the ‘tangibles’ dimension. Every stratum had a factor for the tangibles dimension. This
may reveal that for users of these remote channels, the aspect of equipment/facilities
quality constitutes a separate evaluation from overall service quality. This conclusion is in
line with their remote nature, as the physical aspect of the branch and the equipment used
is a secondary aspect for these channels’ users.
The splitting of the ‘tangibles’ dimension is interesting in light of the results of
Parasuraman, Berry, and Zeithaml (1991) in their assessment of the SERVQUAL scale.
In their re-evaluation of the scale, the tangibles dimension that originally (Parasuraman et
al., 1988) was unidimensional, appeared subdivided into two dimensions, one pertaining
to the physical facilities/equipment, and the other to employees/communication materials.
• Perceived ease of use did not emerge as a distinct factor. Although not expected, it is
in accordance with the in-depth interviews carried out, where this attribute was not
mentioned as relevant for the usage decision of a technology-based delivery channel.
• Unexpected factors, such as ‘Perceived service quality of technology-based / human-
based channels’, suggest that channels have intrinsic aspects of service quality that are not
revealed in a common service quality dimension covering service quality aspects that are
generic to the company whatever the distribution channel considered.
Due to the exploratory nature of the research, to length restrictions of the questionnaire
and to the original nature of some of the constructs of the model, it was decided not to use
scales already defined in the literature, except for the case of the perceived service quality
construct. The measures used were newly developed by the author, based both on existing
measures and on insights brought by the in-depth interviews conducted.
Reliability and validity
In empirical research, there is the need to determine the extent to which a particular
measurement represents a certain construct. Latent variables are not observed, and so the
validity and reliability of their measurement have to be established. The measures (i.e., the
constructs) used were assessed in terms of reliability and validity. Almost all variables
satisfied the recommended reliability values for exploratory research. Some of the variables
with lower reliability values were retained due to their theoretical interest (Nunnally, 1967;
Murphy and Davidshofer, 1988). Discriminant validity was assessed by comparing if the
correlation coefficient between a scale and any other scale was lower than its Cronbach’s
alpha coefficient (Gaski and Nevin, 1985; Andaleeb, 1995). From the overall results, it was
16