Ask a project team for their Requirement Traceability Matrix and you often get a spreadsheet that was filled in the night before an audit, with every row marked “Covered” in green. That version is useless. A real traceability matrix is a living map that answers questions nobody can answer otherwise, and it is worth understanding what those questions are before deciding how much effort to put into one.

What it is, stripped down

A Requirement Traceability Matrix (RTM) records links between requirements and the things that verify them. At its most basic it has one row per requirement and columns for the test cases that check it. Most teams extend it to include test execution status and any defects raised against those tests. Some go further and link requirements to design documents or code modules.

A small slice for a loan origination system might look like this:

Req IDRequirementTest casesStatusDefects
LN-101Applicant must be 21 to 60 years oldTC-201, TC-202, TC-2033 passed–
LN-102PAN must be verified before credit checkTC-210, TC-2111 passed, 1 failedDEF-88 (open)
LN-103EMI shown must match the sanction letterTC-220Not run–
LN-104Rejected applicants receive reason by SMS–––

Four rows already tell a story. LN-104 has no tests at all. LN-102 has an open defect. LN-103 has a test that has not run. Someone looking only at the overall pass rate would miss all three.

Forward, backward and both

Traceability runs in two directions, and each direction answers a different question.

Forward traceability goes from requirement to test. It answers: is every requirement being checked? The empty cell for LN-104 is a forward traceability gap. Without the matrix, that gap is usually found by a customer.

Backward traceability goes from test to requirement. It answers: why does this test exist? A test that traces to no requirement is either checking something undocumented (which may mean the requirement is missing), testing something that has been removed, or simply waste. In large regression suites, backward traceability is how you find the hundreds of tests that nobody dares delete because nobody knows what they protect.

Bidirectional traceability is having both. Most standards and regulated industries that ask for traceability mean this.

Three questions it answers that nothing else does

1. What does this change break? Suppose the business changes the age rule in LN-101 from 21–60 to 21–65. The matrix immediately lists TC-201 to TC-203 as tests that must be updated. If the requirement had been linked to design and code as well, you would also see which modules are affected. This is impact analysis, and on a large system it can be the difference between updating the right three tests and discovering a fourth one failing in production.

2. Can we release? A release decision based on “92 percent of tests passed” is weak, because the failing 8 percent might all sit on one critical requirement. Rolling status up by requirement turns the question into something business owners understand: “Requirements LN-101 and LN-103 are verified; LN-102 has an open defect in PAN verification; LN-104 is untested.” A product owner can make a real decision from that.

3. Can we prove it? In banking, insurance, pharma, medical devices and aviation, auditors and regulators may want evidence that each requirement was verified. The RTM, with links to execution records, is that evidence. Indian banks and NBFCs going through IT audits, or vendors building systems for them, frequently face this. The matrix is not just a QA convenience there; it is part of the deliverable.

How matrices go wrong

Requirements that are too big. If a single row reads “User management module”, one passing test can mark it covered while ninety percent of the behaviour is unchecked. Traceability is only as precise as the requirements. Rows should be small enough that one person can say, without hedging, whether they are verified.

Counting links instead of judging coverage. A requirement with one test is “covered” in the same way as a requirement with fifteen. The matrix shows that a link exists, not that the tests are good. LN-101 is an age range, so it needs boundary tests at 20, 21, 60 and 61 at least; three tests may or may not be enough depending on what they check. Someone still has to look.

Maintaining it by hand in a spreadsheet. Hand-kept matrices drift out of date within weeks, because nobody updates the spreadsheet when they rename a test or split a story. Teams that keep traceability useful usually let their tools do it: linking test cases to user stories inside Jira with a test management plugin, or in Azure DevOps, TestRail, qTest or similar. The matrix becomes a report generated on demand rather than a document anyone edits.

Only filling it at the end. An RTM built after testing tells you what happened. One built during test design tells you what is missing while there is still time to fix it.

Traceability in agile teams

Agile teams sometimes treat the RTM as a waterfall relic. The need does not go away; it changes shape. User stories carry acceptance criteria, each criterion maps to one or more tests (often automated), and the link lives in the story tracker or in the test code itself through tags. When a BDD scenario is tagged with the story ID it verifies, the pipeline can produce a traceability view on every build. That is the same matrix, kept current automatically, without a separate document.

The level of rigour should follow risk and regulation. A marketing microsite does not need formal traceability. A system that calculates loan EMIs, processes insurance claims or controls medical dosing does, and in those settings a well-kept matrix is one of the cheapest forms of protection a team can have.

A quick self-check

If your current matrix is genuinely useful, you should be able to answer each of these in under five minutes: Which requirements have no tests? Which tests trace to nothing? Which requirements currently have open defects? Which tests need updating if requirement X changes? If any of those takes an afternoon of digging, the matrix exists on paper but not in practice.

Written by the RP Strativa Team. The requirement and test IDs in the example are illustrative.

Leave a Reply

Trending