Database Reference
In-Depth Information
Table 4: Additional Analysis of the WebSurvey (*** = sign. 99 %, ** = sign. 95 %, * = sign. 90 %)
R1
R2
R3
R4
R6
R5
R7
Corr
R_avg
0.912
(***)
0.877
(***)
0.855
(***)
0.862
(***)
0.912
(***)
0.875
(***)
0.209
Corr H 0.890
(***)
0.515 (*)
0.827
(***)
0.579
(**)
0.890
(***)
0.943
(***)
0.000
R8
R9
R10
R11
R12
R13
R14
Corr
R_avg
0.482
0.592 (**)
0.912
(***)
-0.048
0.565 (*) 0.706 (**)
0.534
(*)
Corr H 0.401
0.601 (**)
0.890
(***)
-0.306
0.311
0.749
(***)
0.047
with the highly signifi cant correlation between the heuristic and the group's average opinion
of 0.810, we cautiously conclude a positive relation between our cohesion metric and the
corresponding intuition of experts on this matter . Obviously, the set of respondents and
questions is very small. Furthermore, we have no hard evidence that the presented dilemmas
look like real practical problems; it is based on the author's personal experiences only.
The considerations of the respondents give us some insight into the limits of a cohesion
metric like the one we defi ned. When a design consideration is very specifi c, the cohesion
metric may be a bad implementation. However, when these considerations are less explicit
or mixed, then the cohesion metric seems like an attractive and valid quantifi cation thereof.
An expert group's opinion is then reasonably well refl ected by the metric.
CONCLUSIONS
On the basis of our evaluation, the author is positive about the value of the cohesion
metric for both distinguishing weakly cohesive activities and the support it can offer to
decide between design alternatives. Obviously, these results must be interpreted with cau-
tion, as discussed earlier. In particular, the interviews with workfl ow designers showed that
very specifi c design considerations are not well implemented by the cohesion metric. The
presented cohesion metric should be used when more explicit considerations do not lead to a
decisive result. In this sense, it can be used as the 'fi nishing touch' for a workfl ow design.
The possibilities to extend this research are many. A following step may be the use of
the cohesion metric in question in an actual project, which involves the design of a work-
fl ow process in a real setting. Several of the respondents have indicated their willingness to
cooperate within such a practical test. It would be a good opportunity to test the heuristic
on real design dilemmas.
As stated before, the introduced cohesion metric only supports the designer in making
a decision with respect to activity defi nition. It does not suggest any clustering or ordering
itself. An extension of the heuristic so that it effi ciently generates optimal activity defi ni-
tions itself is the ultimate but challenging next step of this research. One interesting idea is
Search WWH ::




Custom Search