ISTQB Advanced (CTAL-TA v4.0) Mock Exam #3 — Questions & Answers
Every question in this mock exam, with the correct answer marked and a written rationale behind each option below — for reading and review, not a timed run.
Chapter 1 · The Tasks of the Test Analyst in the Test Process — 8 questions
A Test Analyst joins a project for a municipal parking-permit portal. At which test levels is the Test Analyst, as described in the CTAL-TA syllabus, typically most involved?
Static analysis and code review sessions, where the Test Analyst measures cyclomatic complexity, checks compliance with the team's secure coding guideline and signs off each merge request before it enters the main branch.
Static analysis of code and complexity measurement are Technical Test Analyst tasks; the Test Analyst's static work targets requirements, user stories and models.
Performance and load testing, where the Test Analyst configures virtual users and analyses response-time percentiles under peak permit-renewal traffic.
Performance testing is covered by the Technical Test Analyst and the CT-PT specialist syllabus, not by the Test Analyst role.
Component testing and component integration testing, where individual code units and their interfaces are verified against the detailed design, the coding standards and the unit-level acceptance criteria agreed with developers.
Component and component integration testing are structural, developer-facing levels that belong to the Technical Test Analyst and developers, not the Test Analyst.
System testing, system integration testing and acceptance testing, where business-facing behaviour of the integrated product is verified against requirements and user needs.
The syllabus places the Test Analyst at the business-facing levels: system testing, system integration testing and acceptance testing.
The Test Analyst works at the levels where the product is judged from the business and user perspective: system, system integration and acceptance testing. Structural, code-facing and performance work belongs to the Technical Test Analyst.
Before the first execution cycle of a hospital pharmacy system, the Test Analyst orders the test cases into suites, defines a short set of checks that confirm the build is stable enough to test, and documents the pre-conditions each suite needs. Which test activity is being carried out?
Test implementation, which organises test cases into procedures and suites, prepares a smoke test and establishes pre- and post-conditions before execution starts.
Building suites, defining a smoke test and documenting pre-/post-conditions are the test implementation tasks named in the syllabus.
Test execution, which runs the suites against the build, compares actual with expected results and logs each anomaly found during the run.
Test execution starts once suites exist and the build is under test; the described work prepares for execution rather than performing it.
Test design, which turns prioritised test conditions into high-level and low-level test cases, identifies the test data each case will need and decides which cases are candidates for automation.
Test design produces the test cases themselves; arranging them into executable suites with a smoke test is the following activity.
Test analysis, which examines the test basis for completeness and testability, derives prioritised test conditions from the pharmacy requirements and records the traceability from each condition back to its requirement.
Test analysis studies the test basis and derives test conditions; ordering suites and preparing smoke checks happens after design, not at this stage.
Sequencing test cases into procedures and suites, creating a smoke test and defining pre- and post-conditions are the defining tasks of test implementation.
During test execution of an airline baggage-tracking system, which TWO of the following are tasks the CTAL-TA syllabus assigns to the Test Analyst? (Choose two.)
Looking for defect clusters in the executed results so that additional exploratory or confirmation testing can be aimed at the modules that are failing most.
The syllabus asks the Test Analyst to watch for defect clusters during execution and to react by intensifying testing where they appear.
Recalculating the project's earned value and updating the release budget forecast based on the number of test cases that passed in the current cycle and the defects still open against the release.
Budget and earned-value tracking are Test Manager responsibilities described in the CTAL-TM syllabus, not tasks of the Test Analyst.
Analysing each reported anomaly to decide whether it is a product defect, a test defect or an environment problem before a defect report is raised.
Anomaly analysis, including recognising false positives caused by tests or environments, is an explicit execution-time task of the Test Analyst.
Profiling database query plans to identify the slowest SQL statements executed by the baggage-scan service under the regression workload.
Query-plan profiling is technical, white-box work assigned to the Technical Test Analyst, outside the Test Analyst's business-facing scope.
Rewriting the build pipeline scripts so that the nightly regression suite is triggered automatically after every successful merge to the main branch and its results are posted to the team channel.
Pipeline scripting is an automation engineering task, typically owned by the Technical Test Analyst or developers, not a Test Analyst execution task.
During execution the Test Analyst analyses anomalies (separating real defects from false positives) and monitors for defect clusters. Pipeline engineering, budgeting and database profiling belong to other roles.
A payroll system stores salary, bank account and tax identifiers for every…
A payroll system stores salary, bank account and tax identifiers for every employee. The Test Analyst must decide between copying production data and generating synthetic data for the system test. According to the syllabus, which statement best describes when synthetic test data is the more appropriate choice?
When service virtualization is not available and the payroll interfaces must be exercised end to end with real downstream banks, synthetic data is preferable.
Availability of service virtualization concerns interface simulation, not the choice of data source, and real bank interfaces would still require compliant data.
When the test environment is shared with the operations team and must stay identical to production in structure and content, synthetic data is preferable because it can be loaded faster and refreshed without coordination.
Keeping an environment identical to production argues for production-like data; loading speed is not a criterion the syllabus uses to choose between the two.
When personal or regulated data cannot be exposed in test and the required boundary and negative cases are rare or absent in production, synthetic data generated to specification is preferable.
The syllabus recommends synthetic data where privacy rules or missing edge cases make production data unsuitable, provided it is designed to cover the needed conditions.
When the team wants the test data to age realistically over time, reflect the exact volume and statistical distribution of real employee records and reproduce historical anomalies, synthetic data is preferable to a production copy.
Realistic volume, distribution and data ageing are the strengths of production-derived data; synthetic sets rarely reproduce them without extra effort.
Synthetic data is chosen when production data would breach privacy rules (GDPR, HIPAA) or does not contain the boundary and negative cases needed; production data wins on realism and volume, which is why the syllabus also discusses pseudonymisation and anonymisation.
A pharmacy ordering module calls an external drug-interaction service that is not yet available in the test environment, and the module itself is invoked by a scheduler that also does not exist yet. Which pair of test doubles does the Test Analyst need, and for which side of the module?
A stub to stand in for the unavailable drug-interaction service that the module calls, and a driver to invoke the module in place of the missing scheduler.
A stub replaces a called component and returns canned responses; a driver replaces a calling component and triggers the unit under test. Both are named correctly here.
A driver to stand in for the unavailable drug-interaction service that the module calls, and a stub to invoke the module in place of the missing scheduler.
This reverses the two terms: a driver calls the module under test, while a stub is called by it, so the assignments are swapped.
A stub for both sides, one returning canned drug-interaction results and another one periodically invoking the module in place of the scheduler.
A component that actively invokes the module under test is a driver by definition; calling it a stub misclassifies the calling-side double.
A mock object for the drug-interaction service and a service-virtualization layer that replaces the entire module so that the scheduler can be tested alone.
Replacing the module under test would leave nothing to test; service virtualization is an alternative to stubbing the external service, not to the module.
Stubs replace components that the item under test calls; drivers replace components that call the item under test. The external service needs a stub, the missing scheduler needs a driver.
In a keyword-driven framework for an online tax-filing system, the keyword 'SubmitAnnualReturn' internally performs 'LogIn', 'OpenReturnForm', 'EnterIncomeFigures' and 'ConfirmSubmission'. How does the syllabus classify 'SubmitAnnualReturn' and what is its main benefit?
It is a domain-layer keyword that must be re-implemented for every supported browser; its benefit is that failures are reported at the level of individual clicks and field entries, which simplifies debugging for the automation engineer.
Browser-specific implementation belongs to the test interface layer, and reporting at the click level is a property of low-level steps, not of a composite keyword.
It is an atomic keyword because it maps to a single business outcome; its benefit is that it can be executed without any test interface layer.
A keyword that is itself made of other keywords is composite, not atomic, and every keyword still needs the test interface layer to reach the application.
It is a verification keyword because it ends by confirming submission; its benefit is that each check it performs is reported separately, so a failed assertion can be traced to a single step without re-running the sequence.
The keyword performs a sequence of actions with a confirmation step; classifying it purely as verification ignores the action keywords it contains.
It is a composite keyword built from atomic keywords; it lets business-level test cases stay short and readable while the detailed steps are maintained in one place.
