Free course: Functional Safety Basics on Udemy. The first 100 learners enrol free, until 7 December. Enrol now →

Learn

What Is Verification and Validation (V&V)?

Published 13 November 2026

Verification and validation (V&V) are two complementary activities that together check, through objective evidence, whether a product, service or system meets its requirements and fulfils its intended purpose. Verification is confirmation that specified requirements have been fulfilled. Validation is confirmation that the requirements for a specific intended use or application have been fulfilled. These are the definitions standardised in ISO/IEC/IEEE 15288, the international standard for system life cycle processes, and closely related definitions are used in quality management.

Both definitions rest on objective evidence: each activity ends in a record of results showing whether the requirement in question has been met.

The pair is commonly summarised as two questions: verification asks whether the product was built right, and validation asks whether the right product was built. The formulation is usually attributed to Barry Boehm, who used it in Software Engineering Economics (1981), and it has become standard shorthand in systems engineering. The terms are used most often in software, but they apply equally to hardware and to whole systems. In practice the two terms are sometimes used interchangeably, and their definitions vary between organisations. The distinction is nonetheless deliberate, because the two activities check a product against different references.

This guide covers what each activity is checked against, where the practical dividing line falls, the methods used, how V&V runs through the life cycle, the evidence and decisions it produces, related terms, its use across industries and its limits. It closes with a case study: the London Millennium Footbridge.

What each activity is checked against

Verification: against the specification

Verification judges a product against its specification. Each requirement, typically written as a "shall" statement, is shown to be met. Verification is traced to the approved set of requirements and to the configuration baseline: the approved specifications, drawings and parts lists. It also confirms that no error was introduced when inputs were transformed into outputs, and that the chosen verification approach would have detected such an error if one had been introduced.

Validation: against the stakeholders' needs

Validation judges a product against the needs it was meant to meet: the expectations of the customer and other stakeholders, and the concept of operations, which describes how the system will be used. It shows that the product accomplishes its intended purpose in its intended environment when operated by its intended users. It also covers the converse, confirming that the system does not allow unintended users to interfere with its intended use, or allow it to be used in unintended ways.

Why the difference matters

A product can therefore pass verification and fail validation. This happens when it is built exactly to a specification that does not capture what the stakeholders need. Verification closes an inner loop between the design and the requirements as written; validation closes an outer loop between the delivered system and the purposes and people it was built for.

Two loops. Verification, the inner loop, checks the design and product against the requirements as written. Validation, the outer loop, checks the delivered system against the stakeholders' needs and the concept of operations. Validation (outer loop): the delivered system against the stakeholders' needs and the concept of operations Verification (inner loop): the design and product against the requirements as written Design andproduct Requirementsas written Stakeholders'needs and theconcept ofoperations
Two loops: verification checks the design and product against the requirements as written; validation checks the delivered system against the stakeholders' needs and the concept of operations.

When validation by simulation produces an unsatisfactory result, the cause may lie in the design, in the simulation model, or in the requirements themselves. System requirements should be regarded as flexible up to a point: a requirement found to be unduly driving up complexity, cost or risk is analysed critically and, where appropriate, raised with the customer. Problems found when users validate a finished system may likewise be resolved by changing the concept of operations rather than the hardware or software. The later such a problem is found, the more costly its correction, and in some cases a late validation failure can cause a project to fail or be cancelled.

Where the practical dividing line falls

The definitions separate the two activities by their reference. In practice, the conditions under which evidence is gathered usually separate them too.

  • Verification is typically carried out during development, often on components and subsystems, in a laboratory, on a test stand or in another environment where conditions are tightly controlled and prescribed.
  • Validation is typically carried out during or after system integration, on the complete system, in a real or simulated operational environment, with real or representative users who have no special knowledge of the design.

A frontal crash test run at prescribed speeds and angles is verification, whereas months of driving a finished vehicle in desert heat and in snow are validation. A satellite shaken on a vibration table to a load spectrum defined for its launch vehicle is primarily verification. Letting children play with a new toy in a nursery before committing to mass production is validation.

This distinction is a rule of thumb, not part of the definition. Where it is very hard to design isolated tests that check requirements one at a time, as with specialist medical equipment whose performance depends on how clinicians use it, evidence may have to come from a pilot deployment in a realistic setting almost from the start. The confidence obtainable before operation also depends on novelty: a system similar to many already in service can be largely verified before use, while a first-of-its-kind system carries residual unknowns into its first operation.

