Table of contents
At a glance
How do you check critical manual data entries under Annex 11?
For critical data entered manually, Annex 11 section 6 calls for an additional accuracy check by a second person or validated electronic means. Identify which errors can occur and their consequences. Check assignment, unit and source alongside the number itself. A range check or redisplay of the entered value does not automatically provide an independent comparison with the correct expected value.
- The potential impact of a deviation determines the criticality of an input.
- An electronic control must align with the actual failure scenario.
- Control point, correction, and subsequent use are interrelated.
1. Identify critical data based on their usage
The basis is Annex 11 January 2011, and on October 7, 2026, it continues to be listed in the Human Medicines section of EudraLex. § 6 explicitly deals with critical manually entered data. The passage does not prescribe that any entry should be checked by two people.
Start with the following decision: Does the input, for example, alter a calculation, select a material, or assign a laboratory result? Describe a plausible error and its potential consequence. A manually transcribed sample identifier can be just as significant as a concentration. The term "critical" should be made understandable through the workflow, not merely by a tick mark in the form.
2. Consider more than typos
Trace the entry from origin to usage. Potential errors include a swapped sample, incorrect source revision, misplaced decimal point, missing unit, or transfer of an old result. A correct value from a wrong report may render a simple plausibility check completely inadequate.
Capture also the working conditions: Is the source at the input location legible? Are multiple samples being processed simultaneously? Does the interface clearly indicate which dataset is being entered? Such questions serve as independent methodological aids for implementation. They prevent controls from functioning only on paper, while operations risk causing errors.
3. Select the additional control appropriate from the list
A second person can compare the original source, destination and meaning. An electronic check might use an independent reference or test a defined mapping. Demonstrate that the check suits its purpose. The mere presence of a warning message proves little about accuracy.
The table presents internal selection questions, not an officially mandated priority order. Multiple controls may interact synergistically. What is critical is which error class is actually captured and which gaps remain. An acceptable value range may protect against certain magnitude errors, but does not necessarily prevent the occurrence of two swapped results that both fall within the range.
| Input scenario | Appropriate testing concept | Remaining question |
|---|---|---|
| Sample identification from report | Source and target data sets to be additionally cross-checked | Is the correct report version being used? |
| Sample weight and unit | Verify value, unit, and conversion against source | Does the control also identify a plausible erroneous value? |
| Status of approved decision | Identity and decision reference comparison | Does the control prior to intended use function? |
| Free comment without significant impact | Assess criticality and typical documentation control practices | Does a later amended comment nonetheless affect a quality decision? |
4. Example: Single weighing and sample identity
The approved source record for sample P-017 gives the weighed amount as “2.500 mg”, using the German thousands separator. This means 2,500 mg. The target expects grams. Divide 2,500 mg by 1,000 to obtain 2.5 g, recorded as 2.500 g to the stated precision. The source format is an explicit assumption of this example.
The evaluation of results may only commence in the intended model after identity and weight have been verified. The table distinguishes correct conversions from intentionally embedded errors. In a real-world project, source format, display, storage, and additional calculations would need to be separately defined.
| Case | Input in target | Predefined expected result |
|---|---|---|
| Correct | P-017; 2.500 g | Source and destination values are consistent after unit conversion. |
| Order-of-magnitude error | P-017; 2500 g | Not accept as valid transfer |
| Plausible false value | P-017; 2.450 g | Detect a discrepancy from the source value even if the value looks plausible |
| Identity error | P-071; 2.500 g | Detect incorrect sample; correct number is insufficient |
| Unverified status | P-017; 2.500 g without control documentation | Do not start the planned downstream analysis yet |
5. Schedule the check before the data affects a decision
The practical question is: When could a wrong entry cause harm or lead to an incorrect decision? Set the control point accordingly. If a number immediately triggers a critical calculation, an audit can only occur during the subsequent monthly review—potentially too late.
This does not mean that Annex 11 requires the same technical lock for every input. You must evaluate the process including permissible intermediate states. When data is temporarily stored, it should be clearly evident that the information has not yet been validated for its intended use. Define who may authorize actions in the absence of a reviewer or in the event of electronic control failure, and specify which activities must be suspended until such time.
6. Make corrections and re-evaluations traceable
A known error requires a controlled correction pathway. The relationship between the original input, the correction, and the subsequent re-evaluation result must be clearly traceable. For electronic modifications, risk-based audit trail controls must be applied; for paper records, the original information must remain identifiable in accordance with Chapter 4.
In addition, assess downstream impacts. A previously entered incorrect weighing value may influence calculated results. Simply correcting the number without reviewing the affected analyses does not resolve the issue. Whether an incident constitutes a critical deviation and whether further root cause investigations are required shall be determined upon evaluation; not every typing error automatically qualifies as a critical deviation.
