Training Effectiveness Beyond the Read-and-Understand Signature

What a Read-and-Understand Signature Actually Records

Consider what the electronic record contains after an operator completes a read-and-understand assignment. It contains a user identifier, a document number, a document version, a timestamp, and in a validated system an audit trail entry showing that the assignment moved from open to complete. Nothing in that record describes performance. Nothing in it describes the task. The record is an acknowledgment of receipt.

That is a real control and it should not be dismissed. Document control depends on knowing who has seen which version, and revision management depends on being able to push a changed procedure to everyone affected by it. Read-and-understand is the mechanism that makes those things work. The problem is not that sites use it. The problem is that many sites use it as the whole training program and then present it as proof of competence when a regulator asks a different question.

The gap between acknowledgment and capability

A person who has read a sixteen-page aseptic gowning procedure has acknowledged a document. Whether that person can gown without touching the outer surface of a sterile sleeve is a separate question that the acknowledgment does not answer. A person who has read the deviation procedure has acknowledged a document. Whether that person can distinguish a critical from a major classification on a real event is a separate question. In both cases the site holds a record of the first thing and is being asked about the second.

Practitioners who work in this area have been direct about how the acknowledgment behaves in practice. One widely read industry piece describes the pattern in which the volume of assigned reading itself becomes the message, with a consultant reporting that a global biopharmaceutical organization assigned more than one hundred procedures within the first four to six weeks of onboarding, at least half of which had nothing to do with the job in question.¹ When a program is built that way, the completion metric measures throughput, not learning.

Why the habit persists

Three forces keep the read-and-understand model in place, and none of them is regulatory.

The first is that it is easy to administer and easy to report. A percentage-complete number for training assignments is one of the simplest metrics a quality system can produce, and it appears on management review slides because it is available, not because it is informative.

The second is defensive documentation. A site that assigns every procedure to everyone can never be told it failed to train someone on a document. The overassignment is a hedge against a finding, and it produces a training matrix in which the signal is buried under noise.

The third is system inertia. Most learning management systems were configured years ago around document assignment as the core object. Adding observed performance, practical assessment, or task qualification to that structure looks like a system replacement, so it does not get proposed. In practice it usually is not a replacement, which is the subject of a later section.

What the Regulations Actually Say, and What They Do Not

Before designing a better model it helps to read the source texts carefully. They are shorter than most people expect and they say specific things.

21 CFR 211.25: enable the person to perform

The United States requirement for personnel qualifications in finished pharmaceutical manufacturing is three paragraphs long. Paragraph (a) states that each person engaged in the manufacture, processing, packing, or holding of a drug product “shall have education, training, and experience, or any combination thereof, to enable that person to perform the assigned functions.” It continues: “Training shall be in the particular operations that the employee performs and in current good manufacturing practice (including the current good manufacturing practice regulations in this chapter and written procedures required by these regulations) as they relate to the employee’s functions.” The third sentence adds that current good manufacturing practice training “shall be conducted by qualified individuals on a continuing basis and with sufficient frequency to assure that employees remain familiar with CGMP requirements applicable to them.”³

Four obligations are visible in that text, and they are not the same obligation.

• An outcome standard. The combination of education, training, and experience must be sufficient to enable performance of the assigned functions. The regulation is written in terms of capability, not attendance.
• Task specificity. Training must be in the particular operations the employee performs. A general procedure library assigned to a whole department does not meet this on its own.
• A qualified trainer. Current good manufacturing practice training must be conducted by qualified individuals. Most sites can name the trainer. Fewer can produce the basis on which that trainer was qualified.
• A frequency judgment. Continuing basis and sufficient frequency are both undefined in the regulation, which means the site owns the rationale.

