Efficiency, security and Sharia compliance drive mobile banking satisfaction in Afghanistan

Mobile banking in Afghanistan appears to succeed or fail on a handful of very specific qualities, according to a new survey-based study of customers of Azizi Bank’s mobile application. Researchers from the study, published in Discover Analytics, distributed a structured questionnaire to active users of Azizi mobile banking services in six major Afghan cities and analyzed the responses with partial least squares structural equation modeling. Of the eight service quality dimensions they tested, only four turned out to be statistically significant drivers of customer satisfaction: efficiency, privacy and security, Sharia compliance, and site organization. The remaining four dimensions, including reliability and responsiveness, which are staples of classical service quality theory, showed no statistically significant relationship with satisfaction in this sample.

The study employed a quantitative, deductive research design guided by a positivist paradigm. The target population comprised individuals who had used Azizi mobile banking within the past 30 days, verified through a screening question at the start of the questionnaire. Using non-probability purposive sampling, the researchers initially collected 300 responses and retained 250 valid ones after excluding 50 incomplete, duplicated, or low-quality submissions. An a priori power analysis in G*Power, assuming eight predictor constructs, a medium effect size, a significance level of 0.05 and statistical power of 0.80, indicated a minimum sample of 109 respondents, so the final sample comfortably exceeded that requirement. The questionnaire was developed in English, translated into Pashto and Dari, reviewed by bilingual checkers, and pretested with 15 respondents before distribution through WhatsApp groups, email, Facebook and Instagram.

The theoretical backbone of the study is the E-SERVQUAL framework, an instrument for measuring electronic service quality that the authors modified for the Afghan context. The conventional six dimensions, site organization, responsiveness, reliability, user friendliness, personal needs, and efficiency, were retained, and two additional dimensions were added: Sharia compliance, capturing customers’ expectations that services conform to Islamic principles by avoiding riba (interest) and gharar (excessive uncertainty), and privacy and security, reflecting concerns about the protection of personal and financial data. The authors describe Sharia compliance as a genuinely novel extension of the framework, while privacy and security is treated as a contextual adaptation of an established e-service quality dimension.

The demographic profile of the sample is heavily skewed in ways readers should keep in mind. All 250 respondents identified as Muslim. Men made up 86 percent of the sample (216 respondents) versus 14 percent women (34 respondents). The largest age group was 18 to 23 years old, at 34 percent, followed by 24 to 29 year olds at 26 percent, so the sample skews young. A majority held bachelor’s degrees (65.2 percent), and the largest occupational category was a mixed “other” group at 31.6 percent, followed by entrepreneurs and the self-employed at 23.2 percent. Geographically, Kabul contributed the largest share of respondents at 32.8 percent, with the remainder distributed across Nangarhar, Balkh, Kandahar, Paktia and Herat provinces.

Descriptive statistics painted a broadly positive picture. Composite mean scores across the nine measured dimensions ranged from 4.10 for reliability to 4.17 for efficiency on a five-point scale, with an overall satisfaction mean of 4.12. Privacy and security scored 4.15, user friendliness and personal needs 4.14, and Sharia compliance 4.13. Standard deviations were low, between 0.53 and 0.59, indicating strong consensus among respondents. The lowest score belonged to reliability, and a minimum response of 1 on the item measuring error-free performance suggested that a small subset of users had encountered transaction failures or service inconsistencies.

The measurement model performed largely as expected. Factor loadings ranged from 0.738 to 0.957, all above the conventional 0.70 threshold, with customer satisfaction items loading highest at 0.943 to 0.957. The reliability construct showed a Cronbach’s alpha of 0.950 and composite reliability of 0.891, with an average variance extracted of 0.673, above the 0.50 benchmark for convergent validity. Discriminant validity checks were mostly satisfactory but not flawless. Under the Fornell-Larcker criterion, the square root of the average variance extracted for reliability (0.820) was slightly exceeded by its correlation with user friendliness (0.849), a marginal violation the authors attribute to the practical reality that reliable systems are often perceived as easy to use. Three construct pairs also marginally exceeded the conservative 0.85 HTMT threshold, though all remained below the more lenient 0.90 cutoff and their bootstrap confidence intervals included values below 0.85.