7. Assess efficacy using real error types
A successful run with the correct input demonstrates only part of the system's capability. Use targeted erroneous inputs that the procedure is designed to detect. Examples include a plausible deviation and a swapped sample identifier. The expected result is determined prior to execution from an independent source.
During personnel audits, verify that the comparison actually encompasses both source and target. For electronic systems, assess their reference data, configuration, and error response behavior. A check mark without a clear and traceable object leaves unclear what was evaluated. The scope and documentation should be aligned with the specific risk, not with a generic minimum number of test cases.
8. Prepare a feasible input control prior to implementation
The department identifies the data source and potential root causes of errors, IT defines electronic functionalities, and QA evaluates integration into the quality process. A concise decision for each critical input scenario is generally more helpful than a general instruction to document everything in principle.
Use the checklist for a specific work location. If a question remains unanswered, identify the gap and establish an appropriate action plan. The independence of a theoretical audit can be compromised if, in practice, no readable source, no time, or no qualified representative is available.
- Critical data item and potential consequences identified.
- Valid source, identity, format, and unit defined.
- Additional control based on captured error types justified.
- The time prior to the relevant use is determined.
- Correction and re-evaluation including subsequent impacts are governed.
- Appropriate fault cases and independent target prepared.
Project experience and collaboration
Design the input control around the actual workstation
We start where a critical number or identifier is actually transcribed. The process owner explains its source and downstream use. We follow a plausible error to its potential consequence and assess whether a second person can make a reliable comparison or an electronic check would be more suitable. We prepare failure cases and evidence for the chosen control. A clearer work instruction may suffice. A confusing entry screen may require changes to the workflow.
Frequent Questions on Manual Data Entry Control
Must every manual entry be verified by two individuals?
Section 6 refers to critical data and permits the use of validated electronic means in the presence of a second person. Therefore, evaluate the potential error sequence first. A general field-by-field double check does not constitute an accurate representation of this passage.
Sources: [1]
Is a departmental review already sufficient?
Only if it adequately covers the relevant error risks in the specific context. A value within the allowed range may still be incorrect or incorrectly assigned to another sample. Describe the captured error class and, if necessary, supplement it with an appropriate additional control.
Is it sufficient to enter the same number twice?
Double entry alone does not constitute an independent accuracy check. The same incorrect source or the same assignment error can be accepted twice. Verify what the actual double entry actually detects and how identity, unity, and correct source are additionally safeguarded.
Sources: [1]
Can a barcode replace manual control?
An appropriate automated identity capture can reduce certain transfer errors. However, it must be demonstrated that the correct code is linked to the correct object and is processed accurately. A barcode alone neither validates the numeric value nor the integrity of the underlying dataset.
Must the unit be verified during input validation?
Can unverified entries be temporarily stored?
A provisional state is not universally prohibited. What is essential are identifiable statuses, protection measures, and authorized downstream usage. If an unverified input could already trigger a critical decision, a robust procedure must be in place to prevent such usage until the required control is performed.
What does the second person document?
The evidence should clearly indicate which input was verified against which source and resulting output. The timing and assignment to the reviewing personnel should be included in the appropriate documentation process. A general summary heading lacking a clear subject may leave significant questions unanswered.
Does every identified keyboard error require a root cause analysis?
Incidents must be appropriately reported and evaluated. Annex 11 § 13 requires root cause identification for critical incidents. Assess impact, recurrence, and control failure, rather than automatically equating each corrected spelling error or routinely excluding it.
Sources: [1]
What do we do when the electronic monitoring fails?
Prior to proceeding, clearly define which activities are suspended and whether an appropriate alternative procedure exists. A second person can only assist if they possess the required source and qualification and if the procedure remains under control. An improvised workaround does not substitute for a failed control.
Sources: [1]
How do we assess an existing instruction without major reconstruction?
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.
- EU GMP Annex 11: Computerised SystemsRevision January 2011; Effective date 30.06.2011; Access date 07.10.2026. Citations: §§1–2, 4.7, 6, 9–10, 12–13, 16
- EU GMP Chapter 4: DocumentationRevision January 2011; Effective date 30.06.2011; Access date 07.10.2026. References: §§4.1, 4.8–4.9
- PIC/S PI 041-1: Good Practices for Data Management and IntegrityPI 041-1, 01.07.2021; Access date 07.10.2026. Citations: §§9.1.5–9.1.6; §9.4; §9.5; §9.7
- 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 medicines / Annex 11
- 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: .
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.
- Computer system validation →
Bring your GxP systems into validated operation, from laboratory software, MES and eQMS to ERP. We define the scope, plan risk-based tests using GAMP 5 and hand the evidence to QA for release.

