Evaluating E-Learning Systems Success: A Case of Sri Lanka

E-learning, the product of technology and education, has emerged as a powerful medium of learning particularly in the higher education sector. Significance of e-learning in educational services has led to a massive growth in the number of e-learning courses and systems offering different types of services followed by the COVID 19 pandemic. Thus, evaluation of e-learning -systems is critical to ensure successful delivery, effective use, and positive impacts on learners. Survey data of Sri Lankan university student’s sample tested the research hypotheses. Quantitative assessment of determinants of PLS-SEM results confirms 81.7% explanatory power of the predictor variables in explaining the variance of E-Learning Success among which Instructor Quality, Learner Quality, Service Quality, Support System Quality and Technical System Quality found significant predictors. Implications invite revisiting the theoretical models to assess the e-learning effectiveness by incorporating multifarious factors while developing elearning system of multifractality found critical for the success of any electronically driven learning experience.


INTRODUCTION
The advancement of information technology has paved way to success in many sectors such as health, finance, transportation, and agriculture…etc. In line with that wave of e-transformation, education sector too has integrated technology to its various deliverables in order to meet the expectations of stakeholders effectively. As such, Electronic Learning (E-Learning) simply is the technological adoption of education that is been vastly practiced by many educational institutions nowadays.
Choudhury and Pattnaik (2020) defined elearning as transfer of knowledge and skills, in a well-designed course content that has established accreditations, through an electronic media like the Internet, Web 4.0, intranet and extranets. The main stakeholders of e-learning include learners, faculty, administrative & technical staff, and employers (Choudhury & Pattnaik, 2020). E-learning found to be having a greater effect on academic performance (Abbasi et al., 2020;Almaiah et al., 2020;Ebner et al., 2020;Maldonado et al., 2011, Radha et al., 2020. It encompasses a range of activities: from supported learning to blended learning and to pure elearning (Cruz-Jesus et al. 2016;Aboagye, 2021;Radha, 2018). Online learning systems provide benefits for stakeholders located around the world. Advantages of e-learning for learners include an increased accessibility to information, better content delivery, personalized instruction, content standardization, accountability, on-demand availability, self-pacing, interactivity, confidence, and increased convenience. Minimization of costs, enabling a consistent delivery of content, and improved tracking are among the other benefits of e-learning to the faculty (Sander, 2020;Al-Maroof, 2021). E-learning reduces classroom and facilities cost, training cost, travel cost, printed materials cost, labor cost, and information overload (Sander, 2020;Choudhury & Pattnaik, 2020;Al-Maroof, 2021). E-learning initiatives call for considerable investments in technology such as hardware costs, software licenses, learning material development, equipment maintenance, and training (Abbasi et al., 2020;Al-Maroof, 2021). Al-Maroof (2021) concluded that e-learning has huge potential and can reduce costs in comparison to a traditional class room environment after initial course development.
Nowadays, educational technologies have quickly evolved along with the prompt development of ICTs (Al-Emran andShaalan 2015, 2017;Salloum et al., 2017;Ali et al., 2018). The last two decades have witnessed an increase in the prevalence of the internet due to the reason that universities and other educational institutions have made investments in information systems (For instance Moodle, Blackboard, Google Class…etc.) so as to help in face-to-face as well as distant course delivery (Tarhini et al., 2013;Teo et al., 2020). Using e-learning along with networked computers facilitates transmitting the digitized knowledge from the online sources to the end-user devices, such as a laptop, desktop and handheld devices (Misra et al. 2014;Behera 2013;Salloumet al., 2019, Shahmoradi et al., 2018. In par with the other parallel development, the global online learning industry is sporting massive annual growth of 19% or more per year, and it's set to be a $243 billion industry within next two years following the COVID 19 pandemic (Sander, 2020). The United States is still at the forefront of the industry in terms of market size, but other regions such as Europe, Latin America, and Asia are also starting to become increasingly prevalent players in the industry (Sander, 2020). The demand for e-learning platforms also raised due to the COVID 19 pandemic situation in the world. The lockdowns restricted physical presences and encouraged to continue studies with e-learning platforms. Consequently, the education activities across the globe are moving with the aid of e-learning systems where the quality of elearning systems matters today more than ever.
Among the top most concerns of the e-learning are the quality of elearning deliverables. This has received a substantial level of attention by scholars resulting immense number of research outcomes those attempted to clarify different facets of elearning quality (Ali & Ahmad,2011;Fathema, Shannon & Ross, 2015;Mohammadi, 2015;Mtebe & Raphael, 2018;Sander, 2020). In a nutshell, majority of these studies have examined individual aspect of key determinants of e-learning systems success ignoring the synergistic effects of all determinants affecting the success of e-learning systems (Eom & Ashill, 2018;Janelli, 2018). Alternatively, some has looked in to the direct relationships between e-learning quality factors and usage or satisfaction which is again not addressing the system as a whole operating unit (Janelli, 2018;Mtebe & Raphael, 2018;Sander, 2020).
Success of e-learning systems found to be multifaceted (Sander, 2020). Hence, any assessment should primarily account both the individual effect and the combined effect of the predictors. Additionally, the level of influence reported to vary by the context itself too (Janelli, 2018;Ebner et al., 2020). On account of the fact that e-learning success factors vary in terms of their relative significance based on the context, different strategies have been adopted to deal with these factors.
For  (2003) once reported that technology integration within education in developing countries is lagging due to cultural, political and economic concerns where the objective of e-learning is to provide basic education to a large number of poor students. This is very different from the objective of e-learning in developed countries, which aims to develop an effective knowledge economy and enhance lifelong education (Gulati, 2008;Hubalovsky, 2019). Regardless of these challenges, opportunities still exist to improve the effectiveness and success of elearning (Ebner et al., 2020;Thiyagarajan & Suguanthi, 2021). Besides, the critical evaluation of the success factors of e-learning systems will aid in satisfying the expectations of all its stakeholders. As an emerging nation, Sri Lanka has a great potential to move forward with technological advancements. Investments in such pre-assessed, well-planned and goal-oriented technological systems are of greater demand than unplanned and ad-hoc investments on system development or modification. Therefore, the various aspects which determine the success of e-learning systems will be a prime concern for its further developments and meeting leaner needs. Motivated by these empirical lapses, the present study focused on evaluating the e-learning system success referring to the context of a developing country; the Sri Lanka. attitudes and interactions, those play a vital role in e-learning success (Cheng,2011;Liaw, Huang & Chen, 2007;Selim, 2007;Al-Samarraie et al., 2018). Yet, only very few if not no studies have analyzed how the collective effect of e-learning success factors can explain the success of e-learning systems. With the widespread use of e-learning platforms, a further investigation is timely important to evaluate the success of e-learning systems accounting observed multi dimensionality of the construct.