Multicollinearity diagnostics flagged one construct. Variance inflation factors ranged from 2.344 to 4.562 for seven of the eight predictors, well below the conservative threshold of 5, but user friendliness registered a VIF of 7.258, indicating substantial conceptual overlap with other constructs, particularly reliability and efficiency, and its coefficient should be interpreted with caution. The structural model explained 79.0 percent of the variance in customer satisfaction (R2 = 0.790, adjusted R2 = 0.783), with both values statistically significant in bootstrapping. Effect sizes told a consistent story: efficiency had the largest effect (f2 = 0.151, a medium effect), while Sharia compliance (0.077), privacy and security (0.055), site organization (0.040), personal needs (0.034), user friendliness (0.030) and reliability (0.021) showed small effects, and responsiveness (0.017) fell below the 0.02 threshold for a small effect. Positive Q2predict values between 0.657 and 0.706 for the three satisfaction indicators indicated strong out-of-sample predictive relevance.

The hypothesis tests delivered the study’s headline result. Efficiency was the strongest and most significant driver of satisfaction (β = 0.380, p < 0.001), followed by privacy and security (β = 0.217, p = 0.015), Sharia compliance (β = 0.194, p = 0.010) and site organization (β = 0.157, p = 0.017). Reliability, despite a moderate positive coefficient of 0.241, did not reach statistical significance (p = 0.097). Responsiveness (β = −0.120, p = 0.213), user friendliness (β = 0.015, p = 0.874) and personal needs (β = −0.001, p = 0.992) were not supported; the authors caution that the negative coefficients for responsiveness and personal needs are statistically non-significant and should not be read as evidence of meaningful negative effects.

The authors offer tentative, explicitly untested explanations for the non-significant findings. Reliability may function as a hygiene factor, a basic requirement whose adequate provision does not increase satisfaction, and it had the lowest composite mean score in the sample. Responsiveness may matter less in highly automated self-service environments where direct interaction with service personnel is limited. User friendliness and personal needs may be uniformly perceived as satisfactory, reducing their power to explain variance in satisfaction. Each of these interpretations was not directly examined in the cross-sectional data and is presented as a hypothesis for future work rather than an established mechanism. The authors also suggest that the non-significant dimensions may operate indirectly through mediating variables such as trust or perceived value.

The study’s limitations are considerable and stated plainly. The cross-sectional design prevents causal inference and cannot track how perceptions change over time. Reliance on self-reported data leaves the results open to common method and social desirability bias, and the authors recommend complementing surveys with objective usage data such as transaction frequency and session duration. The focus on a single bank’s customers in six major cities limits generalizability, particularly to rural Afghanistan, and factors such as brand reputation, digital literacy and infrastructure quality were not modeled. The authors call for longitudinal designs, qualitative follow-up on the non-significant dimensions, comparative studies across other Islamic banks and regions, and cross-cultural replication in other Muslim-majority countries. Within those bounds, the practical message for Afghan banks is specific: prioritize strict Sharia compliance, transaction efficiency, cybersecurity investment and professional, well-organized application design, while maintaining adequate performance on reliability and responsiveness as basic operational requirements rather than statistically supported satisfaction drivers.

Subject of Research: A cross-sectional survey study of mobile banking service quality dimensions and customer satisfaction among Azizi Bank users in Afghanistan using a modified E-SERVQUAL framework analyzed with PLS-SEM.

Article Title: Mobile banking service quality and customer satisfaction in the Afghanistan banking sector through a modified E-SERVQUAL framework

Article References: Sahar, R., & Munawaroh, M. (2026). Mobile banking service quality and customer satisfaction in the Afghanistan banking sector through a modified E-SERVQUAL framework. Discover Analytics, 4(1), Article 46. https://doi.org/10.1007/s44257-026-00097-y

Image Credits: AI Generated

DOI: 10.1007/s44257-026-00097-y

Keywords: mobile banking, customer satisfaction, E-SERVQUAL, Afghanistan, Islamic banking, Sharia compliance, PLS-SEM, service quality, cybersecurity, survey research, Azizi Bank, digital finance