Composite keywords aggregate atomic ones, keeping test cases concise and centralising changes to the underlying sequence.
Composite keywords group atomic keywords into a higher-level, business-readable step. They keep test cases concise and put the step sequence in one maintainable place.
A reviewer flags a low-level test case for a school grading system: it contains…
A reviewer flags a low-level test case for a school grading system: it contains eleven steps, of which five re-check preconditions that the previous step has already established, and its expected result repeats the same assertion three times in different words. Which test case quality criterion from the syllabus is most clearly violated?
Completeness, because a test case with repeated assertions is likely to have left out other expected results that should have been listed.
Nothing in the description says results are missing; the defect is surplus content, which is the opposite failure to incompleteness.
Conciseness, because the test case contains redundant steps and repeated assertions that add length without adding verification value.
Conciseness requires a test case to contain no unnecessary steps or repeated content; the redundancy described violates exactly this.
Consistency, because the expected result is phrased in three different ways and therefore contradicts itself, leaving the executor unsure which formulation to compare against.
Rephrasing the same assertion is repetition, not contradiction; consistency would be violated only if the versions demanded different outcomes.
Feasibility, because eleven steps with repeated checks cannot be executed in the time normally allotted to a single low-level test case.
Feasibility concerns whether the case can be executed at all in the environment; a long but executable case is still feasible.
Among the nine quality criteria (correctness, feasibility, necessity, understandability, traceability, consistency, precision, completeness, conciseness), redundant steps and repeated assertions are a conciseness defect.
A utility company is replacing its legacy electricity-billing engine with a new…
A utility company is replacing its legacy electricity-billing engine with a new one. Tariffs combine time-of-use rates, seasonal multipliers and social discounts, so the correct invoice amount for an arbitrary meter reading cannot be worked out reliably by hand. The legacy engine is still running in production and is trusted for the existing tariffs. The Test Analyst needs an expected result for thousands of generated meter readings before the migration cut-over. Which approach to the test oracle problem is the most appropriate first choice here?
Use metamorphic testing: double every meter reading and check that the new engine's invoice amount doubles, treating any other ratio as a failure.
Metamorphic relations are useful where no reference exists, but with tiered tariffs and fixed discounts the invoice does not scale linearly, so this relation would produce false failures.
Use property-based testing: assert that every generated invoice amount is a positive number with two decimal places and no reading is billed twice.
These properties are valid sanity checks but far too weak to confirm tariff correctness; they cannot detect a wrong multiplier or a missed discount.
Use the legacy billing engine as a pseudo-oracle: feed identical readings to both engines and investigate every invoice where the amounts differ.
A trusted existing implementation of the same function is the classic pseudo-oracle; it provides expected results at scale for the tariffs it already supports.
Use a human oracle: ask a billing specialist to calculate a stratified sample of 40 invoices by hand and accept the engine if all 40 match.
A human oracle for 40 cases gives narrow coverage of thousands of readings and ignores the trusted automated reference that is already available.
When a trusted implementation of the same functionality exists, it can act as a pseudo-oracle, supplying expected results for large generated data sets. Metamorphic relations, human oracles and generic properties are weaker fits for a tiered-tariff calculation with a legacy reference available.
Chapter 2 · The Tasks of the Test Analyst in Risk-Based Testing — 3 questions
During risk assessment for a national e-prescription service, the Test Analyst separates the factors that influence the likelihood of a failure from those that influence its impact. Which grouping matches the syllabus?
Likelihood: severity assigned by the product owner. Impact: priority assigned by the developers, adjusted by the test manager during risk monitoring.
Severity and priority are attributes of defects that have already been found, not the inputs the syllabus lists for assessing the likelihood and impact of a product risk.
Likelihood: complexity of the feature, quality of the test basis and use of new technology. Impact: frequency of use, financial and reputational damage, legal and safety consequences.
Technical and process factors drive how likely a defect is; business, legal and safety consequences and how often the feature is used drive the impact of a failure.
Likelihood: frequency of use, financial and reputational damage, legal and safety consequences. Impact: complexity of the feature, quality of the test basis, use of new technology and the experience level of the development team.
This reverses the two groups: business consequences describe impact, while complexity and basis quality describe how likely a defect is to exist.
Likelihood: number of stakeholders interviewed, size of the risk workshop and the number of checklists consulted. Impact: number of items already in the risk register, the retrospective findings and the age of the oldest open risk.
Workshop size, interview counts and register length are properties of the risk identification process, not factors used to assess an individual risk.
The syllabus lists technical factors (complexity, new technology, quality of the test basis) as drivers of likelihood, and business factors (frequency of use, financial or reputational damage, legal and safety consequences) as drivers of impact.
A public-transport ticketing platform releases every two weeks. The automated…
A public-transport ticketing platform releases every two weeks. The automated regression suite has 6,200 tests and takes 19 hours; the release window allows 5 hours. The team has: a requirements traceability matrix that maps every test to a requirement; a defect history showing that 61% of escaped defects in the last year were in fare-calculation and refund modules; commit logs listing the modules changed in each release; and code-coverage reports per test. For the next release only the refund workflow and the mobile-app login screen were changed. The Test Analyst must select the tests to run. Which selection approach is best supported by the available information, and why?
Impact analysis using the commit logs and the traceability matrix to select the tests that exercise refund and login functionality, supplemented by the fare-calculation tests flagged by defect history.
The syllabus names impact analysis as the most reliable selection technique; the commit logs and traceability matrix make it directly applicable, and defect history adds the historically risky fare module.
History-based selection alone, running every test that has failed at least once in the past year plus every test attached to the fare-calculation and refund modules, because the defect history is the most objective data set the team has and the failing modules are known.
History-based selection ignores what actually changed this release; the login change is new and would be missed if no login test has failed before.
Coverage-based selection, running the smallest set of tests whose combined code coverage reaches the suite's total coverage figure, because it minimises execution time while preserving the coverage level that stakeholders have come to expect from the full suite.
Coverage-based selection optimises for breadth, not for the changed areas; a minimal covering set may exercise refund code only incidentally and misses risk information.
Random selection of the tests that fit in five hours, re-drawn for each release, because it removes tester bias and eventually covers the whole suite over successive releases.
Random selection leaves the two changed areas to chance; the syllabus lists it as inferior to risk-based and impact-based approaches when change information exists.
When the changed components are known and traceability exists, impact analysis is the most reliable way to pick regression tests. Defect history is a useful supplement for chronically risky modules, but on its own it ignores what changed in this release.
A risk workshop for an e-prescription service produces this risk register: R1…
A risk workshop for an e-prescription service produces this risk register: R1 'dosage unit converted incorrectly between mg and ml' (likelihood high, impact very high); R2 'prescription PDF renders with the pharmacy logo misaligned' (likelihood high, impact low); R3 'patient with two active insurance plans is billed to the wrong plan' (likelihood medium, impact high); R4 'search for a drug by brand name returns results sorted differently from the specification' (likelihood low, impact low). The Test Analyst has to propose risk mitigation through testing. Which TWO proposals are consistent with the risk levels? (Choose two.)
Apply domain testing and decision table testing to R1 with full boundary and rule coverage, and give it the earliest execution slot so that any failure is found while there is still time to fix it.
The highest-risk item deserves the most rigorous techniques and the earliest execution; domain and decision table testing fit unit conversion rules well.
Treat R2 with the same depth as R1, designing a full decision table for every combination of logo size, paper format and printer driver and executing it in the first test cycle, because its likelihood is also rated high and cosmetic defects are visible to every pharmacist.
Likelihood alone does not determine test depth; a high-likelihood, low-impact risk does not justify the intensity reserved for the highest risk level.
Defer all testing of R1 to the final acceptance week, when the dosage conversion library will be most stable and the clinical reviewers are available, so that the tests do not have to be re-run after each fix and the acceptance sign-off can rely on a single execution.
Postponing the highest risk to the end of the schedule contradicts risk-based test ordering; late discovery of a very-high-impact defect leaves no time for correction.
Cover R3 with a decision table over insurance-plan combinations and a set of scenario tests for dual-coverage patients, executed after R1 but before the lower risks.
A medium-likelihood, high-impact risk still warrants systematic technique-based testing and an early slot, ordered after the very-high-impact risk.