Success of e-learning systems
As stated by Alireza, Fatemeh and Shában (2012) the emergence of modern technologies has promised to provide equal educational opportunities everywhere for everyone and also, diverse courses continuously. In fact, without considering the main components of learning, application of the most advanced and latest technology is in vain, and will have merely advertising aspect rather than educational. On the other hand, since unsuccessful effort in implementing e-learning is reflected in terms of return on investment, the success of elearning is one of the important issues (Govindasamy, 2002). In an e-learning system, not only the learner, but also all stakeholders are important. It is no doubt that internet and other digital technologies are able to support e-learning in an open, flexible and distributed environment. But how? Due to the differences between e-learning and traditional learning in some aspects, effective and successful conversion of traditional courses to e-learning may need a complex attempt and requires accurate planning, monitoring and control (Cantoni, Cellario, and Porta, 2004;Bhat et al.,2018;Fernando et al., 2019). In fact, continuity of global demand growth for elearning and acceptance of virtual communities needs to measure their effectiveness and usefulness in education (Valencia-Arias ;2019, Chopra et al., 2018.  The ESSS model is one which includes seven independent constructs: technical system quality, information quality, service quality, educational system quality, support system quality, learner quality, and instructor quality. In addition, there are four dependent constructs: perceived satisfaction, perceived usefulness, system use, and benefits. Accordingly, the research model for the study developed as presented in figure 2.

