Test Automation Test Case Design Interview Questions

Equivalence partitioning is defined exactly this way in ISTQB and other testing foundations. The core idea is that if one value in a partition catches a defect, the others in the same partition likely would too, so a single representative value per partition suffices as a baseline.

Boundary value analysis targets the edges of an input range because off-by-one errors cluster there. The 2-value approach selects each boundary value along with the value immediately outside it on each side: 0 and 1 for the minimum, and 100 and 101 for the maximum. The set of 0, 1, 100, 101 is therefore the correct choice. A set of 1, 50, 100 only samples the valid range and misses the edges, while 1 and 100 alone never test the just-outside values. The set of 0, 1, 2, 99, 100, 101 adds the values just inside each boundary, which is the 3-value variant rather than the 2-value one.

Equivalence partitioning splits the input into a valid class (8 to 64) and two invalid classes (below 8 and above 64). Boundary value analysis then picks values on each edge of the valid class and just outside them. The set of 7, 8, 36, 64, and 65 characters hits both invalid classes (7 and 65), both valid edges (8 and 64), and one representative interior valid value (36), so every partition and boundary is covered with five tests. The set of 8, 36, and 64 characters never exercises either invalid class, so it misses the just-outside boundary checks. The set of 7 and 65 characters never exercises the valid class at all. Testing every length from 1 to 100 is exhaustive and violates the requirement to avoid testing every possible length.

Boundary value analysis tests values at the edges of each equivalence class. For the inclusive range [10, 50], the inner-edge valid values are 10 (the smallest valid value) and 50 (the largest valid value). The values just outside the range are 9, the nearest invalid value below the lower edge, and 51, the nearest invalid value above the upper edge. All four values must be identified to answer fully.

The core premise of equivalence partitioning is that every value inside a partition produces the same behavior from the program under test. If one representative value passes, all others in that partition are assumed to pass as well, which is what makes the technique an efficient way to reduce the number of test cases while still covering the input domain.

2-value BVA selects each boundary value together with its neighbour just outside the valid range, giving 0 and 1 at the lower edge and 100 and 101 at the upper edge. Those same four integers also cover every equivalence class of integer input: 0 represents the invalid-low class, 1 and 100 represent the valid class, and 101 represents the invalid-high class. Because the boundary values already give one representative per class, adding an interior value such as 50 does not improve the set, which is why the minimal complete set is 0, 1, 100, 101.

Pairwise (all-pairs) testing selects a subset of combinations so that every pair of values from every pair of parameters is covered at least once. It rests on the observation that most configuration defects are triggered by interactions between two parameters rather than three or more. Counting the coverage requirement gives a lower bound: 10 parameter pairs each need 16 value combinations (160 total), and each test covers 10 of them, so at least 16 tests are required, which is far fewer than 1024.

Boundary value analysis fits discount because it is a bounded numeric range, and testing at 0 and 50 plus the values just outside (−1 and 51) targets off-by-one validation errors. State transition testing fits the order workflow because it is a finite state machine with defined allowed transitions, so verifying valid transitions succeed and invalid ones (such as shipping before confirming) are rejected is its core purpose. Equivalence partitioning fits discount because its input domain splits into a valid partition (0–50) and invalid partitions (negative, above 50, non-numeric), and one representative per partition exercises each distinct validation path. Pairwise combinatorial testing is the mismatch: it targets behaviour that depends on interactions between parameters, but the scenario states the query parameters are independent, so each can be tested on its own and covering all pairs adds cost without finding new defects.

The first blank is filled with 'pairwise' (also called 'all-pairs' or 'pairwise combinatorial' testing), the standard name for the technique that builds a covering array in which every pair of parameter values co-occurs in at least one test case. The second blank is filled with 2: the core assumption behind pairwise testing is that defects arise from the interaction of at most two parameters. If higher-order (3-way) interactions can cause failures, pairwise coverage is insufficient and a higher-strength covering array (t-wise combinatorial testing with t = 3) is needed instead.

In state transition testing, 0-switch coverage exercises each individual transition at least once, 1-switch coverage exercises every pair of consecutive transitions where one ends at the state the next begins from, and 2-switch coverage exercises every sequence of three consecutive transitions. All-state coverage only requires each state to be visited. The team's requirement, exercising every pair of consecutive transitions, is the definition of 1-switch coverage.

Standard and Flexible booking types each combine with the 3 Days-before-check-in values and the 3 Membership tier values, giving 3×3 = 9 rules each, or 18 in total. For Prepaid bookings both other conditions are don't-care, so the 9 Prepaid combinations collapse into one consolidated rule. Total distinct rules = 18 + 1 = 19.

A strength-2 covering array guarantees that every value pair from any two parameters appears in at least one generated test. It does not cover 3-way combinations, so a defect that only manifests when three or more parameters take specific values simultaneously can escape even a complete pairwise suite. The statement about covering every 3-way combination is false because triples are not guaranteed. The claim about a guaranteed minimum is false because finding the minimum covering array is computationally hard and tools return heuristic, near-minimal results. The claim about fewer tests than strength-1 is false because a strength-2 array also covers every single value, so it can never be smaller than a strength-1 array for the same parameters.

A mixed-strength (variable-strength) covering array assigns different coverage strengths to different parameter subsets. The fault model has two defect classes: interactions among the critical parameters P1–P3 (at most 2 of them at a time), and interactions between exactly one critical and exactly one non-critical parameter. Strength 2 over P1–P3 guarantees every value pair among critical parameters appears in some test, covering the 90% class. Strength 2 over each critical–non-critical pair covers the 10% class. Nothing is required for non-critical–non-critical pairs or for 3-way interactions, which the model says never cause defects. The full pairwise array over all 12 parameters also detects every fault but adds mandatory coverage of non-critical pairs the model excludes. The strength-3 array over all 12 parameters covers 3-way interactions the model says never occur as defects. The full factorial over P1–P3 with non-critical parameters fixed at one default never exercises critical values alongside non-default non-critical values, so it misses the 10% class.