Allocate R4 the largest share of exploratory sessions because sorting defects are hard to specify in advance, exploratory testing is the technique the team enjoys most, and a visible sorting problem would be noticed by every pharmacist in the first week.
A low-likelihood, low-impact risk should receive the lightest coverage; team preference is not a factor in risk-based allocation.
Risk-based testing allocates technique depth and execution order by risk level: the very-high-impact conversion risk gets rigorous techniques first, the high-impact billing risk follows, and the low-impact cosmetic and sorting risks get the lightest treatment.
Chapter 3 · Test Analysis and Test Design — 22 questions
What is the purpose of adaptive random testing compared with plain random testing?
To spread the randomly generated test inputs more evenly across the input domain, because failure-causing inputs tend to cluster in contiguous regions.
Adaptive random testing selects each new input to be far from those already used, exploiting the observation that failures cluster in regions of the input space.
To replace the random generator with an operational profile so that the most frequently used inputs are tested proportionally more often than rare ones.
Weighting inputs by usage frequency describes operational-profile-based random testing, a different refinement from adaptive random testing.
To provide a recognised coverage criterion for random testing, so that the percentage of the domain exercised can be reported to stakeholders.
The syllabus notes that random testing has no generally recognised coverage criterion; adaptive random testing does not introduce one.
To remove the requirement for an automated oracle by comparing each new random result with the result of the previous random input.
Comparing consecutive random results is not a valid oracle; random approaches still need an automated oracle to judge each outcome.
Adaptive random testing improves the distribution of random inputs across the domain, based on the observation that failure-causing inputs cluster together. It does not add a coverage criterion or replace the oracle.
In metamorphic testing, what is the relationship between a source test case and a follow-up test case?
The follow-up test case re-runs the source test case unchanged on a second environment or platform, and the two outputs are compared with each other to detect environment-specific defects, replacing the need for an expected result in either environment.
Running the same input in two environments is a form of back-to-back or portability testing, not a metamorphic relation between transformed inputs.
The follow-up test case uses the output of the source test case as its input, forming a chain of dependent test cases in which just the final output is verified against a golden result, so that intermediate expected values do not have to be specified.
Chaining outputs into inputs is scenario or workflow testing; metamorphic testing transforms the input and checks a relation between outputs.
The follow-up test case is derived from the source test case by a transformation defined in a metamorphic relation, and the two outputs are checked against the relation rather than against fixed expected results.
This is the definition: a metamorphic relation links input changes to a required output relationship, replacing the need for an exact expected value.
The follow-up test case is the negative counterpart of the source test case, built by invalidating one input value, and the pair is verified by checking that exactly one of the two passes and that the failing one produces a specific error message.
Positive and negative pairs are a general test design pattern; a metamorphic relation does not require one of the pair to fail.
Metamorphic testing defines a relation between changes to an input and the expected change in output. The follow-up case is the transformed input, and the pair passes when the outputs satisfy the relation.
Which TWO of the following does the CTAL-TA syllabus list as benefits of automating test design (for example generating test cases from a model)? (Choose two.)
Defects in the requirements are prevented, because the tool rejects any model whose transitions contradict the written specification.
Tools generate from the model as given; detecting contradictions with the specification is a review activity, not an automatic generation benefit.
Systematic coverage of the model is achieved consistently, avoiding the omissions that manual derivation of many similar test cases tends to produce.
Consistent, systematic coverage without human omission is another benefit named in the syllabus.
The generated test cases no longer need separately specified expected results, because the generation tool derives pass or fail verdicts directly from the structure and transitions of the model.
Generated tests still require an oracle; the syllabus lists the continued need for expected results among the risks, not the benefits.
Test cases can be regenerated quickly when the model or the requirements change, reducing the maintenance effort for large test suites.
Fast regeneration after change is one of the listed benefits of test design automation.
Exploratory testing becomes unnecessary once a complete model exists, since every behaviour of interest is already encoded as a generated test and covered by the tool's coverage report.
Models capture specified behaviour; the syllabus keeps experience-based testing as a complement, not something that generation replaces.
Test design automation brings speed of regeneration and systematic, consistent coverage. It still needs expected results, does not replace experience-based testing and cannot itself detect a model that encodes a wrong requirement.
Why does the CTAL-TA syllabus recommend defect taxonomies (such as Beizer's or Kaner's) as a source when building checklists for checklist-based testing?
Because a taxonomy catalogues the kinds of defects that recur across many systems, so each category can be turned into a checklist item that prompts the tester to look for that defect type.
Taxonomies systematise known defect types; converting categories into checklist items gives the checklist a structured, experience-based foundation.
Because a taxonomy replaces the test charter in session-based testing, so exploratory sessions can be run with the taxonomy as their mission statement and debriefed by counting the categories in which defects were found.
Charters remain the mission statement of a session; a taxonomy may inform a charter or a checklist but does not substitute for it.
Because a taxonomy defines the exact number of test cases required per requirement and per defect category, giving the checklist a measurable coverage criterion that can be reported alongside requirement coverage.
Defect taxonomies say nothing about the number of tests per requirement; checklist coverage is measured against checklist items, not requirements.
Because a taxonomy fixes the severity and priority of each defect type in advance, so the checklist can be executed in descending order of severity without further prioritisation and the most severe categories are exercised in every session.
Taxonomies classify defects by kind, not by a predetermined severity; severity still depends on the system under test.
Defect taxonomies collect recurring defect types. Turning each category into a checklist item gives testers a systematic list of things to look for, which is why the syllabus names them as a checklist source alongside defect libraries.
A neonatal incubator monitor raises a high-temperature alarm when the measured…
A neonatal incubator monitor raises a high-temperature alarm when the measured skin temperature is strictly greater than 38.0 °C; at exactly 38.0 °C no alarm is raised. The sensor reports values with a resolution of 0.1 °C. Applying domain testing to this open border, which pair of values are the ON point and the OFF point for the alarm boundary?
ON point 38.0 °C, which lies on the boundary and stays in the no-alarm domain; OFF point 38.1 °C, the nearest value on the other side, which must raise the alarm.
For an open border (strictly greater than) the ON point is the boundary value itself, which does not satisfy the condition, and the OFF point is one resolution step beyond it.
ON point 37.9 °C and OFF point 38.1 °C, the two nearest values either side of the boundary, leaving 38.0 °C untested because it is ambiguous.
The boundary value is exactly the point most likely to expose a strict-versus-inclusive defect and is the ON point; omitting it defeats the purpose of domain testing.
ON point 38.0 °C, on the boundary; OFF point 37.9 °C, one step below, because OFF points are taken on the side where the condition is false.
37.9 lies in the same no-alarm domain as the ON point, so it does not cross the border; the OFF point must be on the opposite side, which is 38.1.
ON point 38.1 °C, the first value that raises the alarm; OFF point 38.0 °C, the last value that does not, because ON points must belong to the domain being tested.
The ON point is defined by lying on the boundary value, not by belonging to the alarm domain; with a strict inequality the boundary value 38.0 is the ON point.
With an open border defined by a strict inequality, the ON point is the boundary value (38.0, no alarm) and the OFF point is the closest representable value on the other side (38.1, alarm). This pair detects a mistaken '>=' implementation.
An electric-vehicle charging tariff applies when charging power is between 3.7…
An electric-vehicle charging tariff applies when charging power is between 3.7 kW and 22.0 kW inclusive AND session duration is between 15 and 480 minutes inclusive; power is set in steps of 0.1 kW, duration in whole minutes. The Test Analyst wants an OFF point for the lower boundary of duration while keeping power well inside its valid range. Which test input is that OFF point?
Power 22.0 kW, duration 481 minutes.
481 minutes is the OFF point for the upper duration boundary, and 22.0 kW is an ON point for power, so this tests the wrong border.
Power 3.6 kW, duration 60 minutes.
This is the OFF point for the lower boundary of power; duration 60 is an IN value, so it does not test the duration boundary.
Power 11.0 kW, duration 15 minutes.
15 minutes is exactly on the closed lower boundary and belongs to the valid domain, so it is the ON point, not the OFF point.
Power 11.0 kW, duration 14 minutes.
14 minutes is one step outside the closed lower duration boundary (15) while 11.0 kW is an IN value for power, so this is the OFF point for the lower duration boundary.