Who carries out each activity

Verification is normally carried out by the developer. Operational validation of the complete system is carried out by the customer, by an independent test agent acting for the customer, or by the users themselves, in an environment identical or closely similar to that of intended use, and satisfactory results are a prerequisite for production and deployment. In validation, users may operate the system as part of their normal duties rather than follow a scripted step-by-step procedure, since their judgement is the final measure of how well it meets their needs. Each need recorded in the concept of operations is traced to a validation activity. One rule of thumb is that anything important enough to be written down as a need should be validated at least once in real-world use.

Methods

Whoever carries them out, both activities use the same four basic methods:

  • Inspection examines an item, generally without operating it, to confirm properties such as colour, weight or freedom from manufacturing flaws, including by non-destructive techniques such as X-ray and eddy-current inspection.
  • Analysis uses calculation, modelling or simulation under defined conditions to show theoretical compliance, where testing under realistic conditions is not achievable or not cost-effective. Analysis by similarity transfers the verification status of an existing item to a new one. It is valid only if the two are similar in design, manufacture and use, the earlier verification was at least as stringent, and the new operating environment is no more severe.
  • Demonstration shows qualitatively that the system performs its functions, with little or no instrumentation.
  • Test subjects the item to controlled real or simulated conditions and records quantitative data, usually with special equipment, for comparison with predicted behaviour.
The four methods used for both verification and validation: inspection examines an item, generally without operating it; analysis uses calculation, modelling or simulation under defined conditions; demonstration shows qualitatively that the system performs its functions; test applies controlled real or simulated conditions and records quantitative data. Inspection examines an item, generallywithout operating it Analysis calculation, modelling orsimulation under defined conditions Demonstration shows qualitatively that the systemperforms its functions Test controlled real or simulated conditions,quantitative data recorded
The four methods, used for both verification and validation: inspection, analysis, demonstration and test.

The choice of method is a matter of engineering judgement about which will most reliably show conformance, and the method used for a requirement can change as the design matures. The more critical a requirement is to the operation of the system, the more intensive its verification tends to be, through more tests, different tests or a combination of methods. Testing is often the method on which most money is spent, and a single test that checks several requirements at once reduces cost.

Models and simulations play two roles. They are a means of verification and validation within analysis, and they are also artefacts that must themselves be verified and validated before their outputs can be relied on. Translating requirements and designs from natural language into precise models makes specifications analysable and allows parts of design checking and testing to be automated. Physical testing nonetheless retains a large role as modelling and simulation grow.

Across the life cycle and down the system hierarchy

Those methods are applied at every stage of development, not only at the end. Verification and validation are often drawn on the right-hand, integration side of the V-model, in which each level of definition on the left is paired with a level of verification on the right, and test cases are derived from, and traced to, the requirements at that level. The pairing holds for the realised product, but neither activity is confined to it. Both can be applied to any artefact produced on the way to the system: stakeholder needs and requirements, system requirements, models, simulations, architectures, designs and the verification procedures themselves. Validation can be carried out in each development phase using that phase's products, such as models, and not only at delivery. Nor do the activities end at delivery: verification may be repeated in service, and readiness for use reassessed after maintenance.

The V-model. Down the left: stakeholder needs and requirements, system requirements, architecture and design, each verified and validated during definition. At the base: realised elements. Up the right, paired level by level: elements verified and validated then integrated, system verified against its requirements, system validated against the stakeholders' needs. Stakeholder needsand requirements System validated againstthe stakeholders' needs Systemrequirements System verified againstits requirements Architectureand design Elements verifiedand validated,then integrated Realisedelements Verified and validated during definition Dashed lines: each level of definition is paired with a level of verification or validation.
The V-model: each level of definition is paired with a level of verification or validation, and the requirements, models and designs on the left are themselves verified and validated during definition.

Recursion down the hierarchy

Both activities are recursive:

  • Stakeholder needs and requirements are validated against the stakeholders' real-world expectations.
  • System requirements are verified against the higher-level requirements from which they derive, and validated against the stakeholder needs.
  • The architecture and design are verified against the requirements and validated against the needs.
  • Each realised element is verified, and validated in its operational environment, before it is integrated into the next level up, and discrepancies are corrected before that integration.

Integration, verification and validation continue in this way until the complete system has passed validation.

Early detection and early planning

Defects are cheapest to correct when they are found early. The success of a programme may relate to how often and how diligently validation is carried out during early design, while changes of direction can still be made cost-effectively. Errors are best found at the lowest level of integration, where they can be corrected through the design process with the least disruption.

