Information Technology Reference
In-Depth Information
Ta b l e 3 Termination states
for paths in the leaf nodes
(in %)
Leaf node
Path
Termination
13
9
8
1
Success
82
77
64
Quality problems
18
23
36
2
Success
95
95
88
Quality problems
5
5
12
3
Success
95
93
93
Quality problems
5
7
7
temperature for the food and cosmetics market OR with products for the chemicals
market with high chemical resistance requirement), while leaf node 8 is a differ-
ent group (instances with small or medium products for the food and cosmetics
market).
Note that not all the existing and known contextual variables are identified as
influencing the behavior (e.g., the supplier of the raw material was found irrelevant).
4.2 Suggesting Improvements to the Process Model
Phase 1 provides a grouping of process instances according to context groups. In
addition, these are divided into sub-groups with similar behaviors. However, for
improvement purposes a different level of granularity might be needed, both for
the paths and for the termination states. The termination state classification for the
purpose of context identification aims at creating a clear distinction of different
outcomes. Hence, it is at a coarse granularity level, relating mainly to the hard goals
of the process and possibly to a threshold over soft goal achievement levels. When
attempting to suggest improvements that would affect the business results of the
process, a finer granularity level is required, relating to different levels of soft goal
achievement. Considering the paths, some distinctions that were disregarded for the
context identification (e.g., different machines) must be taken into account, as they
might affect the outcomes for a given context.
Process improvement may include three types of action:
1. Providing criteria for path selection in a given situation. These would rely
on the paths, context groups, and outcomes achieved by process instances in
the repository. Considering the granularity level defined as relevant for process
improvement, process instances in each context group should again be clustered
based on path similarity. These clusters are then ranked based on their average
achievement of goals, so the best performing paths for each context groups can
be identified and recommended.
To illustrate path selection recommendations, below are some possible cases
concerning our running example.
 
 
Search WWH ::




Custom Search