tasks (Narsa, Afifa & Wardhaningrum, 2023). Forensic accounting knowledge and abilities can enhance job satisfaction by fostering a sense of competence, diminishing role ambiguity, and facilitating meaningful work—specifically, the detection and prevention of fraud (Ismail, Azizan, & Fahmi, 2019).

First, accounting education as a job resource makes investigation procedures and legal frameworks clearer, which reduces confusion and mental stress. This increase in resources can make people happier, which in turn makes them more involved and diligent in finding fraud (Bakker & Demerouti, 2017; Adejumo & Ogburie, 2025). Second, specialised skills make investigative stages (such analytics, interviewing, and evidence management) more effective, leading to quick victories that boost intrinsic motivation and pleasure, which are associated with greater performance quality (Vuong & Nguyen, 2022). In accounting, higher levels of competence and possibilities for professional growth are linked to higher levels of satisfaction and better audit quality (Ali, Khinger, Subhe, & Al-Orfali, 2024).

The proposed mediation in this study aligns with Human Capital Theory, positing that returns to human capital manifest not alone as direct productivity enhancements but also through elevated job attitudes that facilitate knowledge application and resilience under duress. Research in auditing demonstrates that training and supervisory support enhance satisfaction, hence improving performance and retention (Omer, 2024). Consequently, forensic education and skills should influence fraud detection indirectly through job satisfaction, in addition to their direct impacts.

Hypothesis H3: Job satisfaction has a mediating role in the relationship between forensic accounting education and fraud detection.

Hypothesis H4: Job satisfaction has a mediating role in the relationship between forensic accounting skills and fraud detection.

The population in a study project is called the whole collection of people or events a researcher wishes to investigate or base decisions upon (Saunders & Lewis, 2017). For instance, the population of this study includes accounting practitioners.

A sampling strategy is a technique applied to choose a subset from the population. A sample is this subset. There are various sampling strategies. Firstly, the sampling strategy is divided into probable and non-probable sampling. Since all the accounting practitioners working in the private companies of Saudi Arabia are the population of this study, hence, all of them are probable to be included in this research. Thus, this study first selects probability sampling. Then, the researcher adopts simple random sampling to make sure a sufficient number of accounting practitioners are included in this research (random sampling means anyone working in the workplace can participate if he or she is an accounting professional).

In this regard, sample size is an important factor to ensure that the results from the data analysis fairly depicts the population, a bigger sample size usually yields more consistent results (Hossan, Dato’Mansor & Jaharuddin, 2023). Moreover, sufficient population and suitable sampling techniques guarantee the representation of the population, therefore helping researchers to produce legitimate results. Furthermore, the credibility of research findings depends on proper population identification, sampling strategy, and sample size in line with the research objectives (Kang, 2021). The population of this study are accounting practitioners working in private companies, while the sampling strategy is simple random sampling and the sample size is also calculated using Cochran’s formula (1977). This formula uses two key factors:

Based on that formula, a selection of 372 respondents were determined through the use of random sampling techniques.

A research measure is a tool that helps researchers get real-world data by using surveys so they can assess, or analyse, the data to see if they have met their goals (Van V. & Simmie, 2024). These tools are widely used in social sciences, health sciences, business and education to evaluate the respondents’ opinion (Ghanad, 2023). Instruments are a collection of tools used to gather data, such as surveys, observations, interviews, and other methods that assist researchers in data gathering (Hair et al., 2022). For this study, the questionnaire includes 3 parts. The cover letter is the 1st part that explains the objective of the research while gives other information. The 2nd part provides information on demographic, and the last part is about the tools that will be used to collect the data.

Moreover, the research utilises 3 items for forensic accountant education (Al Samara, 2017; Ozili, 2015; Hegazy et al., 2017; Ozili, 2020; Alabdullah et al., 2014), 4 items for fraud detection (Silverstone & Sheetz, 2004; Bressler & Bressler, 2007; Yuniarti & Ariandi, 2017), and 3 items for job satisfaction (Hancer & George, 2003; Batura et al., 2016) from prior research. The measurement table, Table 1, lists these items.

