Regulations and compliance · Practical guide

Annex 11 data migration: preserving value and meaning

A complete import may lose the meaning of units, statuses or object relationships. For a defined migration into a new GMP data state, set independent expected results and decide how to handle discrepancies. Check that historical decisions remain intelligible. Ongoing monitoring of an operational interface is a separate task.

Table of contents

At a glance

How do you demonstrate that a migration preserves data value and meaning under Annex 11?

Annex 11 section 4.8 requires checks that transfer to another format or system preserves data value and meaning. Define the source, migration scope, transformation and intended use in the target. Compare identities, relationships and states as well as record counts and totals. Document exclusions, deviations and continuing access to historical records. A successful import does not in itself justify releasing the migrated data.

  • The scope of the released transfer must have a identifiable starting point.
  • Equal sums do not exclude incorrect assignments or status discrepancies.
  • Coordinate migration, historical access and system retirement.

1. Values and significance as two inspection objects

Annex 11 January 2011 will continue to be listed in the Human Medicines Section of EudraLex on 7 October 2026. Its migration passage is in § 4.8. § 7 deals with storage and security; these controls can support migration, but do not replace its verification.

A value may remain numerically unchanged yet convey a different meaning: a result may be associated with the wrong sample, a date may lose its temporal context, or a status may be misinterpreted as a usage authorization. Therefore, it is essential to describe what the data means professionally—both before and after transfer. The following methods specifically address this task, without prescribing a universal migration format.

Sources: [1] [4]

2. Define the starting inventory and establish boundaries

State the source, data date, period of capture, and selection criteria. Distinguish between active data, historical records, and intentionally excluded information. An exclusion must be justified and, if applicable, an alternative pathway for continued availability must be provided. "Not in the new system" does not automatically imply "no longer required."

In addition, address changes during transfer. If the source continues to be used, new or modified datasets must be considered in accordance with the agreed procedure. A frozen starting balance and a regulated re-transfer are possible approaches; the suitability of either depends on the operation. Unclear time boundaries complicate any subsequent completeness comparison.

Sources: [1] [2] [3]

3. Explicitly describe all transformations

Prepare a concise mapping between source and target models. Consider units, character encoding, date formats, identifiers, and relationships. For status fields, provide a technical justification for the intended effect. A data record previously deactivated in the legacy system must not be reactivated solely because the target system lacks a corresponding status.

The table serves as an independent decision-support tool. Not every piece of information can be losslessly transferred into every new structure. The implications for future usage and required access rights must then be evaluated. An unverified PDF export is not a general solution for lost metadata or dynamic reporting capabilities.

Own transformation matrix prior to import
Data itemPlanned takeoverAssessed significance
Quantity and unit All quantities in grams Same physical quantity per identical object
Status Map source code to target states No new authorization of use without a valid basis
Historical decision Take decision together with reference Person, time, and subject remain clear
Relationship Link assay probe to associated result No orphaned or misplaced results
Context that was not migrated Historical access tracking Required information remains retrievable and readable

Sources: [1] [3]

Further reading:

4. Example: completed quantity and status model

The defined source inventory contains exactly three material items: A with 2.5 kg and status locked, B with 750 g and status available, and C with 1.25 kg and status locked. The target unit is grams. Expected values are A = 2,500 g, B = 750 g, and C = 1,250 g.

The total sum amounts to 4,500 g. The available quantity is nonetheless only 750 g. If the status of A and B is swapped, the dataset count and total sum remain correct, but the available quantity becomes incorrect. This example illustrates why sum reconciliation and subject-matter assignment address different questions.

Example with independently calculated target values
Object or failure scenarioSource versionExpected result
A 2.5 kg; locked 2,500 g; furthermore locked
B 750 g; available 750 g; further available
C 1.25 kg; locked 1,250 g; furthermore locked
Total inventory Three objects Three associated objects; total 4,500 g
Usage quantity Only B available 750 g available
Intentional status error Available, B locked Detect a deviation despite identical total quantity

Sources: [1] [3]

5. Establish the comparison independently of the migration program

A script used for validation that reuses the same erroneous mapping rule as the import may generate false assurance. Expected values for selected critical cases should be derived from approved sources and through traceable, calculable methods. For large inventories, automated comparisons can be helpful, but must be appropriately designed for their intended purpose.

