Information Technology Reference
In-Depth Information
Main Process
Subprocess Level 1
Subprocess Level 2
Work Instruction
The level „Work Instruction“ permits in turn variants by using the variables
„Start Condition“ and „Input“.
The combination of the three items „Work Instruction“, „Start Condition“ and
„Input“ can be identified unequivocally with the help of a code number
following the process hierarchy.
Acceptance Criteria
For each identifiable test case there exist acceptance criteria
consisting of
expected output and
objective (with reference to the work instruction).
Fig. 3.1 Process hierarchy
In no case the customer should be satisfied by just accepting the test scripts from
the development tests of the supplier, since those are oriented along purely techni-
cal criteria without any focus on the business processes themselves. In this context
it is important to emphasise that testing has to be carried out on the level of work
instructions. This level is depicted in Fig.
3.1
.
Figure
3.1
presents a basic schema. With respect to the complexity of the main
process the hierarchy can be more or less elaborate. The lowest level is the decisive
one, the work instruction level, serving as the yard stick for the test scripts.
When defining test data, the requirements of test scripts have to be taken into
account—especially concerning the combination of possible variations of data
items (variants). From those the data contents and quality are deduced. In some
cases they can be satisfied by a dump from the production data base. In this case one
has to decide whether partial dumps suffice or whether the productive data base has
to be uploaded in its totality. In the latter case one has to make sure that the capacity
of the test system is sufficient and that expected performance will be met.
If the functions are to be tested on the basis of new data to be generated or
employ new algorithms, real data often do not suffice. In these cases synthetic data
have to be provided—sometimes in an initially empty data base. If more than a few
dozen data records are required, the effort to build up bulk complex synthetic data
has to be planned. If possible these data should be generated automatically by
specially written programs.
Another aspect when selecting test data is performance. As already mentioned,
negative performance should not impede acceptance procedures unnecessary. On
the other hand there could be special test scripts explicitly designed to test the
performance of certain functions. This is obviously then the case when for example
performance considerations are behind re-designs of the data model in the first
place.