Planning therefore starts with the requirements. Verification planning begins while system requirements are being defined, setting success criteria, method and strategy for each; validation planning begins while stakeholder needs, operational scenarios and the operational concept are being defined. Early planning improves the chance that the full verification programme will be resourced.

Verification and validation also need their own enabling systems: test facilities, rigs, simulators and instrumentation. These need lead time, and the enabling system used to verify a design is itself designed and integrated while the operational system is being designed.

Evidence, records and decisions

Every activity planned across the life cycle produces evidence, and that evidence is planned, recorded and accepted formally.

The plan and the report

A verification and validation plan states, for each requirement, the method to be used, the test or event in which it will be checked, and the equipment and conditions involved. The allocation is commonly held as a matrix. On some programmes the customer specifies the required method for every requirement, which makes the verification approach contractual.

A verification report records:

  • the requirement verified and its bidirectional traceability;
  • the method;
  • any special equipment, conditions or procedures;
  • the results, with pass or fail declarations;
  • any anomalies, with the corrective actions taken;
  • the versions of the requirement set, the product and the tools used.

A discrepancy leads either to redesign and retest or, where the shortfall is judged minor, to a waiver: a documented authorisation releasing a programme from meeting a requirement that is already under configuration control. Granting a waiver is a risk decision.

Evidence holds only for the configuration that produced it

Verification of a design is performed once, however many units are built, for as long as the design does not change. The principle is summarised as "test as you fly, fly as you test": the configuration tested should be the configuration used, because failures have followed modifications made after an item passed its tests.

The evidence is also what gives a verified artefact its value. A structural model whose predictions have been checked against physical load tests is physically identical to one that has not, but it is far more valuable. On one fighter aircraft programme, finite element models verified to within plus or minus 5% at every location were priced accordingly, some at millions of dollars.

Who accepts the evidence

Acceptance of the evidence rests with an approval authority, the party entitled to decide whether it is sufficient. At the level of the complete system, validation may be performed with or by the acquirer (the customer), as set out in the supplier agreement, while at lower levels a supplier may validate its own elements without the acquirer's direct involvement. The final readiness review before a system enters service checks that all verification and validation activities have been passed, which waivers have been granted and what residual risk remains.

The core pair is well defined, but the vocabulary around it varies between organisations. The US National Aeronautics and Space Administration (NASA) distinguishes four activities:

  • Verification confirms that the system satisfies all specified requirements and, as above, is performed once per design.
  • Qualification is the subset of verification performed at the extremes of the environmental envelope, confirming that the design works with the expected margins. It is also performed once.
  • Acceptance is a smaller subset performed on every unit produced, showing that its manufacture and workmanship conform to the verified design.
  • Certification is the audit by which the body of evidence is presented to a certifying authority, which certifies the design for use.
NASA's four activities. Once per design: verification, that the system satisfies all specified requirements; qualification, at the extremes of the environmental envelope; certification, the evidence presented to a certifying authority. On every unit produced: acceptance, that manufacture and workmanship conform to the verified design. Once per design Verification the system satisfies all specified requirements Qualification at the extremes of the environmental envelope Certification the evidence presented to a certifying authority Every unit produced Acceptance manufacture andworkmanship conformto the verified design
NASA's four activities: verification, qualification and certification once per design; acceptance on every unit produced.

Other systems engineering usage treats qualification more broadly: either as the state reached when all verification and validation actions have been completed and the system is qualified for its intended use, or as the verification and validation of a design followed by the stakeholders' acceptance of it. Certification in the wider sense is a written assurance, typically from an outside authority, that a system meets identified legal or industrial standards, such as airworthiness standards for aircraft, or CE marking in Europe and Underwriters Laboratories (UL) certification in the United States and Canada.

Within organisations, verification and validation are typically done by people other than those who wrote the requirements or did the design, and one way for a customer to avoid repeating a supplier's tests is to send knowledgeable representatives to witness them.

Across industries

The same pair of terms is used wherever products are designed against requirements, with each field adding its own detail:

  • Quality management: ISO 9001:2015, the quality management systems standard (now superseded by ISO 9001:2026), applies the pair to design and development: verification confirms that design outputs meet design inputs, and validation confirms that the resulting products and services meet the requirements for their specified application or intended use.
  • Software: IEEE 1012, the IEEE standard for system, software and hardware verification and validation, sets out V&V processes for software alongside systems and hardware.
  • Computational models: validation evidence is tied to the model's intended use. In building evacuation modelling, which in 2013 had no international standard for the purpose, test sets written for passenger ships were often reused for buildings, outside their original context of use.