The survey data in this study was gathered through the survey questionnaire; hence questionnaires were distributed among the accounting practitioners working in the accounting firms. The researcher then collected the responses from the participants, who were distributed. This is called the drop-off/pick-up (DOPU) method (Junod & Jacquet, 2023). In data collection, drop and pick, or drop-off/pick-up (DOPU), is a survey method that involves delivering surveys in person and then picking them up later (Broussard, Allred & Ross-Davis, 2011). This method can increase response rates and reduce non-response bias (Allred & Ross-Davis, 2010). However, before the data collection, the researcher took permission from the managers of the organisations from where the data were collected.

This study employed SmartPLS M4.0 software. It was used as it helps to analyse collected data with the Partial Least Squares-Structural Equation Modelling (PLS-SEM) approach. SmartPLS enables researchers to assess complex models that incorporate numerous variables, including independent and dependent variables. Additionally, SmartPLS has capabilities for testing moderation and mediation effects (Hair et al., 2024).

The measurement model includes both validity as well as reliability. Reliability in the context of data analysis refers to the consistency of what the instrument measured (Saputra, 2025). The measurements, which are perfect, produce the same results for every replication; anything arising as unreliable is the product of random measurement errors. Though they are not always valid, obtaining dependable measures is a prerequisite for obtaining trustworthy measurements. Cronbach’s alpha was used to find the internal consistency whereby every item is regarded as a replicate measurement of the construct (Hair et al., 2024), and the suggested Cronbach Alpha coefficient for reliability is 0.7. All instruments in this study are adopted, and these were employed in the previous studies as explained in the instrumentation sections, therefore, the adopted instruments have proven reliability and dependability.

The convergent validity is the extent to which a set of variables converge to assess a specific concept (Hair et al., 2012) that also assessed in this study. In this regard, factor loadings, AVE (average variance extracted), and composite reliability (CR) was assessed accordingly recommendations of Hair et al. (2010). Moreover, composite reliability was also assessed that shows the extent to which a collection of items regularly signals the latent construct, marks the second component of convergent validity (Hair et al., 2024). In this regard, composite reliability values were then under investigation. These outcomes should support the convergent validity of the outer model.

Furthermore, discriminant validity was also assessed using the Fornell and Larcker (1981) suggestion that recommends that the AVE of every variable ought to be higher to the square of the intercorrelation that variable and other variables or the square root of the AVE of every construct should be higher than the correlation between a specific variable and other variables. This study will verify the discriminant validity of the variables using both approaches. A collinearity test was done so that there is no multicollinearity in the data (Ghanbar, 2024; Nagar et al., 2024). However, collinearity happens when there is a high level of interrelation among the variables (Subhaktiyasa, 2024). Variance inflation factor (VIF) is recommended to assess collinearity (Hair et al., 2024). If the VIF results are higher than 5 indicating collinearity among predictor constructs (Hair et al., 2024). Moreover, this study also conducted a collinearity test to make sure that the data did not have multicollinearity (Ghanbar, 2024; Nagar et al., 2024). Collinearity, on the other hand, occurs when the variables are very closely related to each other (Subhaktiyasa, 2024). Hair et al. (2024) suggest using the variance inflation factor (VIF) to check for collinearity. If VIF results exceed 5, it signifies collinearity among predictor constructs (Hair et al., 2024).

This research also employed the structural model result from PLS-SEM to test our hypotheses. The t-statistics are usually the benchmarks for testing hypotheses. If the t-statistics for a variable are greater than 1.96, it is significant at the 5% level (Namaziandost, Kargar Behbahani & Heydarnejad, 2024). The present study is capable of determining the p-value for significance level and hypothesis testing. A p-value below 0.05 indicates a 95% confidence level, while a p-value below 0.01 signifies a 99% confidence level (Kumar et al., 2024), elucidating both direct correlations and the outcomes of mediation and moderation (Kumari & Kaur, 2024; Hidayat-ur-Rehman, 2024).

The study was conducted among the accounting professionals in Saudi Arabia. In this regard, the study has found that 79% participants were male while the rest were females. Moreover, 87% were married and the rest were single. Furthermore, the study included the 45% participants from 30–40 years, while 17% from 41–50 and the rest were above 51 years old.