Combine appropriate tests based on scope, identity, relationships, and technical conditions. Justify which fields should be fully compared and where targeted selection is appropriate. There is no general minimum number of data sets required to validate any migration. The selection must specifically include rare but potentially critical states.

Sources: [1] [3]

6. Failures prior to productive use determine

Variations require a clear and justified evaluation: What is affected, what change in significance has occurred, and could the error impact further data? A correction to the transformation rule may necessitate a reprocessing and appropriate revalidation. Clearly document separately what technical correction was made and on what basis the subsequent decision rests.

For handling open items, the implications and permitted usage are decisive. The absence of a clear explanatory text may be assessed differently than an incorrect usage authorization. The term "remaining item" does not render a critical issue acceptable. Release should specify the actually verified inventory and, if applicable, clearly defined usage limits.

Sources: [1] [2]

7. Plan access, rollback, and decommissioning together

The system retirement conclusion does not address all questions regarding the legacy system. Which historical data remains there, who has access to it, and what technical environment is required for retrieval? PIC/S explicitly addresses risks related to long-term readability and potential functional failures during format changes. These limitations should be clearly understood prior to system decommissioning.

A rollback plan must align with actual operations. If work has already begun in the target environment following a transition, returning to an older backup may result in loss of interim data. Therefore, define how interim activities shall be handled. Backups, historical data retention, and migration comparison serve distinct purposes and must be coordinated accordingly.

Sources: [1] [3] [2]

8. Justify the migration decision conclusively

The closure should indicate which data version was imported using which rules, which comparisons were performed, and how deviations were handled. Link source reference, transformation status, test evidence, and user decision. A simple status message "Import successful" leaves the technical and quality aspects unverified.

For preparation, process knowledge, understanding of both data models, and accountable quality assessment is required. Effort increases significantly in cases of ambiguous historical states, inconsistent units, and missing relationships. These factors can be identified early, before a blanket acceptance of timelines displaces the actual required validation work.

  • Source, date, scope and exclusions clearly defined.
  • Transformations and meaning in the process described.
  • Independent scope including unusual conditions prepared.
  • Number, values, identities, relationships, and status appropriately reconciled.
  • Variations, corrections, and repeatable proof assessments are evaluated.
  • Historical access, rollback, and transition into operation clarified.

Sources: [1] [2]

Project experience and collaboration

Verify that migrated data retain both their values and their meaning

Process owners select representative records and deliberately challenging migration cases. IT explains the transformations, and the business establishes what a correct result looks like. An unchanged transfer may justify different reconciliation from changed units, statuses or relationships. We trace missing records and ambiguous mappings individually so totals cannot hide them. Before deciding on use, your team records what has been demonstrated and how open findings affect the decision.

Frequently asked questions about Data migration

Which Annex-11 section addresses data migration?

§ 4.8 addresses data transfers to other formats or systems and the assessment of value and significance. § 7 covers storage and backup. Both may be relevant, but they must not be cited as if an existing backup procedure constitutes valid evidence of migration.

Sources: [1]

Are equivalent dataset sizes sufficient?

No, not as the sole evidence for all relevant properties. Three misassigned data records remain three data records. Additionally, verify the significant values, relationships, and states. The required comparisons follow from the data model, transformation, and subsequent usage.

Sources: [1] [3]

Why is a correct total sum insufficient?

Errors may balance out or lie outside the specified quantity. In the example presented in this article, the total sum remains 4,500 g, even though a status change may distort the allowed usage. Therefore, the comparison must also capture the correct assignment of each object and its respective technical condition.

Sources: [1]

Must all historical data be transferred to the new system?