Paragraph (b) applies the same education, training, and experience standard to supervisors, tied to providing assurance that the product has the safety, identity, strength, quality, and purity it purports to possess. Paragraph (c) requires an adequate number of qualified personnel.³

EU GMP Chapter 2: practical effectiveness, periodically assessed

The European requirement is the one that says what the industry usually attributes to the United States regulation. EudraLex Volume 4, Part 1, Chapter 2 on Personnel came into operation on 16 February 2014 and was revised to integrate the principles of ICH Q10.⁴

Paragraph 2.10 sets the scope: the manufacturer should provide training for all personnel whose duties take them into production and storage areas or into control laboratories, including technical, maintenance, and cleaning personnel, and for other personnel whose activities could affect product quality.⁴

Paragraph 2.11 is the sentence that matters here. Newly recruited personnel should receive training appropriate to the duties assigned to them, and then: “Continuing training should also be given, and its practical effectiveness should be periodically assessed.”⁴ Two words carry the weight. Practical means the assessment addresses the doing, not the reading. Periodically means it recurs, so a one-time qualification at hire does not discharge the expectation.

Paragraph 2.14 adds a point that is easy to miss: the pharmaceutical quality system and all the measures capable of improving its understanding and implementation should be fully discussed during training sessions.⁴ The word discussed does not describe a document assignment.

The World Health Organization guidance on good manufacturing practices for pharmaceutical products carries the same expectation in nearly identical language. Section 10.2 of the main principles annex states that continuing training should be given and its practical effectiveness periodically assessed, with approved programs available and training records kept.⁵ For a company operating across markets, this is one of the few places where the United States and the rest of the world genuinely diverge in wording, and the practical answer is to build to the higher expectation.

ICH Q10: training as a management review output, not an effectiveness clause

ICH Q10 is frequently cited as the source of a training effectiveness requirement. It is not. Reading the guideline as endorsed at Step 4 in June 2008, training appears in two places, and neither creates an assessment obligation.

Under Quality Planning, section 2.3(d) states that management should provide the appropriate resources and training to achieve the quality objectives. Under Management Review, the guidance lists actions the review system should identify, including “Provision, training, and/or realignment of resources,” and lists among the outcomes of management review the “Allocation or reallocation of resources and/or personnel training.”⁶

That placement is more interesting than a requirement would have been. Q10 treats training as a response to what performance monitoring reveals. In the Q10 model, a rising deviation rate on a particular task is an input, and a change to training is one of the permitted outputs. That is a closed loop, and it is the loop most training programs are missing. The program that only tracks completion has no input signal at all, so it has nothing to respond to.

Annex 1: where the expectation becomes explicit and dated

For sterile products the picture is far more prescriptive. Annex 1 of the EU GMP guidelines, published on 22 August 2022 and in operation from 25 August 2023 for all sections other than the lyophilizer requirement, converts training expectations into specific, auditable events.⁷

Paragraph 7.3 requires that all personnel including those performing cleaning, maintenance, and monitoring, and those who access cleanrooms, “should receive regular training, gowning qualification and assessment in disciplines relevant to the correct manufacture of sterile products,” with the training including basic microbiology and hygiene and a focus on cleanroom practices, contamination control, and aseptic technique.⁷

Paragraph 7.4 is the one that removes all ambiguity about frequency. Personnel accessing grade A and B areas should be trained for aseptic gowning and aseptic technique, and compliance with the gowning procedures “should be confirmed by assessment and periodic reassessment at least annually,” involving both visual and microbial assessment using locations such as gloved fingers, forearms, chest, and hood.⁷

Paragraph 7.5 states that unqualified personnel should not enter grade B cleanrooms or grade A in operation, and where that is unavoidable a written procedure and supervision by an authorized person are required. Paragraph 7.6 requires systems for the disqualification of personnel, based on ongoing assessment, an adverse trend from personnel monitoring, or involvement in a failed aseptic process simulation. Once disqualified, retraining and requalification must be completed before the operator returns to aseptic practice, and for grade B or grade A intervention that requalification should include participation in a successful aseptic process simulation.⁷

Five Kinds of Evidence That Demonstrate Effectiveness

If a signature does not demonstrate capability, what does? Five categories of evidence do the work, and a mature program uses more than one for any task it considers critical.

