Software Testing

Intermediate path · lesson 4 of 19

IntermediateTest design5 min readPublished November 12, 2026

Decision table testing

When a result depends on several conditions at once, decision tables list every combination and the expected outcome. They turn tangled business rules into clear test cases, and expose the rules nobody wrote down.

On this page

Key takeaways

  • A decision table lists conditions, actions and one column (rule) per combination of conditions. Each feasible column becomes a test.
  • Building the full table reveals gaps and contradictions in the requirements, often before any code exists.
  • Tables can be collapsed with "doesn't matter" entries to save tests. Keep the full table for high-risk rules.

Where this picks up

Equivalence partitioning handles one input at a time. But business rules often depend on combinations: "approve if the credit history is good and the income is high enough, unless…"

Testing each condition on its own misses bugs in how they combine. Decision tables are built for exactly that.

Anatomy of a decision table

A bank's loan rules:

  • Customers with a bad credit history are rejected.
  • Good credit and an income of at least 3,000 → approved.
  • Good credit, lower income, existing customer → sent for manual review.
  • Good credit, lower income, new customer → rejected.

Three yes/no conditions give 2 × 2 × 2 = 8 combinations. Each one is a column:

R1R2R3R4R5R6R7R8
C1 Good credit historyYYYYNNNN
C2 Income ≥ 3,000YYNNYYNN
C3 Existing customerYNYNYNYN
Approve✓✓
Manual review✓
Reject✓✓✓✓✓

The top part holds the conditions, the bottom part the actions. Each column is a rule, and a test case.

Building one, step by step

  1. List the conditions. Each should be a clear question: yes/no, or a small set of values.
  2. List the actions: every outcome the system can produce.
  3. Fill in all combinations. A neat trick: in the first row, alternate in halves (YYYYNNNN), in the next in quarters (YYNNYYNN), and so on.
  4. Fill in the actions for each column from the requirements.
  5. Mark gaps. Any column the requirements don't answer is a question for the product owner.
  6. Remove impossible columns, such as "age under 18" and "retired", if the system can't produce them.

Collapsing the table

Look at R5–R8: with a bad credit history, the result is always Reject, whatever the other conditions. They can be merged into one column with "–" (doesn't matter):

R1'R2'R3'R4'
C1 Good credit historyNYYY
C2 Income ≥ 3,000–YNN
C3 Existing customer––YN
ActionRejectApproveManual reviewReject

Four tests instead of eight. The trade-off: you now test only one of the four bad-credit combinations. If a developer wrote the rule as "bad credit and new customer → reject", the collapsed table might miss it.

Coverage

ISTQB defines decision table coverage as the number of feasible columns exercised by tests, divided by the total number of feasible columns. Full coverage means one test per feasible column.

For each column, choose concrete values. "Income ≥ 3,000: Y" could be 3,000 exactly, a boundary value, which combines both techniques.

Beyond yes and no

Conditions can have more than two values. These are called extended-entry tables:

  • Customer type: new / regular / VIP
  • Delivery: standard / express / pickup

The number of columns is the product of the values: 3 × 3 = 9. When tables grow too large, look at pairwise testing to reduce them, or split the rules into several smaller tables.

Real-world exampleThe column nobody had written

A tester built a decision table for a gym membership app. The rules covered students and seniors with a discount, and regular members without.

The full table had a column the requirement didn't mention: a customer who is both a student and a senior. Do they get one discount, both added together, or the bigger one?

The developer had implemented "both added together", giving 70% off. The product owner had assumed "the bigger one". The question took five minutes to settle at the requirement stage, and would have been a pricing bug in production.

Lesson: decision tables test the requirement as much as the code. Every empty column is a question worth asking.

Do

Show the decision table to the product owner and developer before testing. Ask them to fill in the columns you couldn't answer.

Don't

Guess the expected result for a combination the requirements don't cover. You'll test your own assumption, not the business rule.

Check yourself

A shipping rule depends on three yes/no conditions and one condition with three values (standard, express, pickup). How many columns does the full decision table have?

Show answer

2 × 2 × 2 × 3 = 24 columns. If some combinations are impossible (for example, express delivery for pickup orders), remove them. If the table is still too large, collapse it or split it into smaller tables.