Not necessarily required to be migrated to the same new application. However, records must remain available according to their intended use and storage conditions. Justify exclusions and alternative access. A decision against migration is not automatic authorization for destruction of the existing system.

Sources: [1] [2] [3]

Can the source continue to operate during migration?

This may be achievable through an appropriate procedure. In such cases, new and amended data must then be reviewed and validated against the baseline. Time limits, reprocessing, and reconciliation procedures should be established. Without these rules, even a correctly executed initial export may result in incomplete data.

Sources: [1] [3]

Are samples inherently unacceptable?

No, sampling is not fundamentally excluded. Its appropriateness depends on risk, transformation, and existing controls. Annex 11, Section 4.8 requires verification that value and significance are preserved during migration; the document does not specify a universal sample size or blanket full inspection. Targeted validation of critical mappings and rare, high-impact scenarios should be conducted. Randomly selected normal cases may provide insufficient coverage.

Sources: [1]

Can a PDF completely replace the old system?

Only if the information and relationships required for the intended purpose are adequately retained. Dynamic data, metadata, or future analysis capabilities may be lacking. Assess the specific suitability; a readable output does not prove general equivalence.

Sources: [3] [2]

What do we do following a faulty status translation?

Assess the scope and implications, identify and correct the root cause under controlled conditions, and determine the relevant repeat inspections. Additionally, verify whether prior decisions were based on a false state. Simply changing the display does not automatically eliminate all consequences.

Sources: [1]

Is a backup always sufficient contingency plan?

No. After productive use of the target system, new activities may exist that are not present in the previous backup. A rollback plan must take into account these interim data and process states. Restoration and professionally consistent continuation must be considered together.

Sources: [1] [3]

Who decides on the use of the target population?

The decision is made by the operationally authorized role as defined in your quality system. It is advisable for process owners, system owners, and quality assurance to jointly evaluate: technical fitness, functional usability, and remaining quality-related questions. Annex 11, Section 2, and the Glossary define the fundamental roles. A third-party service provider may perform the import and provide supporting documentation; however, their technical confirmation does not substitute for a justified decision by your operation.

Sources: [1] [2]

Sources and review date

Source date: 07.10.2026. The relevant area here is human medicinal product GMP. EudraLex incorporates Annex 11 and Chapter 4 as of January 2011; the consulted draft from 2025 is not considered a current regulatory requirement. The completed examples and tables are internal work aids and do not constitute official template documents.

  1. EU GMP Annex 11: Computerised SystemsRevision January 2011; Effective date 30.06.2011; Access date 07.10.2026. Relevant provisions: §§1, 4.1–4.2, 4.7–4.8, 7, 10, 13, 16–17
  2. EU GMP Chapter 4: DocumentationRevision January 2011; Effective date 30.06.2011; Access date 07.10.2026. References: §§4.1, 4.8–4.12
  3. PIC/S PI 041-1: Good Practices for Data Management and IntegrityPI 041-1, 01.07.2021; Access date 07.10.2026. Citations: §9.4, in particular p. 38–39; §9.1.5
  4. European Commission: EudraLex Volume 4Status as of 07.10.2026: Human medicinal product section implements Annex 11 January 2011 and Chapter 4 January 2011; Access date 07.10.2026. Relevant provisions: Human medicinal products / Annex 11 / Chapter 4
  5. European Commission: Consultation 2025 on Chapter 4, Annex 11 and Annex 22Status as of 07.10.2026: closed; 07.07.–07.10.2025. Occurrences: details; revision of Annex 11; links explicitly marked as draft guidelines. Serves solely for status demarcation and is not considered an authoritative source of requirements.

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Address the specific control and verification requirements

Daniel Herrmann Consulting supports you in adapting this procedure to your system portfolio, evaluating gaps, and preparing appropriate validation evidence. The scope and level of involvement are determined by your actual process.

  • GxP system decommissioning →

    Retire legacy systems while preserving data, context and auditability. We plan and manage the decommissioning so that data, metadata and audit trails remain readable and retrievable after shutdown.

Send a project enquiry Customer testimonials →