Why the observation has to be structured

Observation is only evidence if two observers watching the same performance would reach the same conclusion. That requires the standard to be written down before the observation, in behavioral terms, with the failure conditions named. “Demonstrated good aseptic technique” is not a standard. “Did not break first air over an open container at any point during the intervention” is a standard, because it can be scored the same way twice.

The most common failure in observation programs is that the checklist is a restatement of the procedure’s section headings. Every line reads as a step and every line is scored as done. That produces a second document acknowledgment wearing a different name. A useful checklist is written from the failure modes: the things people actually get wrong on this task, drawn from deviation history, aseptic process simulation results, and the observations of the people who train it.

What the evaluation literature says about where programs stop

The four-level evaluation framework introduced by Donald Kirkpatrick in 1959 remains the common vocabulary for this work: reaction, learning, behavior, and results. A 2025 integrative review in Revista Latino-Americana de Enfermagem examined thirteen studies applying the framework to nurse training and found that six evaluated all four levels while the rest used partial combinations, with the results level identified as the most difficult to reach in practice.⁸ That difficulty is not unique to nursing. In pharmaceutical manufacturing the equivalent pattern is a program that measures completion and satisfaction, occasionally measures knowledge, and rarely measures behavior on the job at all.

The practical response is not to build a four-level evaluation for every procedure in the library. It is to decide which tasks warrant level three evidence, which is observed behavior, and to accept level one and level two for the rest. That decision is a risk decision, and it belongs in a documented rationale rather than in whatever the learning system happens to support.

A Worked Example: Visual Inspection of Filled Containers

Abstract principles are easy to agree with. Here is one critical task carried through from the regulatory basis to the record.

Visual inspection of filled parenteral containers is a good example because it is unambiguously critical, because the regulator has said specific things about it, and because it is one of the tasks most often found deficient. Annex 1 paragraph 8.30 requires that all filled containers of parenteral products be inspected individually, that defect classification and criticality be determined during qualification and based on risk and historical knowledge, that different defect types be categorized and batch performance analyzed, and that “a defect library should be generated and maintained which captures all known classes of defects.” That library, the text continues, “should be used for the training of production and quality assurance personnel.”⁷

Two recent warning letters show what happens when the qualification is thin. In a July 2025 letter to a sterile manufacturer, FDA noted that the firm lacked a procedure for inspector training and raised the composition of the challenge kit used for qualification. The same letter recorded that operators were not required to participate in an aseptic process simulation before working in the aseptic processing operation, and that there was no written procedure for gowning qualification.⁹ In a May 2025 letter, FDA cited 21 CFR 211.25(a) against an outsourcing facility and found the visual inspection training program inadequate, including for personnel who had passed qualification but did not follow the established inspection timeframes.¹⁰

The task qualification, built out

Notice what changed and what did not. The read-and-understand assignments on the two procedures are still there. They did not go away, and removing them would break document revision management. What changed is that they are now prerequisites to a qualification rather than the qualification itself, and the site can answer a competence question with a competence record.

An Assessment Design Rubric

Assessment design is where good intentions turn into a question bank that tests reading comprehension. The rubric below is deliberately short enough to apply to a real item set in an afternoon. Score each proposed assessment item against all six criteria. An item that fails any of the first three should be rewritten or removed.

Two design rules that prevent most of the damage

Separate critical items from the aggregate score. If an assessment covers a task where one class of error carries patient risk, that class should be scored on its own with a pass threshold of its own. An eighty percent aggregate that permits a missed critical defect is not a qualification, it is an average.

Write the observation checklist before the training material. Deciding what the person must be observed doing forces clarity about the standard, and the training content then has an obvious target. Programs built in the other order tend to produce content first and then a checklist that summarizes the content, which measures delivery rather than capability.