This was also tested in this study before the analysis of models i.e., measurement and structural. To ensure, it was completed that there is no common method bias effect on the model being tested. In this regard, the variance inflation factor (VIF) values were computed that shows that the VIF values in this study were less than 3.3 i.e., maximum value in this study was 2.665. It fulfills the recommendation of Kock (2015), who argued that if the VIF values are less than 3.3, the results would have no common method bias issue. Thus, based on this result, we proceeded to analyse the measurement model.

This study employed a comprehensive method using CA and CR to thoroughly assess the internal consistency and reliability of variables in model, as recommendation of Hair et al. (2021). Table 1 and Figure 2 below demonstrate that CA and CR values for all variables are higher than 0.7. This reveals that the internal consistency was strong. When checking for convergent validity, this study strictly followed the rules set by Hulland (1999) and Hair et al. (2021). These rules say that each item must have a FL more than 0.40 & an AVE higher than 0.50. Results in Table 1 is acceptable since it reveals that all the items had loadings > 0.70, which is far higher than the FL benchmark. Each construct also met the AVE standard, with values higher than 0.50. This outcome illustrates strong convergent validity in line to the standards of Hair et al. (2021).

This research applied HTMT and Fornell-Larcker criteria to analyse discriminant validity. In this regard, Table 2 shows the HTMT values where it compares correlations between every construct within the same constructs. According to Henseler et al. (2015) if the ratio is less than 0.90 the HTMT is fulfilled. In this regard, the results in this study as mentioned in the Table 2 shows the values are lower than 0.90, hence, the results are feasible to be argued to have good and clear discriminant validity.

Moreover, this study has also checked the Fornell-Larcker criteria where the study found that the diagonal values i.e., square roots of the AVE to the correlation results are lower. Table 3 shows that the diagonal values are more than the related correlation values. According to Fornell & Larcker (1981) the results then show higher and acceptable discriminant validity.

Results of hypothesis evaluation as depicted in Table 4 as well as Figure 3 show that every hypothesis has been accepted. In this regard, for H1 i.e., forensic accounting education (AE) positively related to the fraud detection (FD) as we found that β = 0.213, p = 0.007. Moreover, the findings show that H2 i.e., forensic accounting skills (AS) also enhance FD as we found that β = 0.204, p = 0.032. Moreover, findings of H3 i.e., the mediating influence of job satisfaction in accounting education and JS relationship was also accepted as β = 0.126, p = 0.003. Furthermore, the results also showed that H4 is also accepted that job satisfaction mediates the relationship of forensic accounting skills with JS i.e., β = 0.203, p = 0.002.

Results above substantiates that forensic accounting education and skills positively and significantly influence fraud detection among Saudi accounting professionals, with job satisfaction serving as a mediating factor for both impacts. These results align with the theoretical foundations of the study. Human Capital Theory posits that specialised education and skills enhance productivity and task efficiency (Becker, 1993), whereas the JD–R model conceptualises education and skills as job resources that activate motivational pathways, thereby increasing satisfaction and performance in challenging forensic environments (Bakker & Demerouti, 2007; 2017). The findings align with empirical literature that associates forensic training with professional scepticism, red-flag identification, interview effectiveness, and analytics-driven anomaly detection (Apostolou et al., 2010; Mohd-Sanusi et al., 2013; Atika & Rochayatun, 2023; Kinanti et al., 2023; Nilendu, 2024).

Simultaneously, they build upon previous research by investigating these effects in Saudi Arabia, a market experiencing regulatory enhancement although still exhibiting deficiencies in capability and analytics (CMA, 2023; Nazaha, 2023; PwC Middle East, 2022; Amin & Motta, 2023). Our findings provide a nuanced perspective on the points presented in the introduction, which highlighted systemic governance vulnerabilities and the prevalence of tips over controls in detection (ACFE, 2024; PwC, 2022; PCAOB, 2023). These macro-level realities persist, particularly in the context of cyber-enabled fraud and intricate supply chains (EY, 2023; OECD, 2021). However, the current findings indicate that individual-level competencies and attitudinal states significantly enhance detection, even when formal controls are inconsistent (Bonrath & Eulerich, 2024; Ziorklui et al., 2024). This is in line with what the current literature outline about how specialised skills in data analytics, e-discovery, and interviews can help traditional audit methods that have trouble with collusive or technology-driven schemes (Messier et al., 2018; IAASB, 2025; Michelet et al., 2023).