Research Hypotheses
The hypotheses developed based on the connections in the model are presented in this section.

System quality (SQ)
In the original model of Delone and McLean (2003) the researchers assumed that system quality directly affects use and user satisfaction. Several researchers applied the DeLone and McLean model in the information systems context and found a positive association between system quality and use (Halawi, McCarthy, and Aronson, 2008;Po-An Hsieh and Wang, 2007;Iivari, 2005;Tularam, 2018). In the e-learning systems context, system quality was also proved to be strongly related to use (Balaban, Mu, and Divjak, 2013;Garcia-Smith and Effken, 2013;Lin, 2007;Marjanovic et al., 2016). Based on these findings, researchers therefore, assume that the higher the technical quality of the elearning system, the more satisfied the users are. Also, if users find the e-learning system compatible with their requirements, this would positively make users utilize it and consider it useful. Thus, the following hypothesis is proposed: H1: Technical system quality positively influences success of e-learning system

Information quality (IQ)
The relationships between information quality and each of the three constructsuse, satisfaction, and usefulnesshave been studied empirically by e-learning researchers. For example, Klobas and McGill (2010) and c) found a significant relationship between information quality and both use and satisfaction with the Learning Management System (LMS). The relationship between information quality and perceived usefulness was found significant in the study of Chen (2010) with e-learning systems in an organizational context, and a similar result found by Lwoga (2014) with web-based LMSs. Therefore, we may assume that improved quality of information in the e-learning system will positively lead to an increase in perceived usefulness, perceived satisfaction, and system usage. Thus, we hypothesize that: H2: Information quality positively influences success of e-learning system

Service quality (SQ)
The construct has been utilized in the information systems field. For example, the relationship between SRQ and satisfaction was confirmed by Chen and Cheng (2009) in an online shopping system. The direct relationship between SRQ and use was found significant by Wang and Liao (2008) in an egovernment system. Similarly, in the context of e-learning, the relationship between SRQ and satisfaction was found significant in the Roca et al. (2006) and  models. The relationship between SRQ and perceived usefulness proposed in the conceptual model developed by Pham (2019), Hagos, Garfield, and Anteneh (2016) and Lwoga (2014) was shown empirically to be significant in the study conducted by Al-Sabawy (2013) and Ngai, Poon, and Chan (2007). Accordingly, the following hypothesis are proposed: H3: Service quality positively influences success of e-learning system Hassanzadeh et al. (2012) found that educational system quality positively and directly influences user satisfaction and indirectly the use of the system, which indicates that educational features in the e-learning system, and facilities like discussion forums, chat-rooms, collaborative learning tools, can result in user satisfaction and maximizing their usage of the e-learning systems. Social interaction was employed as a key factor of success in computer supported collaborative learning (CSCL) and found to have a significant effect on student learning (Xing, Kim, and Goggins, 2015;Nikolić, 2018;Nikolić, 2019). The relationship between educational system quality and perceived usefulness was found significant for webbased e-learning systems in the study undertaken by Liu, Liao, and Peng (2005) (2015) found a positive relationship between educational system quality and satisfaction. In addition, the relationships between diversity in assessment materials, and learner interaction in the e-learning system with perceived satisfaction, were found significant by Cidral et al. (2018). Further, the relationship between educational system features and usefulness was found significant by Liu et al. (2005) for a web-based e-learning system. The same results were obtained by Liaw and Huang (2013) where a significant relationship between the interactive learning environment construct with both perceived usefulness and perceived satisfaction was found. Therefore, the following hypothesis about educational system quality are proposed:

Educational system quality (ESQ)
H4: Educational System Quality positively influences the success of e-learning system

