Database Reference
In-Depth Information
If it looks confusing to you, you are not alone. It certainly looks Greek to us (pun intended). For most of us, it
has been many years, if ever, since we have worked with advanced mathematics. And although we could go back
and relearn this calculation, it may not be time well spent.
Instead, what we need is an example that can be used to test the calculation. To get one, have someone who
is proficient at performing these calculations create a set of inputs and outputs. For instance, they provide three
or four examples as
test cases
. The test team then uses each of the provided inputs to create outputs using the BI
solution and subsequently verify that the proper output always occurs.
Of course, this means that more documentation must be generated in order for the test team to efficiently do
their job. This can become an issue as management tries to coordinate the documentation effort in conjunction
with the testing effort. And although we discuss documentation further in the next chapter, for now keep these
points in mind:
•
Simple documentation is better than no documentation.
•
Use test case documentation even on small BI solutions.
•
Provide input and output examples for both simple and complex operations.
teStING FOr SMaLL SOLUtIONS
often, something as simple as an Excel spreadsheet can prove almost as useful as professional testing software.
For example, you can create a spreadsheet that maps the test cases for your team to validate (Figure
18-2
).
using this spreadsheet as a guideline, your testers can review your inputs and expected outputs, create reports
or queries using the data warehouse or cube, and compare the actual outputs to the expected outputs. After each
test, the testers note the results in a pass/fail column along with any additional useful notes.
Figure 18-2.
A simple way to track test cases with Excel
This is a very low-tech option, but remember,
something
is much better than
nothing
. All too often testing
is left out of the process because it is believed to be too costly to implement on small Bi solutions. From
our example you can see that this is a mistaken assumption. Keep in mind that once the credibility of your
