This is a subject of great concern to testers. You don't want to be wasting your time testing something that has already been tested, but at the same time you don't want to have any defects slip through and find that you are the one to blame because you didn't do enough testing.
Do you want to waste time in testing something that has already been tested?
Do you want to be blamed for a defect slipping through because you didn't do enough testing?
The answer to both of these questions is "No" (or it should be).
So let's take a look at a hypothetical example to get a handle on how to determine when enough is enough.
We have a retailer that assigns their stores one of four categorisations, A, B, C, or D, according to the level of income in the store's area. The most affluent areas have an A store, whilst the least affluent areas have a D store. When the store is determining the mark up on the basic price of a product, they run the basic price through a Category Pricing Engine (CPE), which calculates prices like this:
Category Factor
A 1.5
B 1.4
C 1.3
D 1.2
A product with a basic price of $10 will be priced at 10 x 1.5 = $15.00 when placed in a Cat A store, and at 10 x 1.2 = $12.00 when placed in a Cat B store. The test team has been provided with the codes and basic prices for 10 products to test the CPE functionality and results.
So how many tests do we need to do?
For complete testing, you could test each product for each category - which means 10 products multiplied by 4 categories to make 40 tests. This is unnecessary as you would be duplicating your tests. When faced with a situation where you are considering how much of the test data to use, ask yourself what would be your approach if there were 10 times as much test data. In this case, if you had the codes and basic prices for 100 products, would you conduct 400 tests to cover everything?
What you need to test is the functionality, and in this case the functionality is that a basic price is multiplied by a factor according to the store category. When considered like this, you need only 4 tests - one to verify that each category is being calculated by the correct factor.