For a closed border (>= 15) the ON point is 15 and the OFF point is the nearest value outside the domain, 14. Keeping power at an IN value isolates the duration boundary under test.
A video-streaming app is configured by three parameters: Device (SmartTV, Phone…
A video-streaming app is configured by three parameters: Device (SmartTV, Phone, Browser), Resolution (SD, HD, 4K) and Login method (password, QR code). Two constraints apply: QR-code login exists on SmartTV only, and 4K playback is not offered on Phone. A pairwise tool produced the following candidate tests. Which TWO tests are infeasible under the constraints and must be replaced? (Choose two.)
SmartTV, 4K, QR code
SmartTV supports both 4K and QR-code login, so this combination is feasible and useful.
Phone, HD, password
HD on Phone with password login violates neither constraint; it is a valid test.
Phone, 4K, password
4K is not offered on Phone, so this combination cannot be executed and violates the second constraint.
Browser, HD, QR code
QR-code login exists on SmartTV only, so a Browser test with QR code violates the first constraint.
Browser, SD, password
SD playback on Browser with password login is unconstrained and feasible.
Constraints exclude specific value pairs from the combination set. Phone with 4K and Browser with QR code each violate one constraint; the other three tests remain feasible.
A public-transport ticket machine has four independent parameters: Ticket type…
A public-transport ticket machine has four independent parameters: Ticket type (adult, child, senior, student), Zone (1, 2, 3), Payment (card, cash) and Validity (single, day). The Test Analyst chooses 'adult, zone 1, card, single' as the base choice. How many test cases does base choice coverage require?
48
48 is the exhaustive count 4 x 3 x 2 x 2, which base choice coverage is designed to avoid.
11
11 is the total number of parameter values (4 + 3 + 2 + 2); base choice does not run one test per value because the base test already covers four of them.
12
12 is roughly the size of a pairwise set for these parameters (the two largest have 4 x 3 = 12 pairs); base choice coverage is a weaker, smaller criterion.
8
Base choice coverage needs the base test plus one test per non-base value: 1 + (3 + 2 + 1 + 1) = 8.
Base choice coverage = 1 base test + the sum over parameters of (number of values - 1): 1 + 3 + 2 + 1 + 1 = 8 tests.
A hospital system imports prescriptions as PDF files from external clinics. The Test Analyst plans fuzz testing: millions of malformed PDFs will be generated automatically and fed to the import service. Which prerequisite must be in place for this random approach to produce usable results?
An automated oracle that judges each run without a predetermined expected value, for example 'the service must not crash or hang and must return a well-formed rejection for any invalid file'.
Random and fuzz testing generate inputs whose exact expected output is unknown, so an automatable property-style oracle is essential.
A manual review step in which a pharmacist or clinical reviewer inspects a stratified sample of the generated files and records the expected import result for each before the run starts, so that the results can be judged against clinical expectations.
Manual expected results for millions of files are infeasible and contradict the purpose of automated random generation.
A recognised coverage criterion for random testing, agreed with the stakeholders before the run, so that the percentage of the PDF specification exercised can be reported afterwards and used as the exit criterion for the fuzzing activity.
The syllabus states that random testing lacks a generally recognised coverage criterion; its value comes from the automated oracle and volume, not coverage figures.
A list of equivalence partitions for every PDF field and structure element, so that the generator draws exactly one representative from each partition per run and the resulting coverage can be reported per partition.
Partition-based selection is equivalence partitioning, a different technique; fuzzing deliberately generates inputs outside any planned partitions.
Because random and fuzz inputs have no precomputed expected result, they require an automated oracle expressed as properties (no crash, no hang, well-formed error). Partitioning, manual review and coverage criteria do not fit this technique.
For a pharmacy system the CRUD matrix for the 'Prescriber' entity shows a…
For a pharmacy system the CRUD matrix for the 'Prescriber' entity shows a function for each of Create, Read, Update and Delete, and a static check confirms the matrix is complete. During dynamic testing the Test Analyst finds that deleting a prescriber leaves open prescriptions pointing to a prescriber that no longer exists, and that updating a prescriber's licence number is not reflected in prescriptions issued earlier that day. Which TWO statements are correct? (Choose two.)
Static completeness checking of the matrix would have detected both defects if the matrix had been reviewed more carefully together with the data model, because the relationships between Prescriber and Prescription are visible there.
Static checking can only show whether functions exist; it cannot reveal wrong run-time effects on related data.
The matrix was in fact incomplete, because Delete should not have been listed for an entity that other entities reference, and the static completeness check should have flagged the missing 'archive' function that would replace it.
Completeness concerns whether each required function exists; whether Delete should cascade or be refused is a behavioural consistency question.
Both findings are consistency defects: the CRUD operations exist, but their effect on related data is wrong, which is exactly what dynamic CRUD testing is meant to reveal.
CRUD consistency testing executes the operations and checks their effects across related entities; these are consistency, not completeness, failures.
The findings show that CRUD testing is unsuitable for entities with relationships to other entities and should be replaced by state transition testing for this entity, with prescriber deletion modelled as a terminal state.
CRUD testing is designed to cover relationships between functions and entities; the findings demonstrate its value rather than its unsuitability.
Negative tests, such as attempting to delete a prescriber who is still referenced by prescriptions, are a legitimate part of CRUD consistency testing.
The syllabus includes negative tests in consistency testing to check how the system handles operations that should be refused or cascaded.
A complete CRUD matrix (static completeness) does not mean the operations behave correctly; dynamic consistency testing, including negative tests, is what exposes orphaned references and stale data.
A car-park exit gate is modelled as an extended finite state machine with states…
A car-park exit gate is modelled as an extended finite state machine with states Closed and Open. Transitions: Closed -> Open on 'ticket scanned' with guard [ticket paid]; Closed -> Closed on 'ticket scanned' with guard [ticket not paid], displaying 'payment required'; Open -> Closed on 'vehicle passed'. Which test set achieves 0-switch coverage of this model?
Scan an unpaid ticket at a closed gate; scan a paid ticket at a closed gate; drive a vehicle through the open gate; drive a vehicle through the closed gate.
The last test is an invalid event for the Closed state; it exceeds 0-switch coverage of valid transitions while adding a case the model does not define.
Scan an unpaid ticket at a closed gate; scan a paid ticket at a closed gate; drive a vehicle through the open gate.
0-switch coverage requires every valid transition to be exercised once; with a guard, each guard outcome is a separate transition, giving three transitions.
Scan a paid ticket at a closed gate; drive a vehicle through the open gate; repeat the paid scan and the vehicle passage once more to confirm the cycle is stable.
This omits the guarded self-transition for an unpaid ticket, so one of the three valid transitions is never exercised.
Scan a paid ticket at a closed gate; drive a vehicle through the open gate; scan a ticket while the gate is open.
Scanning while Open is not a modelled transition; it may be a useful negative test but does not contribute to covering valid transitions, and the unpaid case is missing.
In an extended finite state machine the guard splits one event into distinct transitions. 0-switch coverage needs all three valid transitions: unpaid scan (stay Closed), paid scan (Open) and vehicle passed (Closed).
An EV charging session has states Idle, Charging, Paused and Complete, with six…
An EV charging session has states Idle, Charging, Paused and Complete, with six transitions: Idle -> Charging (plug in), Charging -> Paused (pause), Paused -> Charging (resume), Charging -> Complete (battery full), Paused -> Idle (unplug) and Complete -> Idle (unplug). How many distinct two-transition sequences must be exercised to achieve 1-switch coverage?
12
12 would assume every transition has exactly two valid successors; three of the six transitions have only one.
36
36 is 6 x 6, the count of all transition pairs regardless of whether the second can follow the first; most of those pairs are not valid sequences.
6
6 is the number of single transitions, which is 0-switch coverage; 1-switch requires pairs of consecutive transitions.
9
Counting valid successors: plug in (2), pause (2), resume (2), battery full (1), unplug from Paused (1), unplug from Complete (1) gives 9 sequences.
1-switch coverage counts valid sequences of two consecutive transitions: Idle->Charging is followed by 2 options, Charging->Paused by 2, Paused->Charging by 2, Charging->Complete by 1, Paused->Idle by 1, Complete->Idle by 1, giving 9.
