LMS Consulting for GxP Training: Fixing Curricula Before You Change Platforms

Why a New LMS Rarely Fixes Training Compliance

The business case for a new LMS usually reads the same way. The current system is old, hard to report from, and disliked by employees. Completion rates are low. An inspector asked for a list of everyone qualified on a procedure on a specific date, and it took two days to produce. The proposed fix is a modern, cloud-based platform with better dashboards and automatic reminders.

Those benefits are real, but they address how training is delivered and reported. They do not address what employees are assigned, why, or whether the assignment shows they can do the job. If a production operator has 240 documents in their curriculum and 90 of them have nothing to do with their tasks, a faster system will remind them more efficiently to read documents they do not need. The overdue rate will stay high because the assignment load is the problem, not the reminder engine.

What the Inspection Data Shows

FDA publishes the frequency of each citation on system-generated Form 483s every fiscal year.¹ For fiscal year 2025 (inspections ending between October 1, 2024 and September 30, 2025), the drugs data set covers 713 Form 483s. Three citation types under 21 CFR 211.25(a) were used a combined 67 times: personnel lacking the education, training, or experience for their functions (28), employees not trained in their particular operations, CGMP, or required written procedures (25), and GMP training not conducted often enough (14). The most frequent drug citation overall was 211.22(d), procedures not in writing or not fully followed, at 243. The year before, the FY2024 drugs data set showed the same three 211.25(a) citation types used 60 times across 561 Form 483s.

Two cautions on reading those numbers. FDA notes that the spreadsheet does not include Form 483s prepared by hand, so the counts are a floor, not a total. And a citation count is not a count of firms. Even so, the pattern is useful. Training appears as its own citation, but the much larger number is procedures not being followed, and a training program that assigns the wrong documents to the wrong people contributes to both.

The regulation itself is specific about scope. Section 211.25(a) says that training “shall be in the particular operations that the employee performs” and in CGMP as it relates to the employee’s functions.² That is a requirement about fit between the person, the task, and the training. No LMS feature satisfies it on its own. It is satisfied, or not, by the curriculum design.

What Moves to the New System Unchanged

The UK MHRA’s 2016 GMP inspection deficiency report gives a set of training examples that read like a list of curriculum problems rather than software problems.³ Inspectors found an operator signed off as competent for manufacturing solutions and suspensions who had not completed training in all the associated duties, such as cleaning the compounding equipment. They found an analyst qualified on one chromatography system who was then treated as qualified on two others, with no check on the differences between them. And they found a site where, in the report’s words, “There was no system for confirming that all personnel that required training had been trained.”

Each of those findings would survive a platform change. The missing duty, the assumed equivalence between systems, and the missing link between role and requirement all live in the curriculum design. Migrate them as they are and the new LMS will report the same gaps with better formatting.

The point is not new to people who build these systems. A 2023 PharmTech article on GxP training matrices quotes John Constantine of Orchestrall: “Many companies try to fit the old way of doing things into their digital LMS, which is not the right way to go.”⁴ The rest of this article is about doing the opposite: fixing the old way first.

Problem 1: Role-to-Curriculum Mapping That No Longer Matches the Work

Most LMS curricula are attached to job titles or HR job codes. That made sense when the curriculum was first built: someone in quality sat with each department head, listed the documents each title needed, and loaded the result. The trouble is that titles and tasks drift apart. A “Manufacturing Associate II” at one site runs the granulator; at another site the same title runs packaging. A new piece of equipment arrives and three people are trained on it, but the curriculum for the title is never updated, so the next person hired into that title is not assigned the training.

How the Mapping Drifts

Mapping drift has a few common sources, and each one is visible in the data if you look:

• Titles that cover different work. One HR job code maps to several sets of tasks across sites, shifts, or product lines, so the curriculum is either the union of all of them (too much) or one of them (too little).
• Individual add-ons that never became rules. A supervisor asks the LMS administrator to add three documents to one person. Two years later, a dozen people have hand-added assignments and nobody can say why.
• Equipment and method qualifications held outside the curriculum. The MHRA chromatography example is typical.³ Qualification on a specific instrument, room, or method is tracked in a spreadsheet or a logbook, so the LMS shows a person as fully trained when the qualification that matters is somewhere else.
• Transfers that add but never remove. When someone moves departments, the new curriculum is assigned but the old one stays, so long-tenured employees carry every curriculum they ever held.

