Answer: (ii) Input conditions at their limits

Boundary Value Analysis (BVA) is a black-box (functional) testing technique that designs test cases using the boundary values of the input domain — the values at, just above, and just below the edges of each valid/invalid partition. The rationale is that errors tend to concentrate at the boundaries of input ranges (e.g., for a range [1, 100], test 0, 1, 2, 99, 100, 101).

Why the others are wrong:

• (i) Average input data — typical/mid-range values are not the focus of BVA; boundaries are.
• (iii) Randomly selected inputs — random testing is a different approach (e.g., fuzzing).
• (iv) Data from previous projects — historical data is not how BVA selects cases.

BVA therefore designs test cases based on input conditions at their limits — option (ii).