LEARN. PRACTICE. GROW.Using risk as an explicit criterion
A teaching matrix assigns likelihood and impact scores and defines level as their product. A risk with 3×5 scores 15 and another with 4×2 scores 8. This supports ordering under the agreed scale; it does not automatically represent lost money or a percentage probability. Discuss concrete consequences, evidence quality, and scale limitations. In a support context, a rare closing failure can have high impact. Visibility of a cosmetic problem does not determine treatment by itself. Record defect severity separately from repair priority and explain the reasons for the decision to the relevant participants.
Calculating effort without promising duration
For a task, estimates are four person-hours in the optimistic scenario, seven in the most likely, and sixteen in the pessimistic. Applying E=(a+4m+b)/6 gives eight person-hours. Measure (b-a)/6 gives two, but should not be presented as a guarantee that the task finishes between six and ten hours. The calculation depends on the estimates and model. Effort also differs from elapsed duration: two people do not guarantee completion in four hours when dependencies, waiting, or a shared environment exist. Decompose work and state availability and precedence before converting effort into a schedule forecast.
Checking the comparison used in a historical estimate
An organization observed a development:testing effort ratio of 5:2 in comparable projects. If development is estimated at 150 person-days, testing is estimated at 150×2/5=60 person-days. The combined total would be 210, but that is not the figure to report as test effort. The ratio is not a universal standard. Compare technology, automation, criticality, experience, reuse, and definitions of included activities. If the new project requires recovery validation previously absent, the analogy may underestimate effort. Record assumptions and revisit the estimate using actual observations while retaining consistent units and scope throughout reporting.
Prioritizing new coverage while respecting dependencies
If R1,R2,R3 are already covered, a test covering R1,R2 adds zero items; another covering R3,R4,R5 adds two. With equal duration and no dependencies, the latter comes first when additional coverage is the selected criterion. Another criterion may be appropriate if risks or constraints differ. A high-risk test can depend on setup with lower nominal priority; run the necessary setup first. Do not confuse priority with an impossible execution order. Within the available window, consider environments, data, people, and sequence so that results correspond to the conditions you intend to assess.
Reporting denominators and preserving initial failures
Out of one hundred planned tests, 72 passed, eight failed, and twenty were not run. Execution is 80%, the pass rate among executed tests is 90%, and passed tests represent 72% of the plan. None of these metrics alone explains residual risk. Show which requirements and risks relate to failures and omissions. If the pipeline becomes green only after retries, retain initial results and investigate synchronization, data, and environment. Ten failures in one hundred first runs do not establish a 10% production failure rate. Population, conditions, and causes differ and require analysis.
Preparing a release decision with limited evidence
An unknown API version compromises interpretation of an integration test even when the schedule says to begin. Make unmet entry criteria visible and resolve them or explicitly agree the limitations. At completion, compare exit criteria with results. If the main failover test has not run, identify the risk, consequences, options, and who can accept residual exposure. Do not turn absence of testing into approval or a rollback document into demonstrated recovery. The report helps accountable parties decide; it neither removes their authority nor creates a guarantee unsupported by the tests performed.
With 72 passed, eight failed, and twenty pending, present 80% execution, 90% passing among executed tests, and critical-risk coverage separately. Explain why the missing failover test changes the recommendation.
Common pitfalls
Converting person-hours directly to duration; reversing ratios; presenting executed-only success without omissions; hiding retries; confusing severity with priority or a green percentage with safe-release probability.
Related topics: Risk and release · Tools and signal
A useful recommendation combines results, gaps, and consequences. Estimates and percentages need clear assumptions and denominators to support an accountable decision.
Reference: ISTQB CTFL syllabus v4.0.1 · CTFL v4.0; syllabus v4.0.1 (2024-09-15)