The Fix: Map Tasks First, Then Roles

EU GMP Chapter 2 asks that newly recruited personnel receive “training appropriate to the duties assigned to them.”⁵ The word that matters is duties. The more reliable way to build a curriculum is to start from the tasks people perform, link each task to the documents and qualifications it requires, and then define roles as bundles of tasks. A person’s curriculum becomes the sum of the tasks they are assigned, not the sum of everything their title ever touched.

In practice this means building three short lists with the department heads, not one long one:

1. A task list. Concrete activities such as “dispense raw materials,” “perform line clearance on Line 3,” or “review HPLC data in the chromatography data system.” Each task has an owner.
2. A task-to-requirement map. For each task, the procedures, work instructions, and qualifications needed to perform it, with the type of evidence each needs (covered under Problem 4).
3. A role-to-task map. For each role, the tasks it performs. Variants by site or line become separate roles rather than one role with everything attached.

The training research literature points the same way. A 2017 review of a century of training research in the Journal of Applied Psychology describes training needs analysis, the step of working out what people need to learn for their jobs, as widely considered one of the most important steps in building training, and notes that it has received little research attention.⁶ Task-first mapping is a needs analysis for a GxP curriculum. It is the step most often skipped when a curriculum is loaded into a new system under deadline.

A Test You Can Run on Your Current Data

Export the current assignments from your LMS and pick two roles in one department. For each person in those roles, ask the supervisor to mark which assigned documents the person uses in their work at least once a year. Then count:

• The share of assigned documents the supervisor marked as not used by that person.
• The number of documents that differ between two people with the same title.
• The number of tasks the supervisor says the person performs that have no matching document or qualification in the curriculum.

The third count is the one that matters most for compliance, because it shows people doing work they have no recorded training for. The first two show how much load can be removed. Running this test on two roles takes a few days and gives you real numbers for the business case, rather than an estimate.

Problem 2: Read-and-Understand Assignments That Have Grown Past Their Purpose

A read-and-understand assignment asks an employee to open a document, read it, and sign that they have read and understood it. It is the quickest form of training to deliver and the easiest to record, which is exactly why it grows. Every new document gets one. Every revision gets one. Nobody removes them.

How Assignments Pile Up

Two industry practitioners writing in Pharmaceutical Online in 2022 described the result plainly. One recounted that a global biopharma client “assigned over 100 SOPs for me to complete within my first four to six weeks there,” and added that “at least half of those assignments had nothing whatsoever to do with my job there.”⁷ They also noted that the time documents are open for reading is, in some cases, a matter of seconds. When the load is that high, employees learn that the goal is to clear the list.

A 2023 BioProcess International article on SOP design makes the same point from the document side. The author describes a training program that included “more than 150 good practice (GxP) SOPs, several of them longer than 50 pages,” and warns that “SOP training requires time and can quickly overburden employees.”⁸ Her recommendation is direct: only employees who perform the activities a procedure describes should be trained on it.

Read-and-understand is not wrong in itself. For a policy that everyone must know exists, or a minor change to a document a person already knows well, an acknowledgment is a reasonable record. The problem is using it as the default for everything, including tasks where the only acceptable evidence is that the person can perform them.

Sorting Assignments Into Training Types

Before migration, sort every document in every curriculum into one of four groups. This is the single most useful piece of curriculum work, because it reduces the number of assignments and tells you which ones need stronger evidence.

The groups are a starting point, not a rule set. Each company will draw the lines differently, and QA should approve the criteria. What matters is that the classification exists and is recorded, so that the new LMS can be configured to treat each group differently instead of treating every assignment as a read-and-understand.

Retiring Assignments Without Opening a Gap

Removing assignments makes some quality leaders nervous, and the concern is fair. If an inspector asks why an operator is no longer trained on a procedure, the answer needs to be on paper. Three practices help:

• Document the rationale per curriculum, not per person. A short curriculum review record, approved by the function head and QA, listing what was removed and why.
• Keep historical completions. Removing a future assignment does not delete past training records. The person’s history still shows they were trained.
• Remove before you migrate. Every assignment removed in the old system is one fewer record to map, test, and reconcile in the new one.