Support system quality (SUP)
In the literature on e-learning system success, supportive issues in the e-learning system such as ethics and policies that outline rules, regulations, guidelines and prohibitions to communicate within the e-learning system, assignments' plagiarism rules, data protection, and other legal and copyright issues of the uploaded materials in the elearning system, in addition to the popularity and policy followed by the organization, all these issues influence the learners significantly (Khan, 2005). For example, in the empirical study conducted by , the use of the LMS at Brunel University has increased significantly due to the encouragement students and academics received from the university to use the LMS in their modules. The researchers stated "the use of U-Link has increased significantly during the last three years. This is mainly because of the increasing popularity of elearning portals." The researchers studied the relationship between supportive system issues and satisfaction and found it significant. On the other hand, the organizational promotion of the e-learning system significantly and positively affected employees' satisfaction in the study conducted by Navimipour and Zareie (2015). As stated by (Al-Fraihat et al, 2020), the popularity of the elearning system, and the policy followed by the organization to promote their e-learning system, play an important role in increasing the usage of the system by academics and learners. Therefore, researchers propose the following hypothesis: H5: Support System Quality positively influences the success of e-learning system

Learner quality (LER)
This construct was successfully operated in several models developed by prior e-learning researchers. Several researchers examined a subset of the learner quality construct, for example, the learner's self-efficacy was studied by Ong, Lai, and Wang (2004) and a significant relationship with perceived usefulness was found. The same result was achieved by Park (2009). McGill and Klobas (2009) ;Rakic et al.(2020) studied the relationship between learner attitude toward LMS use and LMS utilization and found it significant. Additionally, the relationships between student involvement and both use and satisfaction were found significant in the study of Klobas and McGill (2010). Also, the relationships between selfefficacy and a learner's computer anxiety with perceived usefulness were studied by Chen and Tseng (2012). The relationship between learner and perceived satisfaction was found significant in the models of Sun et al. (2008) and . Given the positive relations of the indicators associated with the variety of learner's characteristics, it is more likely that the quality of the learner will influence perceived usefulness and use of the system. Thus, propose the following hypothesis: H6: Learner Quality positively influences the success of elearning system

Instructor quality (INS)
According to (Al-Fraihat et al, 2020) the instructor's role in the success of e-learning has received attention from researchers in the e-learning arena. To clarify, the model developed by Sun et al. (2008) researched the relationship between the instructor dimension, using two indicators (instructor response timeliness, instructor attitude toward e-learning), and satisfaction, and found it positively significant. Similar results were obtained by Cidral et al. (2018) where a positive relationship found between instructor attitude toward elearning and user's satisfaction. Lwoga (2014) employed instructor quality as a separate construct and confirmed a positive significant relationship between instructor quality and both perceived usefulness and user satisfaction. Also, instructor quality has been found to have a significant effect on learners' satisfaction with an e-learning system in the study conducted by Mtebe and Raphael (2018). Thereby the following hypothesis is proposed; H7: Instructor Quality positively influences the success of elearning system

