There are both strengths and weakness in basing our
study on only one case company. A weakness is, of course,
that we cannot claim that our case is empirically representative.
On the other hand, sampling a sufficiently
large number of Lean companies would most likely have
forced us to use cross-sectional survey methods and, consequently,
to rely on respondents’ memories. The benefit of
our approach is the access to a very rich amount of archival
data extending across many years, collected using (almost)
the same measurement instruments over the years (Lean
audits and tiered contribution margins), especially instruments
not created for research purposes. Consequently,
the reliability is very high. Nonetheless, a cross-sectional
survey addressing some of the same hypotheses could
be interesting from an external validity point of view.
Building a bridge between these types of studies would
certainly improve validity, if they provide consistent
results.
Using the Euclidian distance in a regression model with
performance indicators as the dependent variables has certain
limitations. Despite being recommended as the best
statistical technique for addressing a systems fit model
(Gerdin and Greve, 2004), its ability to measure the impact
of individual variables is somewhat limited (Drazin and Van
de Ven, 1985). However, this method is intended for systems
fit models, and these models focus on the impact of
a package (group) of independent variables on a dependent
variable, rather than the impact of single independent
variables.
A limitation of our study is that we cannot claim to
have found causal relations between the control forms
and performance. We can only claim to have found significant
statistical relationships between the datasets.
However, we have validated the results with the actors
of the organization in order to test whether they match
their perceptions of how their Lean program intends
to affect performance. The actors believed in synergy
effects coming from the package of Lean audit scores on
performance, but they have never applied sophisticated
statistical techniques to test this.