Changing the Record Model Without Replacing the LMS

The objection to everything above is usually the same: our learning management system only handles document assignments, and replacing it is a two-year program with a validation effort attached. That objection is understandable and mostly wrong. Very little of what this article describes requires a new platform. What it requires is a different set of objects, and most of those objects can live in systems that already exist and are already validated.

What can be done inside the existing system

Nearly every validated learning management system in life sciences already supports the following, whether or not the site has configured them.

• Curricula tied to job roles rather than departments. Most systems support role-based curricula. Many sites assign at department level because it was faster to configure. Moving to role-based assignment is a configuration change, not a system change, and it is the single highest-value cleanup available.
• Assessment objects with pass criteria. Online assessment with a configurable passing score is standard functionality. The work is writing better items, not enabling the feature.
• A qualification or certification object with an expiry date. Many systems have this and it goes unused. Where it exists, it gives you the qualification state described earlier without any new software.
• Prerequisites and sequencing. Making the document acknowledgment a prerequisite to the assessment, and the assessment a prerequisite to the observed performance sign-off, enforces the order without a workflow build.
• Instructor-led or on-the-job training records with an assessor signature. Standard functionality in most platforms. The observation checklist can be attached as the completion evidence.

What belongs outside the learning system

Some things should not be pushed into the learning management system even when it can technically hold them.

Retiring redundant read-and-understand assignments

This is the part that frees the capacity to do everything else, and it is the part sites are most nervous about. The nervousness is understandable, because removing an assignment feels like removing a control. Handled properly it removes noise and improves the control.

Work through it in this order.

Handling the transition without opening a compliance gap

Four rules keep the transition clean.

Never leave a person unqualified during a change. Existing training records remain valid until the new qualification is issued. A cutover that invalidates prior records on a date, without a bridging assessment, creates a population of people performing tasks with no current qualification on file. Grandfather existing performers explicitly, with a documented basis such as a satisfactory observation within a defined lookback window, rather than implicitly.

Change one role at a time. A pilot on a single role, taken through a full requalification cycle, produces the evidence that the model works and surfaces the configuration problems while they are easy to fix. Enterprise-wide cutovers in training systems have a poor record for exactly the reason they have a poor record everywhere: the exceptions are invisible until the volume is high.

Keep the acknowledgment record where the document control depends on it. Revision management still needs to know who has seen the current version. The change described here is about what the acknowledgment is used to prove, not about deleting the mechanism.

Write the new expectation into the training procedure before the first qualification is issued. If the site’s training procedure still describes read-and-understand as the qualification method, the first qualification record issued under the new model is a departure from procedure. Revise the governing procedure first, then issue.

What an Inspector Asks When They Doubt Training Effectiveness

Investigators rarely open with a training question. Training comes up when something else has already gone wrong, and the shape of the questioning follows a predictable path once it does.

The inspection program tells you where to look

FDA compliance program 7356.002, Drug Manufacturing Inspections, issued 16 September 2022 with an implementation date of 17 October 2022, structures a drug inspection around six systems: quality, facilities and equipment, materials, production, packaging and labeling, and laboratory control. Training is not a seventh system. The program states that the organization and personnel, including appropriate qualifications and training, “will be evaluated as part of that system’s operation.”¹²

Two consequences follow. First, training and qualification of personnel appears as a listed coverage area under the quality system and under each of the operational systems, so a site is examined on training four or five times in a single inspection, each time in the context of a specific operation. Second, and more consequential for how you prepare, each system section of the program carries the same instruction: the establishment’s adherence to written procedures “should be verified through observation whenever possible.”¹²

The sequence of questions

Once an investigator has doubt about competence on a specific task, the questions generally run in this order. Each one narrows.