Limits of verification and validation

Whatever the field, verification and validation have limits.

They cannot be exhaustive

Exhaustive verification and validation is not possible for any system of significant complexity. Not every combination of actions under every operating situation can be covered, and even a relatively small software program can contain more logical paths than can practically be tested, so a finite set of test cases must be selected. For automated driving systems, a purely statistical demonstration by road testing has been estimated to need several hundred million kilometres of driving, depending on the assumptions made, repeated for every modified version; the entire paved road network of the United States is about 4.3 million kilometres long. Approaches for such systems therefore describe classes of scenarios at several levels of abstraction, and use simulation to search for the rare combinations of conditions in which critical behaviour occurs.

Passing every planned test does not guarantee correct operation

Failures in operation often occur outside the scenarios that were tested, so test conditions are made as representative of real operation as possible, and the representativeness of simulated conditions and stand-in components is itself examined. The test equipment is part of the evidence: sensors out of calibration lead to wrong conclusions, and test equipment has to be matched to the conditions of the test.

Residual risk remains

Every system worth building enters service with some residual risk; the requirement is not that all risks are reduced to zero. Testing is one means of reducing technical risk, yet when budgets are constrained it is often among the first activities cut, trading a cost saving for technical risk and potential rework.

Case study: the London Millennium Footbridge

The London Millennium Footbridge passed its design verification, yet did not perform acceptably once it was in use. A shallow suspension footbridge across the Thames engineered by Arup, it opened on 10 June 2000. During design its response to pedestrians was assessed with a modified BS 5400 approach for vertical excitation, using a higher input force than the code recommends, together with the lateral and torsional response to eccentric vertical loads. Tests with a few people on the bridge in May 2000, before opening, confirmed the calculated accelerations.

Between 80,000 and 100,000 people crossed on the first day, with up to about 2,000 on the deck at once. When large numbers were on the south and central spans, unexpected lateral vibrations built up, and a significant number of pedestrians had difficulty walking and held the balustrades. No excessive vertical vibration was observed. Numbers were restricted on 11 June and the bridge was closed on 12 June.

Modal testing after closure found the frequencies of the modes that had vibrated within 10% of the design predictions, and the damping higher than assumed. Arup concluded that the analytical model was valid, and that the movement came from a lateral loading effect not anticipated during design. Their investigation attributed it to what they called synchronous lateral excitation: pedestrians falling into step with a small sway, which increases it. They found that the effect was not tied to the bridge's innovative form and could occur on any bridge with a lateral frequency below about 1.3 Hz loaded by enough pedestrians. Later researchers have proposed other explanations for the pedestrian forces.

The retrofit added fluid-viscous dampers (37 were specified) and tuned mass dampers, and the bridge reopened on 22 February 2002. Arup reported that a clause on the phenomenon had been added to the draft British bridge loading standard BD 37/01, that the Institution of Structural Engineers' Standing Committee on Structural Safety had issued a warning, and that Arup had written to government and professional bodies with the findings. Their summary lesson was that satisfactory performance below the critical number of users is no guide to performance above it.

The principle. The bridge was built to its design and its design verification held, but the design basis did not capture the bridge's actual use: a dense crowd walking on it. Only operation by real users in real use exposed that shortfall: a validation finding, not a verification one.

Summary

Verification confirms that a product meets its specified requirements; validation confirms that it meets the needs of its intended users in its intended use. The two check against different references, so a product can pass verification and fail validation. Both rest on objective evidence from inspection, analysis, demonstration and test; both apply at every level of the system and every stage of the life cycle; and the defects they reveal are cheapest to correct when found early. Neither can be exhaustive, so every system enters service with some residual risk, which is weighed at the final readiness review.

Referenced standards

  • ISO/IEC/IEEE 15288, the international standard for system life cycle processes.
  • ISO 9001:2015, the quality management systems standard (withdrawn in 2026 and replaced by ISO 9001:2026).
  • IEEE 1012-2016, the IEEE standard for system, software and hardware verification and validation (since superseded by IEEE 1012-2024).
  • BS 5400, a modified form of whose approach was used to assess the Millennium Footbridge's response to vertical pedestrian excitation.
  • BD 37/01, the British bridge loading standard, to whose draft a clause on synchronous lateral excitation was added.