METHODS
Present study adopts EESS model (Al-Fraihat et al.,2020) based on its greater explanatory power and inclusion of wider range of predictive variables such as technical, human and social. An empirical study of quantitative approach tested the EESS model based on the LMS of a Sri Lankan state university; Wayamba University of Sri Lanka. Data collection led by instrumentalization of an online questionnaire among the level 03 undergraduates who are enrolled to Moodle based LMS of Wayamba University of Sri Lanka. Moodle was selected to test the model of the study because the University of Wayamba has adopted Moodle as the main e-learning system designed to support teaching and learning materials and activities, and to provide a number of interactive activities including forums, wikis, quizzes, surveys, chat and peer-to peer activities, serving most of the departments and students. In addition, Moodle is widely used in the education sector generally and in higher education specifically. The online survey assessed the success of ELS which is a Moodle based LMS. Sample size determination followed the "10-times rule method" which is a commonly used classic rule for deciding the sample size of Partial Least Square -Structural Equation Modelling (PLS-SEM). (Hair et al., 2011;Peng and Lai, 2012). There, the sample size should be greater than 10 times the maximum number of inner or outer model links pointing at any latent variable in the model (Goodhue et al., 2012). This yielded 70 (7*10) sample units whistle the researchers succeed in drawing 263 valid responses via online survey of selected group. The sampling frame was a list of ELS IDs of all internal undergraduates of WUSL. Using lottery method, 5 times of minimum required sample size was drawn so as to avoid the potential problem of low responses. Resultantly, authors received 263 out of 350 (75%) emailed questionnaires. Undergraduates, the study's unit of analysis offered an evaluation of the ELS properties based on system design, system delivery, and system outcome. The refined instruments (Al-Fraihat et al, 2020) based on measurement model validity and reliability indexes composed of 52 items falling in to seven exogenous variables (predictors) namely, Technical System Quality (TSQ), Information Quality (IQ), Service Quality (SQ), Educational System Quality (ESQ), Support System Quality (SSQ), Learner Quality (LQ), and Instructor Quality (IQ). The endogenous variable; E-Learning System Success (ELSS) contained of four reflective first-order constructs namely, Perceived Satisfaction (PS), Perceived Usefulness (PU), and Use (U) and Benefits (B) of ELSs. 5-point Likert scale was the measure of the responses in which the "1" stands for "Strongly Disagree" and 5 denotes "Strongly Agree". The questionnaire was pre-tested for its clarity and easy understanding through a pilot study and the face validity was achieved by obtaining the experts views of the same. Partial Least Square -Structural Equation Model (PLS-SEM) deemed to be well explaining the relationship of nexus of latent variables. It generates less contradictory results compared to regression analysis and facilitates analyzing the relationships of multiple independent and dependent variables. Further, PLS-SEM is good at increasing the parsimonious of the analysis (Hair, et al., 2011;Hair, et al., 2014;Ringle, et al., 2012;Wong, 2013). Hierarchical Component Model (HCM) of the collected data was developed using Smart PLS version 3.