1. Who performed this step? Traceability from the record to the person. Straightforward for most sites.
2. Show me their training record for this task. The point at which many sites produce a list of acknowledged documents and the conversation changes tone.
3. How did you determine they could perform it? The capability question from 211.25(a). An acknowledgment does not answer it.
4. Who qualified them, and how was that person qualified? The qualified individuals clause. Frequently unanswerable.
5. When did you last observe them doing it? The question the compliance program sets up. If the answer is at initial training three years ago, the follow-up is predictable.
6. Show me someone who failed. A qualification system with no failures in its history is a system whose criteria do not discriminate. This question is asked more often than sites expect.
7. How do you know the training worked? The Chapter 2 question in plain form. It is answered with trended data, not with a completion percentage.

What the public letters show

The enforcement record is useful here because it shows how training findings actually appear, and they rarely appear as a standalone training citation.

An August 2026 warning letter to a sterile manufacturer illustrates the pattern well. FDA recorded multiple instances of inappropriate aseptic technique in which operators blocked unidirectional airflow at critical locations, and noted that the firm’s own procedures did not adequately define acceptable hand positioning, intervention speed, or first-air protection requirements. The response was found insufficient on the oversight of operator practice.¹³ No training regulation is cited in that observation. The finding is about aseptic process control. But the root of it is a task standard that was never defined precisely enough to train against or observe against.

That is the general shape. Training deficiencies surface as production findings, laboratory findings, and investigation findings. A review of FDA training-related findings across 2025 identified letters citing 21 CFR 211.25(a) against firms whose quality units had failed to establish an adequate training program, whose operators had not received adequate current good manufacturing practice training for the products being made, and whose visual inspection and analyst qualification arrangements did not hold up.¹⁴ The pattern is consistent across firm size and product type.

The observation is older than it looks. An analysis of a full year of FDA warning letters, published in 2011, found procedure management and training among the recurring problems, with citations recording that there were no records showing that current good manufacturing practice training specific to employees’ duties had been performed, and in one case that senior staff were unaware of an unsigned procedure titled Training that required current good manufacturing practice training of employees.¹⁵ Fifteen years later the same finding categories recur, which suggests the problem is structural rather than a matter of individual sites being careless.

Preparing the answer before the question

A site that wants to answer the seven questions above needs four artifacts, and none of them is exotic.

• A list of critical tasks with a written rationale for why each is on the list and why others are not. This is a risk document and it should be reviewed on a cycle.
• A qualification record per critical task per person, with basis, issue date, expiry, and assessor.
• An observation history showing periodic observed performance during routine operations, with results, including the ones that were not satisfactory and what followed.
• A periodic effectiveness review that compares qualification status against task-attributed events and observation results, reaches a conclusion, and feeds management review as ICH Q10 contemplates.⁶

Conclusion

The read-and-understand signature is not the problem. It is a document control mechanism doing its job, asked to carry a burden of proof it was never built for. The mistake is structural: a site adopted an acknowledgment record as the answer to a capability question, then built a decade of reporting on top of it, and now the reporting is the reason the model is hard to change. Regulators have been clear enough about the difference. The United States regulation is written in terms of enabling a person to perform assigned functions, in the particular operations they perform. The European and World Health Organization texts add that the practical effectiveness of continuing training should be periodically assessed. Annex 1 goes furthest and defines a qualification state with an expiry, a reassessment interval, and a mechanism for disqualification. None of these describes a signature.

What we see in practice is that the change is smaller than it looks. The work is to name the tasks that warrant real qualification, define what good performance is in terms specific enough to observe, remove the assignments that were only ever defensive, and close the loop by feeding what you observe back into the program. Most of it can be done in the systems a site already runs and has already validated. The two-year platform replacement is usually a way of not starting. Sakara Digital works with pharma and biotech organizations rebuilding this kind of qualification model inside their existing quality and learning systems. If you are looking at a training program that reports ninety-eight percent complete and cannot tell you whether anyone can do the work, we are happy to have that conversation.

For Further Reading