A monthly payroll run is modelled in BPMN: after 'collect timesheets', a parallel gateway forks into 'calculate gross pay' and 'calculate deductions', a joining gateway waits for both, and then 'generate payslips' runs. Which test most specifically targets the behaviour of the joining gateway?
A test that stops the run after 'collect timesheets' and confirms that neither calculation has begun, verifying that the parallel gateway is reached solely with complete timesheet data and that the two branches start at the same moment.
This tests the entry to the fork, not the join; it says nothing about whether payslips wait for both calculations.
A test in which one parallel branch is deliberately delayed so that it finishes well after the other, verifying that payslip generation does not start until both branches have completed.
The join's defining behaviour is waiting for all incoming branches; making one branch late is the way to observe whether the join is honoured.
A test that chooses the 'deductions' path at the gateway and verifies that gross pay is skipped when there are no timesheet changes for the month.
A parallel gateway does not choose between paths; treating it as a decision node misreads the notation and tests behaviour the model does not contain.
A test that runs 'calculate gross pay' and 'calculate deductions' one after the other in a single thread and checks that the payslip totals are arithmetically correct for every employee category, including those with no deductions in the month.
Sequential execution never exercises the concurrency that the fork and join introduce, so it cannot reveal a join that fires early.
In BPMN a joining parallel gateway must wait for every incoming branch. The characteristic test introduces a timing difference between branches and checks that the following activity does not start before the slower branch ends.
An airline online check-in use case has a main flow (enter booking reference…
An airline online check-in use case has a main flow (enter booking reference, choose seat, receive boarding pass). In one alternative path the passenger adds a checked bag and pays a fee before the boarding pass is issued; in another the passport validity check fails and check-in is refused. How should a Test Analyst classify these two paths when deriving use case tests?
Adding a bag is an extension flow that rejoins the main flow and still ends with a boarding pass; the failed passport check is an exception flow that ends the use case without achieving its goal.
Extension flows add optional behaviour and return to the main path; exception flows handle errors and terminate the use case unsuccessfully.
Adding a bag is an exception flow because it costs the passenger money and involves a payment step outside check-in; the failed passport check is an extension flow because it extends the checks performed before the boarding pass would be issued.
Cost is not a classification criterion; a path that ends the use case without its goal is an exception, and a path that returns to the main flow is an extension.
Both are exception flows, because each one departs from the main flow at a decision point, and every departure from the main flow is treated as an exception in use case testing so that each receives its own test case.
Use case testing distinguishes extensions (optional, successful) from exceptions (error handling, unsuccessful); lumping them together loses that distinction.
Both are extension flows, because in each case the passenger's action or document, rather than a system fault, causes the deviation from the main flow, and extension flows are the category used for everything the passenger can trigger during check-in.
Who triggers the deviation does not decide the type; the refused check-in fails to reach the goal and is therefore an exception flow.
Use case tests cover the main flow, each extension flow (optional behaviour that rejoins the main path) and each exception flow (error handling that ends the use case without its goal). Adding a bag is an extension; a refused check-in is an exception.
A student-discount rule for a museum ticketing system has three conditions…
A student-discount rule for a museum ticketing system has three conditions: 'holds a registered student account' (Y/N), 'student ID has been verified' (Y/N) and 'age under 26' (Y/N). Verification is possible only for a registered student account. Starting from the full decision table, how many rules remain feasible?
4
4 would result from removing every rule involving verification; only the contradictory 'not registered, verified' combinations are infeasible.
6
Of the 8 combinations, the two with 'registered = N' and 'verified = Y' (age Y or N) are impossible, leaving 6 feasible rules.
8
8 is the full table before infeasible rules are removed; two of those combinations contradict the dependency between registration and verification.
7
The infeasible pattern 'not registered but verified' occurs for both values of the age condition, so two rules are removed, not one.
A full table with three binary conditions has 8 rules. The dependency 'verified implies registered' makes the two rules with registered = N and verified = Y infeasible, leaving 6.
A library late-fee decision table has conditions A 'item is reference-only', B…
A library late-fee decision table has conditions A 'item is reference-only', B 'borrower is staff' and C 'returned within the grace period'. The minimised table has three rules: Rule 1 (A = -, B = -, C = Y) -> no fee; Rule 2 (A = Y, B = N, C = N) -> double fee; Rule 3 (A = N, B = -, C = N) -> standard fee. Applying the checksum procedure, which condition combination is missing from the table?
A = N, B = Y, C = N (ordinary item, staff borrower, returned late)
This combination is covered by Rule 3, whose don't-care for B includes staff borrowers.
A = Y, B = N, C = Y (reference-only item, non-staff borrower, returned in time)
Rule 1 has don't-cares for A and B, so any return within the grace period, including this one, is covered.
A = Y, B = Y, C = N (reference-only item, staff borrower, returned late)
Rule 1 covers 4 combinations, Rule 2 covers 1 and Rule 3 covers 2, a checksum of 7 against 8 possible; the uncovered combination is Y/Y/N.
No combination is missing; the checksum shows an overlap between Rule 2 and Rule 3 instead.
Rules 2 and 3 differ on condition A (Y versus N) so they cannot overlap; the checksum of 7 out of 8 indicates a gap, not an overlap.
The checksum procedure counts combinations per rule (each don't-care doubles the count): 4 + 1 + 2 = 7 of 8. The single uncovered combination is a reference-only item returned late by a staff borrower.
A pharmacy inventory search returns every product that matches ALL entered keywords. There is no reference implementation to compare against. Which TWO of the following are valid metamorphic relations the Test Analyst can use? (Choose two.)
Removing one keyword from a query must return a result set that is strictly smaller than the original.
Removing an AND-condition can only add matches, so the follow-up set must be a superset (equal or larger), not smaller.
Entering the same keywords in a different order must return exactly the same set of products.
For a conjunctive match the order of keywords is irrelevant, so permuting them is a valid relation with an identical expected result set.
Misspelling one keyword must return at least as many products as the correctly spelled query, because fuzzy matching widens the search.
A misspelled keyword typically matches fewer or different products; no reliable relation exists in this direction.
Doubling the page size of the result list must double the total number of matching products reported in the result-count header.
Page size affects how results are displayed, not how many products match; the total should be unchanged.
Adding a further keyword to a query must return a result set that is a subset of the original result set (equal or smaller).
Adding an AND-condition can only remove matches, so the follow-up result set must be contained in the source result set.
Valid metamorphic relations follow from the semantics of an AND search: adding a keyword narrows (subset), permuting keywords leaves the set unchanged. Page size, misspellings and the reversed direction for removal do not yield dependable relations.
Reviewing six months of production defects for a tax-filing application, a Test…
Reviewing six months of production defects for a tax-filing application, a Test Analyst classifies them with Beizer's defect taxonomy and finds that four of the six most costly defects were boundary-handling errors on numeric and date fields. How should this finding be used to improve checklist-based testing of new screens?
Retire the checklist for these screens and replace it with unscripted exploratory sessions of 90 minutes per screen, because the defects escaped despite a checklist being in place and experienced testers are better at finding boundary problems than lists are.
Dropping the checklist discards the systematic prompt; the finding calls for sharpening the checklist, not for abandoning it.
Use the taxonomy to re-prioritise the open defect backlog by category so that boundary defects are fixed first and reported to management as a separate trend line, leaving the checklist unchanged until the next annual review of test assets.
Backlog ordering is a defect-management action; it does not change how new screens are tested and so does not prevent recurrence.
Add checklist items that, for each numeric and date input on a screen, require testing the exact boundary value, the nearest values on either side and the empty value, and apply them during every checklist session.
A taxonomy-driven checklist converts the dominant defect category into concrete, repeatable checks that testers apply to each new screen.
Add a single checklist item reading 'check boundaries on all inputs' so that the list stays short and easy to maintain, and rely on each tester's judgement and experience to decide what that means for a given screen and which values to try.
An item without concrete guidance leaves the interpretation to chance; the value of a taxonomy-based checklist lies in specific prompts.
Defect taxonomies inform checklists by turning frequent defect categories into explicit checklist items. Concrete boundary checks per input field make the recurring defect type visible in every future checklist session.