Moreover, importantly, the mediation through work satisfaction validates the JD–R motivating process and aligns with findings that resources improve satisfaction, thereby increasing task persistence and performance (Christian, 2014; Vo, Tuliao, & Chen, 2022; Narsa et al., 2023). This elucidates the reason why training alone occasionally produces minimal improvements: neglecting satisfaction factors such as autonomy, recognition, access to tools, and supportive supervision limits the return on human capital investments (Omer, 2024; Veledar et al., 2024). In Saudi Arabia, where Vision 2030 aims to improve governance and market integrity, the results show that improving forensic skills at the practitioner level, along with work design that keeps people happy, can significantly improve detection effectiveness along with ongoing regulatory reforms (CMA, 2023; Nazaha, 2023).

Findings in this research have several theoretical implications. First, it expands the JD–R concept into the field of forensic accounting by demonstrating a motivational pathway from job resources, specifically, forensic accounting education and skills to job satisfaction and subsequently to performance in fraud detection (Bakker & Demerouti, 2007; 2017). While JD–R forecasts such mechanisms in high-demand environments, empirical examinations at the micro-level of fraud detection are still limited; the validated mediation reinforces the argument that attitudinal states serve as immediate channels through which resources convert into outcomes (Christian, 2014; Del Pozo-Antúnez et al., 2018).

Second, the study enhances Human Capital Theory within a specialised assurance context by differentiating domain-specific forensic competencies (e.g., analytics, digital evidence, interviewing) from general accounting knowledge, and by illustrating both direct and indirect returns to detection quality (Becker, 1993; Ismail et al., 2019; Kinanti et al., 2023). This aligns with the HCT concept of complementarities—training generates greater returns when integrated into practical situations that facilitate application and acknowledgement (Okeke, 2016; Yaqoub et al., 2023).

Third, the results fill in gaps in fraud research that are too big or too little. Although existing literature focuses on governance, culture, and the robustness of internal audits (PwC, 2022; Bonrath & Eulerich, 2024; Xu et al., 2024), our findings indicate that individual resources and satisfaction are significant predictors of detection, even in dynamic systems such as Saudi Arabia’s (CMA, 2023; Nazaha, 2023). This theoretically prompts the development of multilevel models that combine organisational controls and market oversight with practitioner competencies and job attitudes. Finally, in a context characterised by digitisation and capacity limitations, the study enhances the generalisability of JD–R and HCT to emerging markets and proposes boundary conditions—workload, ethical climate, and technology enablement (Michelet et al., 2023; IAASB, 2025)—that future theories should explicitly integrate.

The practical implications includes the suggestion that combining capability-building with job design that makes people happier can help find more people. For practitioners, targeted enhancement in analytics, digital evidence management, and investigative interviews improves effectiveness and perceived impact, hence increasing satisfaction and commitment in complicated situations (Michelet et al., 2023; Nilendu, 2024; Kinanti et al., 2023). Accounting firms should make it a standard practice to use integrated programs that combine technical training with JD–R levers like autonomy in planning, high-quality tools and data, and supportive supervision. It is also important to keep an eye on detection KPIs along with engagement and satisfaction, since they are linked to performance (Bakker & Demerouti, 2017; Omer, 2024; Veledar et al., 2024).

Moreover, professional organisations can connect certificates and CPD with skills in e-discovery, AI-assisted analytics, and chain of custody, using regional case libraries to speed up experiential learning (Awwad & Abdelsattar, 2025; Awosika, Shukla, & Pranggono, 2024). Furthermore, the findings could be useful for the accreditation of forensic training providers, and sector-specific guidance, especially for SMEs and family businesses and policymakers. Moreover, the findings can be useful to enhance the forensic modules, hands-on laboratories, and internships in various level education programs to speed up entry-level preparation and minimise the time it takes for students to become competent (Apostolou et al., 2010; Ismail, Azizan, & Fahmi, 2019). These initiatives together address the known gap between the region’s increasing risk and the uneven maturity of practitioner analytics (PwC, 2022; EY, 2023; OECD, 2021).