The endogenous variable, E-Learning
System Success composed of four first order measures of which the measurement model was first analyzed for its reliability and validity (Hair, Sarstedt, Ringle, and Mena, 2012). Factor loadings of all the items leading to four constructs satisfy the threshold value 0.7 at the 95% confidence level (Table 1). For all measures of internal consistency, all constructs scored well above the threshold value of 0.7 as recommended by Nunally (1978). This indicates the high reliability of all four first-order constructs. Factor Loading and Average Variance Extracted (AVE) are considered standard measures of the convergent validity (Hair, et al., 2017;Byrne, 2016;Bagozzi, and Yi, 1998;Fronell, and Larcker, 1981). The AVE values of these constructs fall in between 0.700 and 0.859. Convergent validity of the constructs considered adequate when the AVE exceed 0.5 (Bagozzi, and Yi, 1998;Fronell, and Larcker, 1981). Additionally, factor loadings of latent variables those greater than 0.708 theorized to be explaining minimum 50% or more of the indicator's variance of it (Hair, et al., 2017). Here, the factor loadings of all the indicators of the firstordermodel are between 0.726 and 0.942. Accordingly, it is evidenced that the all constructs satisfy the convergent validity criterion.
Next, the firstorderconstructs are examined for their discriminant validity. For an acceptable level of discriminant validity, Fronell, and Larcker, (1981) recommended that the AVE of a latent variable should be higher than the squared correlations between the latent variables and all other variables (Chin, 2010;Chin, 1998b;Fronell, and Larcker,1981). Table  2 demonstrates the correlation matrix with the square roots of AVEs on the diagonal line (in Bold) which indicates an acceptable level of discriminant validity according to Fronell, and Larcker criterion (i.e. AVE criterion). Additionally, cross loadings are also used as a discriminant validity measure where it is expected for each indicator to load highest on the construct it is associated with (Henseler, et al., 2015;Voorhees, et al., 2016). Examination of loading of each indicator on its respective latent variable ensured that all are loaded highest on the latent variable for which they are assigned. Thus, all the constructs of first-order model confirmed to be holding acceptable level of discriminant validity.  Table 3 shows the key measures of validity and reliability of the second-order constructs.  (Hair, et al., 2017;Byrne, 2016;Bagozzi, and Yi, 1998;Fronell, and Larcker, 1981). The AVE values of these constructs fall in between 0.534 and 0.813. Convergent validity of the constructs considered adequate when the AVE exceed 0.5 (Bagozzi, and Yi, 1998;Fronell, and Larcker, 1981). Additionally, factor loadings of latent variables those greater than 0.708 theorized to be explaining minimum 50% or more of the indicator's variance of it (Hair, et al., 2017). Here, the factor loadings of all the indicators except U, TSQ1, TSQ10, TSQ2, and TSQ9 of the second-order model were between 0.700 and 0.941. Yet, U, TSQ1, TSQ10, TSQ2, and TSQ9 respectively loaded 0.537, 0.690, 0.635,0.695, and 0.669 on their corresponding latent constructs. Based on Byrne's (2016) recommendation for factor loadings equal to or greater than 0.5, indicator U is accepted since the AVE value of the construct is greater than 0.5 (0.732). Further, the factor loadings equal to or greater than 0.6 can also be accepted, provided that the corresponding AVE value is greater than 0.5 (Byrne, 2016). The contributing AVE value of TSQ latent construct is 0.534. Accordingly, it is evidenced that the all constructs satisfy the convergent validity criterion. Next, the second-order constructs are examined for their discriminant validity. For an acceptable level of discriminant validity, Fronell, and Larcker, (1981) recommended that the AVE of a latent variable should be higher than the squared correlations between the latent variables and all other variables (Chin, 2010;Chin, 1998b;Fronell, and Larcker,1981). Table  4 demonstrates the correlation matrix with the square roots of AVEs on the diagonal line (in Bold) which indicates an acceptable level of discriminant validity according to Fronell, and Larcker criterion. Additionally, cross loadings are also used as a discriminant validity measure where it is expected for each indicator to load highest on the construct it is associated with (Henseler, et al., 2015;Voorhees, et al., 2016). Examination of loading of each indicator on its respective latent variable ensured that all are loaded highest on the latent variable for which they are assigned. Thus, all the constructs of second-order model confirmed to be holding acceptable level of discriminant validity.  Sarstedt, 2014). Significance of path coefficient can be assessed using P value and t-value of the path. As such, path coefficients of those the P value is less than 0.05 (for 95% confidence level) and tvalue greater than 1.96 (for 2tailed test) are considered significant (Hair, et al., 2017). Bootstrapping of second-order model results that some paths are statistically significant while some doesn't (Table 5).  (2020) Except Education System Quality -> E-Learning System Success path (P = 0.249, t value = 1.154) and Information Quality -> E-Learning System Success path (P = 0.816, t value = 0,233) all other paths possess significant path coefficients, where P < 0.05 and t value > 1.96 (Hair, et al., 2017). The significant paths should be next assessed for their multicollinearity model (Hair, Hult, Ringle, and Sarstedt, 2014). Variance Inflation Factor of PLS algorithm is used in deciding on the possible multicollinearity issues. As to Hair, et al., (2017) no multicollinearity will be presented if the VIF values are less than 5.0 (Hair, et al., 2017). VIF values of all the inner model constructs are well below the threshold value (< 5.0). Hence, it is confirmed that the structural model constructs are free of multicollinearity problems (Table  6). Now the path significance is assessed and the absence ofmulticollinearity is ensured. Next coefficient of determination (R 2 ) is examined to weight the explained variance. PLS algorithm of second-order model resulted in 0.817 of R 2 value. Based on independent variables' ability to account 81.7% variance of the dependent variable, it is concluded that there is a substantial level of influence by the E-Learning system qualities on the Success of E-Learning System (Hair, et al., 2017;Chin, 1998;Cohen, 1988). The above R 2 value is depicted in the structural model of figure 3. The effect size (f square) of PLS algorithm is the next measure of the structural model. The effect size is defined as "the increase in R 2 relative to the proportion of variance of the endogenous latent variable that remains unexplained" (Cohen, 1988, Henseler et al., 2009). As to Hair et al., (2017) and Cohen (1988), 0.35f 2 value is regarded larger effect size, 0.15 -f 2 value: medium effect size and 0.02 f 2 value equals to smaller effect size. Table 7 contains the effect size of corresponding latent constructs. Based on the decision criterion, Education system Quality and Information Quality appear not having any effect while other independent variables possess small effect size on the variance of R 2 (Cohen, 1988, Henseler et al., 2009. Finally, the structural model is assessed for its predictive relevance via the blindfolding (q Square). Stone-Geisser Predictive relevance suggests that the Q 2 value larger than 0 (0 <) indicates that exogenous constructs have predictive relevance over endogenous construct (Stone, 1974;Geisser, 1975;Hair, et al., 2017). The blindfolding of Construct Cross validated Redundancy results 0.579 Q 2 value which is well above the threshold value of 0. It implies that the E-learning system qualities have predictive relevance over E-Learning System Success.  (Kim, Trimi, Park, and Rhee, 2012;Mohammadi, 2015;Xing, Kim, and Goggins, 2015) and Information Quality (Klobas & McGill, 2010;Eom et al., 2012;Chen, 2010;Lwoga, 2014). E-learning system here has viewed as a triangular conception in which teacher (i.e. facilitator) and the learner interact with each other via a technical platform: the system. All the significant predictors to e-learning success found closely attached to either of these three pillars of an elearning system. Hence, their power in affecting the success of e-learning system can be ramified. For instance, two leading predictors, Instructor Quality and Learner Quality directly associated with two of three pillars. They represent the live components of E-learning systems who are the contributors and as well the beneficiaries of Elearning systems. Thus, it is inveterate that the instructors and leaners to a greater extent should be responsible for success of the e-learning systems to which they are connected. Additionally, the rest of the significant predictors; Support System Quality, Technical System Quality, and Service Quality are elements of the other pillar; the technical platform. Hence, the results proved that the success of elearning system in a way is a communal contribution of all three parties to the system. On contrary, the insignificant factors appear loosely connected with either of the three main components of the e-learning system. Thus, the findings are believed to be revealing the factuality of the presumed relationships. The implications flag that the organizations need to emphasize on creating learning opportunities, knowledge sharing, and tapping knowledge at both individual & corporate levels developing an e-learning culture in the process. Additionally, the study spotlighted the fact that even though global delivery of elearning is highly talked about, the real potential of e-learning depends on the local environment to a large extent (Ali, 2008). Elearning undoubtedly is a source of competitive advantage (Choudhury & Pattnaik,2020). However, as pointed out in the paper, partners to the learning experience need to observe the changing dynamics of the learning environment and should follow an agile approach that enable adoption and diffusion of e-learning tools on a continuous basis.