For a video-streaming client, defect analysis shows that most escaped defects…
For a video-streaming client, defect analysis shows that most escaped defects appear when particular combinations of user settings interact, for example a specific subtitle language together with a particular audio track and a given playback quality, while each setting behaves correctly on its own. Which technique class should the Test Analyst prioritise for the next release?
State transition testing, modelling each setting change as an event and covering every transition of the player's state machine with 1-switch coverage.
The defects depend on which values are combined, not on the order of state changes, so a transition model does not address the observed failure pattern.
Decision table testing, listing business rules for when subtitles and audio tracks may be combined with each quality level and covering each rule once.
There are no business rules restricting combinations here; the settings are free choices whose interactions, not rules, cause the failures.
Combinatorial testing, using pairwise coverage across the settings so that every pair of setting values is exercised together at least once.
Defects caused by interactions between parameter values are the target of combinatorial techniques; pairwise coverage addresses them economically.
Use case testing, covering the main flow and each extension flow of the 'watch a video' use case with realistic personas.
Use case testing covers workflows; it does not systematically vary setting combinations and would leave most interactions untested.
When failures arise from the interaction of independent parameter values, the syllabus maps them to combinatorial techniques such as pairwise testing. The other techniques target sequencing, rules and workflows.
A team generated its regression tests for a benefits-eligibility engine from a…
A team generated its regression tests for a benefits-eligibility engine from a state model using a model-based testing tool. All generated tests passed, yet a production defect appeared: applicants with a temporary work permit were wrongly rejected. Investigation shows that the model treated 'temporary permit' as 'no permit', which is also how the requirements document described it, although the legislation says otherwise. Which risk of automated test design does this case illustrate?
The coverage report was misread: the model was covered completely according to the tool, but the generated tests were executed partially in the pipeline because of a time-out, and the missing tests happened to include the temporary-permit path, so the pipeline configuration is the root cause.
The stem says every generated test passed after execution; partial execution is not the cause described.
Model-based tools cannot represent conditional transitions such as permit types with guard conditions, so eligibility rules of this kind should be tested with hand-written decision tables instead, and the model should be limited to the navigation flow of the application where it adds value.
Modern models support guards and conditions; the problem was the content of the model, not a limitation of the notation.
Generated tests are only as correct as the model; a model that reproduces a misunderstanding in the specification yields tests that confirm the wrong behaviour, so the model must be validated against stakeholder intent and not just the written specification.
The tool did its job; the defect lies in the model's fidelity to the real requirement, which is a listed risk of test design automation.
The tool generated too many tests for the pipeline to run in the available window, so the suite was pruned by a script that removed tests with similar names, and the test for temporary permits was among those removed without anyone reviewing the list.
No test for the correct temporary-permit behaviour could have existed, because the model itself encoded the wrong rule.
A recognised risk of test design automation is that the generated tests inherit every error in the model. When the model mirrors a wrong specification, all tests pass and the defect escapes, so models need validation against the true requirement, not only the document.
An insurance premium is set by a decision table with three conditions: 'driver…
An insurance premium is set by a decision table with three conditions: 'driver age under 25', 'vehicle value above 40,000' and 'one or more claims in the last three years'. A Test Analyst's test set contains eight tests, one per rule, using ages 20 and 35, vehicle values 25,000 and 60,000, and claim counts 0 and 2. All eight tests pass. Evaluate the adequacy of this test set for the feature.
The set should be replaced by pairwise testing over the three conditions, which covers all value pairs in four tests and is therefore more efficient than eight rule-based tests, leaving time to add a fourth condition for the driver's licence age that the business has been requesting.
Pairwise coverage would not guarantee that each business rule is exercised, trading away the rule coverage the feature needs without adding threshold tests.
The set is excessive: the table should be minimised using don't-care entries first, which would reduce the eight tests to about four while keeping every distinct premium outcome, and the freed effort should be spent on exploratory testing of the quotation screens, where usability defects are more likely.
Minimisation reduces rules only where an outcome does not depend on a condition, and it does nothing to address the untested thresholds, which is the real gap.
The set achieves full rule coverage but leaves the condition thresholds untested; it should be augmented with domain testing on ages 24 and 25, values 40,000 and 40,001 and claim counts 0 and 1 to detect off-by-one errors in the condition evaluation.
Decision table coverage checks the combination logic; the thresholds embedded in the conditions need boundary or domain testing, which the chosen mid-range values do not provide.
The set is adequate: every rule of the full table is exercised once with representative values, which is the coverage criterion the syllabus defines for decision table testing, and the passing results demonstrate that the combination logic and the premium amounts are correct, so no further tests are needed.
Rule coverage is necessary but the conditions contain numeric thresholds; the syllabus recommends combining techniques so that boundary defects in the conditions are also caught.
The test set covers the combination logic but every condition hides a threshold (25, 40,000, at least one claim) that mid-range values cannot check. The appropriate improvement is to combine decision table testing with domain or boundary value testing on each condition.
A smart-meter billing feature has had twelve escaped defects in a year. Root…
A smart-meter billing feature has had twelve escaped defects in a year. Root cause records show nine occurred on tariff-switch dates (1 April and 1 October season changes, and the daylight-saving change), where readings just before or after midnight were billed at the wrong rate; the other three appeared only for specific combinations of tariff plan, meter type and region. Which combination of techniques best addresses this defect profile for the next release?
Domain testing on the date and time boundaries of each tariff switch (readings at, just before and just after the switch instant) combined with pairwise testing over tariff plan, meter type and region.
Nine defects are boundary errors on time, which domain testing targets directly; the remaining three are interaction defects, which pairwise coverage addresses economically.
Exhaustive testing of every combination of tariff plan, meter type, region and calendar date across a full year, so that both defect groups are covered by a single technique and the coverage figure can be reported as 100% to the regulator.
Exhaustive combination with dates is unmanageably large and treats the dominant boundary problem as a combination problem rather than testing the switch instants precisely.
Scenario-based testing of one realistic year of consumption for a typical household, covering all four season changes and the daylight-saving switch in a single end-to-end scenario that is replayed against each release candidate.
A single typical scenario uses one tariff, meter type and region and passes through each switch once at an arbitrary time, so it neither isolates the boundary instants nor covers combinations.
State transition testing of the meter lifecycle (installed, active, replaced, decommissioned) with round-trip coverage, because tariff changes are state changes of the meter and each season switch can be modelled as an event that moves the meter to a new billing state.
The defects concern billing at time boundaries and parameter interactions, not the meter's lifecycle states; a lifecycle model would not exercise the switch instants.
The defect history splits into two classes: boundary errors at tariff-switch instants (domain testing on date/time borders) and interaction errors across configuration parameters (combinatorial testing). Selecting techniques by the observed defect types is the syllabus approach to choosing the most appropriate technique.
Chapter 4 · Testing Quality Characteristics — 4 questions
A payroll module implements every function listed in the requirements, and each…
A payroll module implements every function listed in the requirements, and each function is the right one for the users' tasks, but the overtime calculation rounds to the nearest hour instead of the nearest minute, producing wrong amounts. Which ISO/IEC 25010 sub-characteristic of functional suitability is deficient?
Functional appropriateness, because rounding to the hour does not facilitate the user's task of paying accurate overtime and a different function should have been chosen.
Appropriateness concerns whether the functions chosen suit the users' tasks; here the function is the right one but its result is wrong.
Functional completeness, because a rounding-to-the-minute function is missing from the set of implemented functions and the requirement is therefore not fully covered.
Completeness asks whether all specified functions are present; the overtime function is present, it simply computes incorrectly.
Functional correctness, because a required function exists and is appropriate but produces results that deviate from the specified precision.
Functional correctness concerns whether functions provide the right results with the needed degree of precision; wrong rounding is a correctness failure.
Interaction capability, because the user cannot see the rounding rule on screen and therefore cannot recognise the wrong amount.
Interaction capability is a usability characteristic; the defect lies in the computed value, not in how the user interacts with the system.
ISO/IEC 25010 distinguishes completeness (all functions present), correctness (functions give correct results with the required precision) and appropriateness (functions suit the tasks). A wrong rounding result is a correctness defect.
Which TWO statements about accessibility testing are consistent with the CTAL-TA syllabus? (Choose two.)
