Regulations and compliance · Practical guide

GAMP 5: evidence for agile validation across increments

GAMP 5 and agile delivery can work together when you can show which version of a function was tested for which use. A closed ticket or successful sprint review supplies only part of that evidence. Two example increments link requirements, the actual system version, results and permitted use through frequent changes to a GMP system.

Table of contents

At a glance

How can we keep each iteratively developed version linked to its requirements and evidence?

Link each usable increment to an identifiable feature version, the affected requirements and relevant evidence. Assess the impact of changes on previously demonstrated functions and existing data. Make the use decision for that version and its known open issues. Agile development does not require a fixed document package per sprint. In GMP scope, it still needs justified validation, change control and traceable evidence.

  • Clear distinction between status, verified status, and permitted use.
  • The evidence received is based on actual change effects.
  • Identify open findings and functions that remain disabled in the release.

1. Begin with the intended use

Agil describes the operational workflow. Whether a function is GMP-related depends on its usage and impact. Therefore, specify which process is supported, what data is generated, and which decision depends on it. This article's example assumes human medicinal product GMP; other application areas require a separate requirement mapping.

The public ISPE description of the second edition of GAMP 5 supports iterative and incremental approaches. This does not imply a requirement for a specific sprint duration or a particular tool. Annex 11 requires, in the relevant context, a justified lifecycle and appropriate validation evidence. Both perspectives are compatible when changes and proof-of-standards are effectively managed.

Sources: [1] [2]

2. Define a verifiable basis per increment

A verifiable foundation must indicate what was tested. This may include software version, configuration, relevant master data, interface status, and test environment. The required depth depends on the system. A version number alone is insufficient if settings alter the technical behavior independently.

In addition, clearly define the requirements applicable to this increment and those functions that are not yet intended for use. A team may work on later-stage features while continuing to use an earlier version. However, it must be unambiguously established which artifacts correspond to which baseline. A changing "current" link lacks sufficient historical context to support such mapping.

Sources: [2]

3. Example: two stations of a sample import process

A laboratory is implementing an interface for sample requisition entries. Increment A imports requisitions and flags duplicate requisition identifiers. Increment B adds a modification function for already entered requisitions. The professional assumption is: a requisition must not inadvertently generate a second instance through modification. The table describes planned evidence requirements.

A has twelve cases, comprising eight new orders and four repeat transmissions. The expected outcome is eight orders and four identifiable rejections. B adds six change cases, four allowed and two prohibited. The final number of orders remains eight. A change does not create an additional order.

Own incremental matrix with subject-matter calculation of required values
StatusBehaviour and RationalePlanned evidenceExpected decision
A: First import Create a new identifier and reject a duplicate 8 new + 4 repeated transfers = 12 cases 8 orders; repetitions identifiable, evaluate first import
B: Change Intentionally modify existing orders; reject unauthorized changes 4 permitted + 2 unauthorised changes = 6 additional cases Still 8 orders, with traceable revision history
B: Interaction Change may affect the duplicate checking. Select and check the affected A cases again Assess the extent of the intended use when interaction has been resolved
B: Scope not included Automatic deletion remains disabled Check the configuration and permission No release of a non-intended deletion function

Sources: [2]

4. Link requirements and results purposefully

A user story can express a process requirement if it is sufficiently precise. “As a user, I want to change orders” leaves open who may amend which fields and under what conditions. Add the rules and expected error handling. Acceptance criteria are especially useful when they define a boundary that can be tested.

Link the requirement to the specific test objective and result. Existing evidence may be reused if its validity for the new state can be justified. A mere count of linked test cases provides little insight into coverage. In the example, the new change function must be linked to the existing duplicate rule, even though the requirement text remains unchanged.

Sources: [2]

5. Decide on adoption, supplementation, and repetition

Our decision matrix categorizes different types of changes. It does not specify a fixed minimum number of regression tests. The starting point is always which previous assertion becomes uncertain due to the change. Even a configuration change without new program code can have a significant technical impact.

Explicitly justify, when an existing validation report remains valid without change. This approach is more efficient than routinely repeating all tests or simply reviewing only new tickets. In cases of unclear dependencies, uncertainty must first be resolved; the absence of a technical overview does not constitute evidence that no impact exists.

Internal assistance document for release decision justification
ChangeAffected statementAppropriate actionJustify the rationale.
New change field Only authorized fields are editable New Positive and Negative Cases Authorization and Professional Implications
Common import logic changed Duplicates are detected Affected first import cases recur Dependency between A and B
Unchanged help display No detectable changes in data processing Assess impact; adopt appropriate existing evidence Justification for why critical functions remain undisturbed
Unknown supplier patch Effect not yet determinable Obtain change information; verify remaining gaps No blanket acceptance when the basis is unclear

Sources: [2] [3]

6. Do not let open findings disappear into the backlog

A backlog entry may represent a technical task, but it does not substitute for the evaluation of a defect in relation to the intended use. Assign the finding to the relevant site. Describe its impact, if applicable, including any effective mitigation measures, and determine whether the intended use remains acceptable.

In this example, creating a duplicate order on retry is more than a cosmetic issue. A manual control can support a decision only if its operation, coverage and effectiveness are adequately demonstrated. A promise to fix the issue later establishes neither a control nor permitted use. Link subsequent corrections to the original finding and appropriate evidence.

