According to Bollen and Long (1992), not only testing a
proposed model but also comparing rival models has
become an emerging consensus in structural equation
modeling (SEM). Our model posits that functionality and usability influence their outcomes only through the key
mediating variable of customer satisfaction. Because of
limited empirical support, our model does not permit direct
paths from functionality and usability to purchase intentions.
A nonparsimonious rival model would be the one
positing only direct paths from each of the precursors to
the outcome (Morgan and Hunt, 1994). There is no indirect
relationship in the rival model (Fig. 2). For example,
functionality and usability are not allowed to mediate any
of the relationship. Though there is no theoretical
foundation to support the rival model, there are some
discussions and empirical studies which consider functionality
and usability to be independent variables which
directly influence customer purchase intentions (White and
Manning, 1997; Liu et al., 2000).