WCAG conformance is a requirement for public-sector websites alone; commercial products need accessibility testing when users with disabilities request it or when a contract names a specific conformance level.
Accessibility obligations arise from various laws and contracts and apply broadly; waiting for user requests contradicts the preventive approach the syllabus takes.
The Web Content Accessibility Guidelines define three conformance levels, A, AA and AAA, and testing should be planned against the level the product is required to meet.
WCAG conformance levels A, AA and AAA are the reference framework the syllabus names for accessibility testing.
Accessibility testing is complete when an automated scanner reports no violations against the chosen WCAG level, because the success criteria are written to be machine-checkable and manual review adds cost without adding coverage.
Many WCAG criteria require human judgement (for example meaningful alternative text); automated scanners cover only part of the criteria.
Legislation such as the UK Equality Act 2010 and the US Americans with Disabilities Act can make accessibility a legal requirement, so the Test Analyst should treat it as part of usability testing scope.
The syllabus lists these acts as examples of legal drivers and places accessibility within the Test Analyst's usability responsibilities.
Accessibility is a sub-characteristic of compatibility in ISO/IEC 25010, so it is tested by the Technical Test Analyst together with interoperability and coexistence, and the Test Analyst contributes personas for the assistive-technology users involved.
Accessibility belongs to the usability characteristic and is the Test Analyst's responsibility, not a compatibility sub-characteristic.
The syllabus frames accessibility testing around WCAG levels A/AA/AAA and legal drivers such as the UK Equality Act and the US ADA, as part of the Test Analyst's usability scope. It is not a compatibility topic and is not fully automatable.
Before any end users are involved, two experienced interaction designers step through a new pension-calculator screen against a set of established usability principles and record every deviation. Which usability evaluation technique described in the syllabus is this?
An interoperability review, in which specialists check that the calculator exchanges pension data correctly with the employer's systems.
Data exchange between systems is a compatibility topic; the activity described evaluates the user interface, not integration.
A usability questionnaire such as SUMI, which measures perceived satisfaction after the users have completed their tasks.
Questionnaires collect users' subjective ratings after use; no user has interacted with the screen in the described situation.
A usability test session, in which representative users perform realistic tasks based on personas while their behaviour is observed.
Test sessions require representative users performing tasks; two designers inspecting the screen are not users under observation.
A usability review, in which specialists inspect the interface against heuristics or guidelines without involving representative users.
Expert inspection against usability principles is the review technique; it is performed by specialists rather than by users.
The syllabus describes three usability evaluation techniques: usability reviews (expert inspection against principles), usability test sessions (observed users) and questionnaires such as SUMI or WAMMI. An expert walk-through is a usability review.
A school-management application must be installed on premises by each school's…
A school-management application must be installed on premises by each school's own IT staff following the vendor's guide, and must also be able to move from one supported database product to another with minimal changes. Which pairing correctly assigns these two concerns to flexibility sub-characteristics and to the responsible role?
Both concerns are portability topics handled by the Technical Test Analyst, since they involve infrastructure and configuration rather than business functions that the school staff would recognise.
Installability and adaptability are explicitly placed with the Test Analyst; only scalability and replaceability sit outside the role.
On-premises installation is installability, a Test Analyst topic; moving between database products is replaceability, a Technical Test Analyst topic, because it concerns swapping one infrastructure component for another.
Replaceability concerns replacing the product itself with another product for the same purpose; adapting to a different database is adaptability.
On-premises installation is coexistence, and moving between databases is interoperability; both belong to compatibility rather than flexibility.
Coexistence and interoperability describe how the product works alongside or exchanges data with other software; neither describes installing or adapting the product.
On-premises installation is installability, and moving between database products is adaptability; both are Test Analyst responsibilities within flexibility.
The syllabus assigns installability and adaptability to the Test Analyst, while scalability is covered by CT-PT and replaceability by the Technical Test Analyst.
Within flexibility (formerly portability), the Test Analyst covers installability and adaptability. Scalability is a performance topic and replaceability is a Technical Test Analyst topic; coexistence and interoperability belong to compatibility.
Chapter 5 · Software Defect Prevention — 8 questions
The CTAL-TA syllabus groups defect prevention practices into three categories. A team introduces a mandatory root cause analysis for every escaped defect, followed by a change to the coding guideline or the review checklist. Into which category does this practice fall?
Preventing the introduction of defects, because the changed coding guideline stops developers from writing the defect in the first place, and the syllabus classifies any practice that changes how code is written as introduction prevention.
Although the outcome influences future coding, the practice starts from defects that already escaped; the syllabus classifies analysis-driven changes as recurrence prevention.
Preventing defects through modelling, because the analysis produces a cause-effect model that is later reused for model-based testing.
Modelling for defect prevention refers to building system or behaviour models before implementation; a cause-effect diagram is an analysis aid, not a system model.
Preventing the recurrence of defects, because the practice analyses defects that have already occurred and changes the process so the same cause does not produce them again.
Root cause analysis followed by process change is the core of the third category, mitigating recurrence.
Preventing defects from leaking to later phases, because analysing escaped defects improves phase containment effectiveness by showing which phase should have caught each defect and strengthening that phase's checks.
Phase containment practices catch defects within the phase that created them; root cause analysis of escapes addresses why they happened, not where they were caught.
The three groups are preventing defect introduction, preventing leakage to later phases (phase containment) and preventing recurrence. Root cause analysis of escaped defects with subsequent process change belongs to the third group.
Which TWO statements about analysing test results for defect prevention are consistent with the CTAL-TA syllabus? (Choose two.)
Structural coverage analysis is a Test Analyst technique for measuring how many requirements have at least one associated test case, and it is reported alongside defect density per requirement.
Structural coverage measures code elements executed; requirement coverage is a different, black-box measure.
A failing test does not by itself prove a product defect; the failure may be caused by the test, the data or the environment and must be analysed first.
The syllabus explicitly warns that a failed test is not automatically a defect and requires anomaly analysis.
Test gap analysis compares recently changed code with the code exercised by tests to identify changed areas that have not been tested.
The syllabus describes test gap analysis as combining change information with test coverage to reveal untested changes.
Defect detection percentage is calculated as the number of defects found by testing divided by the number of test cases executed, so that a higher DDP indicates more effective test cases.
DDP relates defects found by testing to the total defects found (including those found later, in production), not to the number of tests run.
The defect arrival pattern should be flat throughout the test phase in a healthy project, and any peak indicates a defective build.
The Rayleigh model expects arrivals to rise, peak and then fall; a peak is normal, not a sign of a defective build.
Test gap analysis exposes changed but untested code, and every failed test must be analysed before it is logged as a defect. DDP is defined against total defects, structural coverage measures code, and the Rayleigh pattern includes a peak.
For a review of the requirements of a ticket-refund feature, the Test Analyst…
For a review of the requirements of a ticket-refund feature, the Test Analyst gives each reviewer a concrete procedure to follow: one reviewer writes draft test cases from the requirements, another writes a user manual section, a third traces every requirement to a business goal. Each reviewer reports the problems that surfaced while performing their procedure. Which review technique is being applied?
Scenario-based reviewing, in which reviewers walk through expected usage scenarios written in advance by the author of the requirements.
Scenario-based reviewing follows given usage scenarios; it does not ask reviewers to draft tests or manuals as a way of finding defects.
Perspective-based reading, in which reviewers actively produce a work product from a stakeholder viewpoint and defects are found through that activity.
Perspective-based reading assigns each reviewer a perspective with a concrete procedure that yields a product (tests, manual, traceability), which is exactly the setup described.
Checklist-based reviewing, in which each reviewer answers a fixed list of questions about completeness, consistency and testability.
Checklist-based reviewing works through predefined questions; no checklist is described and reviewers here are constructing artefacts instead.
Role-based reviewing, in which reviewers read the document from the standpoint of an end user, an operator or a maintainer without producing any artefact.
Role-based reviewing assigns viewpoints but does not require reviewers to create a work product; the procedural, productive element here points to perspective-based reading.
Perspective-based reading is the most rigorous of the five review techniques: reviewers adopt a stakeholder perspective and follow a procedure that produces something (test cases, a manual, traceability), surfacing defects in the process.