CONCLUSION
Students today are exposing to different learning environments to gain the maximum value in learning experiences. Natural and man-made disasters such as COVID 19 pandemic, often threaten the continuation of physical learning experiences. This, together with many other socio economical drivers, have thrived the demand for elearning. Every institution is unique and has its own strengths in conducting online courses. The essence of quality education, in any form, is to ensure that learning objectives are achieved efficiently and effectively, without sacrificing the standards of the educator and institution. Although recent attention has increased e-learning evaluation, the current research base for evaluating e-learning is inadequate.
Given the significance of the investments in implementing e-learning programs, the assessment of their success / effectiveness can't be misjudged (American Society for  Training  and  Development , 2001). In that light, the present study contributes to the existing body of knowledge of e-learning systems by proposing a new model of assessing the e-learning systems success. Findings confirms that the proposed model holds superficial capacity to predict the variances in e-learning system's success. The previous studies (Al-Fraihat, et. al., 2020) offer confirmation of the theoretical implications of the EESS model in the context of developing countries. Study supports the practical implication of ensuring not only the technical systems quality but also the instructor, learner, service, support system, and technical system in any attempt to enhance the success of e-learning systems (Abbasi et.al, 2020: Aboagye et.al, 2021: Ali et.al., 2018: Mtebe and Raphael, 2018