Sources: [2] [3]

7. Integrate development and operations at the handover point

For the decision on usage, the specialist process, technical readiness, and quality requirements should be integrated. This does not imply a universal signature combination. In your organization, clearly define who assesses the technical scope, who is responsible for the specific technical provisioning, and who has the authority to approve usage.

A successful automated run validates only the tests performed under the applicable conditions. Additionally, migration, authorizations, work instructions, or support may be relevant. Small teams particularly benefit from a concise handover note: current status, permitted functions, supporting evidence used, open items, and actions to take in case of issues. This note may reference received detailed evidence.

Sources: [2]

8. Checklist for the next usable increment

Evaluate completeness based on a specific decision sequence, not on the number of generated documents. Could a subsequent handler identify why Station B was authorized and which evidence from A remains valid? If oral project knowledge is required, a critical link is missing.

The checklist is a standalone support tool. It can be integrated into an existing release procedure; an additional parallel validation archive is not automatically required. Key factors are controlled assignment, availability, and a traceable decision process.

  • Intended use and permitted functional scope are defined.
  • Software, configuration, and relevant data base are identifiable.
  • Requirements and validation objectives are assigned to the increment.
  • Impacts on previous functions and existing data are assessed.
  • Accepted evidence and additional investigations are justified.
  • Findings including effective limitations are approved.
  • Operational readiness and specific usage decisions are documented.

Sources: [2]

Project experience and collaboration

Keep requirements and evidence linked across iterations

With the process owners, developers and Quality, we follow a requirement through several planned iterations. We examine when it is refined, which risks change and which test applies to each version proposed for release. Short evidence cycles may suit small, independently usable functions. Closely coupled steps also need testing together. We define handovers so that completion of the technical work is clearly distinguished from the decision on regulated use.

Frequently asked questions about Traceable agile validation under GAMP 5

Does GAMP 5 allow for agile development?

The public description of the second edition supports iterative and incremental approaches. It does not establish a specific agile methodology. For GMP use, it remains essential that requirements, risks, changes, and documentation are traceable throughout the product lifecycle.

Sources: [1] [2]

Do we need to perform a complete validation after every sprint?

Not at all. First, check whether a stand is intended for regulated use and which functions have changed. The scope of evidence depends on the impact. Unchanged appropriate evidence can continue to contribute; relevant interactions must not be overlooked.

Sources: [2]

Can a user story replace a requirement?

It may carry a requirement if the intended purpose, rules, and boundaries are adequately described. A brief request lacking verifiable acceptance criteria is often insufficient. What is important is consistent alignment with relevant validation evidence, not the title of the document used.

Sources: [2]

What baseline do we need for configurable software?

Capture the characteristics that determine the tested behavior. In addition to program version, this may include configuration, permissions, interfaces, and relevant master data. The scope is system-specific. A version number alone is insufficient if significant settings can be altered independently of the version.

Sources: [2]

Can previous tests for a new increment be considered valid?

Yes, if the statement remains relevant to the new state. Justify this based on the change and its dependencies. In the example of this article, an unchanged duplicate requirement may still necessitate new testing because the new change function utilizes the same technical processing pathway.

Sources: [2] [3]

Is a Sprint Review already a GMP release?

Not automatically. A Sprint Review can provide valuable technical insights. Whether it also serves as a usage decision depends on authorized governance, existing evidence, and evaluated open items. The naming of a meeting does not imply or indicate the presence of these prerequisites.

Sources: [2]

How do we handle a deactivated new function?

Document that the function is not intended for use and verify the effective restriction. It is also relevant to assess whether the function impacts other processes, even if deactivated. Any subsequent reactivation requires a separate evaluation and decision regarding implications and usage, based on appropriate grounds.

Sources: [2]

Are all open defects to be resolved prior to each use?

A general statement would be too broad. Assess the significance and remaining risk for the intended use. Critical, unresolved implications may preclude usage. A reasonable limitation must be specific, effective, and documented; a later backlog date alone is insufficient.

Sources: [2] [3]

Does a green pipeline indicate appropriate technical suitability?

It first proves that the automated tests set up were successful under running conditions. Whether their scope and result bear the use in the process must be assessed. Unrecorded permissions, migrations or workflows are not verified by the green display.

Sources: [2]

How can small teams keep the effort manageable?

Leverage existing tools and clear linkages. A concise decision with preserved detailed justification is often more beneficial than redundant documentation. Simplify archiving without losing the connection between requirement, status, finding, and allowable use.

Sources: [2] [1]

Sources and review date

Source date: 7 October 2026. Regulatory requirements, guidelines, and industry practices must be categorized separately. Tables, figures, and case examples are to be considered standalone reference materials; they do not constitute completed product testing.

  1. ISPE: GAMP 5 Guide, 2nd EditionPublic publisher description and edition information; no access to paid full text.
  2. EU GMP Annex 11 – Computerised SystemsRevision January 2011; listed in the human medicinal product section. Principle, §§1–17.
  3. EMA: ICH Q9(R1) – Quality Risk Management, Corr.2EMA version dated 23 January 2025, Corr.2. This is Q9(R1), applicable in the EU since 26 July 2023. See sections 1–6, particularly 4.3 and 5.1–5.3.

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Jointly develop the specific evidence process

We help your team integrate development work, technical evaluation, and usage decisions across multiple increments.

  • GAMP 5 consulting →

    Understand how GAMP 5, EU GMP Annex 11 and 21 CFR Part 11 apply to your systems. We translate the relevant requirements into proportionate validation work and evidence your team can defend in an audit.

Send a project enquiry Customer testimonials →