Over a release of a hospital scheduling system, 171 defects were found and removed before go-live and 19 further defects were reported by users during the first three months in production. What is the defect removal efficiency (DRE) for this release, and how should the Test Analyst read it?
90%: 171 of 190 total defects were removed before release; the remaining 10% escaped and should be analysed by phase and root cause to find where containment failed.
DRE = defects removed before release / (defects removed before release + defects found after release) = 171 / 190 = 90%.
90%: 171 of 190 total defects were removed before release, which proves that the 19 production defects were introduced by post-release configuration changes rather than missed in testing.
The percentage is right but the conclusion is not supported; DRE says how many defects escaped, not why they were missed.
89%: 171 of 190 total defects were removed before release, and the figure shows testing was adequate because it exceeds the 85% industry threshold.
The arithmetic is 171/190 = 0.90, not 0.89, and the syllabus does not define a fixed industry threshold that declares testing adequate.
11%: 19 escaped defects divided by 171 removed defects, which is the proportion of effort that should be moved from testing to production monitoring.
19/171 is neither DRE nor its complement; DRE uses the total of all defects as the denominator, and it does not prescribe effort reallocation.
DRE = 171 / (171 + 19) = 90%. The metric quantifies escaped defects; understanding why the 10% escaped requires phase containment and root cause analysis, not a fixed threshold.
A Pareto analysis of 240 defects in an energy-billing system gives these root…
A Pareto analysis of 240 defects in an energy-billing system gives these root cause categories: 'unclear tariff requirements' 96 defects (40%), 'incorrect date and time handling' 72 (30%), 'third-party API changes' 36 (15%), 'UI layout' 24 (10%), 'other' 12 (5%). Which TWO actions follow correctly from the Pareto principle? (Choose two.)
For the two dominant categories, run root cause analysis (for example five whys or Ishikawa diagrams) to find the process causes behind them before proposing corrective measures.
Pareto shows where to look; the syllabus pairs it with root cause techniques to determine what to change.
Focus prevention effort first on 'unclear tariff requirements' and 'incorrect date and time handling', since these two categories together account for 70% of all defects.
The Pareto approach concentrates effort on the few categories that produce most defects; the top two cover 70% here.
Begin with 'UI layout' defects because they are the easiest to fix and produce quick visible wins, and postpone the requirements category until the tariff team is less busy and can attend the root cause workshops.
Ease of fixing is not a Pareto criterion; starting with a 10% category while ignoring a 40% category wastes the analysis.
Treat 'third-party API changes' as the priority, because external causes are outside the team's control and therefore carry the highest risk of recurrence.
At 15% this category is not among the vital few; external origin does not raise its weight in a Pareto analysis.
Distribute prevention effort evenly across all five categories so that no cause is neglected, the distribution becomes flat over time and the team avoids the perception that some stakeholders' defects matter less than others.
Even distribution contradicts the Pareto principle, which directs effort to the vital few categories with the largest share.
Pareto analysis directs defect prevention toward the categories that produce most defects (here 40% + 30%), and root cause analysis of those categories identifies what to change. Spreading effort evenly or starting with easy or external categories misapplies the principle.
In the last regression cycle of an online examination platform, 88 of 620 tests…
In the last regression cycle of an online examination platform, 88 of 620 tests failed. The Test Analyst's anomaly analysis shows that 31 failures came from an expired test certificate in the environment, 12 from test data that another team had modified, and 45 from genuine product defects. What is the appropriate conclusion and action?
The 43 failures caused by environment and data should be reclassified as passed, because the product code did not cause them and the tests would have passed otherwise.
A test cannot be marked passed without executing successfully; the tests must actually be re-run after the environment and data are corrected.
Only the 45 product failures should be logged as defects; the 43 false positives should be resolved by fixing the environment and test data, and the affected tests re-run before reporting the cycle result.
The syllabus stresses that a failed test is not automatically a defect; false positives must be removed from the defect count and the tests re-executed.
All 88 failures should be logged as defects so that the defect density metric stays comparable with previous cycles and the trend chart is not distorted, with the environment and data issues closed later by the developers as 'not a defect'.
Logging known false positives inflates defect metrics and creates noise for developers; anomaly analysis exists to prevent that.
The cycle should be reported as 45 product defects plus 43 environment defects, since an expired certificate and changed data are defects in the test environment component and belong in the same tracker with a different component label.
Environment and data problems are recorded as test incidents or infrastructure issues, not as product defects in the cycle's defect count.
Anomaly analysis separates product defects from false positives caused by the test environment or test data. False positives are not defects and are not passes either: fix the cause, re-run the tests, and report only confirmed defects.
Across four releases of a logistics platform, seven escaped defects involved…
Across four releases of a logistics platform, seven escaped defects involved wrong delivery dates around month ends and leap days. Each was fixed locally in the component where it surfaced. An Ishikawa analysis of all seven shows two shared causes: three components contain their own copies of date arithmetic, and none of the regression suites include tests at month-end or leap-day boundaries. The team proposes four actions. Which one addresses the recurrence at its root?
Add an exploratory testing session focused on dates before every release, staffed by the two most experienced testers and using a charter that names month ends and leap days, relying on tester experience to find the next month-end problem before customers do.
Exploratory sessions may catch some defects but leave the duplicated logic in place and make prevention depend on individual attention rather than a repeatable test set.
Raise the severity of all date-related defects to 'critical' and add a dedicated 'date handling' label in the defect tracker, so that they are fixed within one day of being reported, escalated to the release manager and cannot reach the next release without a sign-off.
Faster fixing shortens the life of each defect but does nothing about why new date defects keep being introduced and missed.
Assign each of the three components to a different senior developer who is instructed to review the date code carefully during the next sprint, document the edge cases found and report back at the sprint review.
Ad hoc review by three people preserves the three divergent implementations and adds no boundary tests, so the shared causes remain.
Replace the three copies with one shared, reviewed date-handling library and add a boundary test set for month ends, leap days and year ends that every consuming component's regression suite must include.
This removes the structural cause (duplicated logic) and the process cause (missing boundary tests) identified by the analysis, so the same class of defect cannot recur unnoticed.
Root cause analysis pointed to duplicated date logic and absent boundary tests. Only the action that consolidates the logic and institutionalises the boundary tests removes both causes; the others treat symptoms or rely on individual diligence.
A Test Analyst classifies a quarter's 150 defects for a telehealth application…
A Test Analyst classifies a quarter's 150 defects for a telehealth application using the FURPS scheme: Functionality 48, Usability 21, Reliability 57, Performance 15, Supportability 9. Over the previous three quarters, Reliability defects have grown from 12% to 38% of the total, while Functionality has stayed near 30%. Root cause analysis of the Reliability defects shows most stem from unhandled failures of a third-party video service that the application calls. Which recommendation is best supported by this analysis?
Reclassify the video-service failures as Supportability defects, because they originate outside the application and are handled by the vendor's support contract, so that the Reliability trend returns to its earlier level and the reporting to management reflects internal quality alone.
Changing the classification hides the trend without removing its cause; the application's handling of the failures is still its own reliability behaviour.
Replace FURPS with Orthogonal Defect Classification for the next quarter, because FURPS cannot distinguish external from internal causes of reliability defects and ODC's trigger and type attributes would let the team separate vendor failures from application logic before deciding on measures.
Switching schemes may add information, but the analysis has already identified the cause; the priority is a prevention measure, not a new taxonomy.
Reduce functional testing effort by a third, since Functionality defects have not increased over three quarters and the functional regression suite is stable, and use the freed time for a general regression pass across all categories and an extra round of usability sessions.
A stable share of Functionality defects does not mean functional testing is over-resourced, and a general regression pass does not address the video-service failure modes.
Introduce a specific prevention measure for the growing category: define required behaviour for every failure mode of the video service in the requirements, review those requirements, and add negative tests using a stub that simulates each failure mode.
The classification isolates a growing category and root cause analysis locates its origin; the recommendation targets that origin with requirements, review and negative testing.
Defect classification reveals the trend (Reliability rising to 38%) and root cause analysis explains it (unhandled external failures). The prevention measure that follows is to specify, review and negatively test every failure mode of the external service; reallocating effort or renaming the category does not prevent recurrence.