An earlier Sakara Digital article, Training Effectiveness Beyond the Read-and-Understand Signature, covers how to design the stronger evidence for the knowledge and task qualification groups. This article stays with the curriculum decision: which documents belong in which group, and how that decision shapes the platform.

Problem 3: SOP Revisions That Retrain Everyone

In many LMS configurations, a new version of a document automatically creates a new assignment for everyone who had the previous version. It is simple to configure and easy to defend. It is also often one of the largest sources of training load, because a busy document set produces a steady stream of revisions, and many of them change nothing about how the work is done.

Why Every Revision Triggers a Full Retrain

The cause is usually a missing step in document change control. The change request records what changed in the document but not what the change means for the people who use it. Without that information, the safest default is to retrain everyone. A typo fix, an updated reference to another document, a new logo, and a changed critical step all create the same assignment for the same audience.

The BioProcess International article notes that the right training approach depends on, among other things, “whether the document is new or only updated.”⁸ That distinction is what most change control forms do not capture.

Classify the Change, Then Target the Audience

The fix belongs in the document change control procedure, not in the LMS. Add two fields to the change request, completed by the document author and approved by QA:

1. A training impact class. For example: no training impact (editorial changes, formatting, reference updates); awareness (clarifications that do not change how a step is done); retraining (changed steps, limits, or decisions); new qualification (a new task, piece of equipment, or method).
2. An affected audience. The roles or tasks that the changed sections affect, which may be a subset of everyone who uses the document.

With those two fields, the training assignment can match the change. An editorial revision triggers no assignment. A changed critical step in section 7 triggers retraining for the operators who perform that step, with the changed text highlighted, and perhaps an observed check if the step is high risk. The other 80 people who use sections 1 through 6 get nothing, or a short notice.

Train Before the Effective Date

Retraining everyone has a second effect that is easy to miss. When a revision goes to hundreds of people, the training window has to be long, so the document often becomes effective before everyone who performs the changed step has been trained. That is a direct gap against 211.25(a), which requires training in the written procedures as they relate to the employee’s functions.² The BioProcess author states the expectation plainly: involved employees should be trained on SOPs, work instructions, and annexes before they become effective.⁸

Targeting the audience makes that achievable. Thirty people can be trained in a week. Three hundred cannot. Once the audience is right, set the effective date after the training is complete, and use the LMS to report anyone in the affected audience who is not yet trained on the effective date. Where a manufacturing execution system or laboratory system can check qualification status before allowing someone to perform a step, that check is worth considering for high-risk tasks.

Retraining as the Default CAPA

A related source of curriculum growth is the CAPA that assigns retraining by default. A deviation occurs, the investigation is short, the root cause is recorded as “human error,” and the corrective action is to retrain the person, or the whole shift, on the procedure. The retraining becomes a new assignment, and over time the curriculum fills with repeat assignments that address no real cause.

The curriculum fix is a rule in the CAPA procedure: retraining is an acceptable action only when the investigation shows a training gap, and a retraining CAPA must name the specific content and the specific people. A second rule helps keep the curriculum clean: CAPA retraining is assigned as a one-time item, not added permanently to the role’s curriculum, unless the document itself changed.

FDA also treats training records as evidence that a corrected procedure is in use. In a May 2026 warning letter, the agency found a response inadequate in part because “You do not provide updated procedures and training records to demonstrate you have corrected your data control deficiencies.”¹⁰ That expectation is another reason to target retraining well. When a procedure is corrected, the records for the people who perform the changed work should be complete and easy to produce, not mixed in with hundreds of unrelated assignments.

Problem 4: Effectiveness Checks Missing From the Curriculum

EU GMP Chapter 2 says continuing training should be given and “its practical effectiveness should be periodically assessed.”⁵ Section 211.25(a) requires CGMP training “with sufficient frequency to assure that employees remain familiar with CGMP requirements applicable to them.”² Neither says how. Most companies answer the question one course at a time, which is why the answer is so often a signature.

Decide at the Curriculum Level

The curriculum is the right place to decide what evidence each item needs, because that is where you can see the whole load for a role. If the sorting exercise under Problem 2 is done, the decision is mostly made: awareness items need an acknowledgment, knowledge items need a check with a pass mark, and task qualifications need observed performance. What remains is to record that decision as an attribute of each curriculum item so that the LMS can enforce it.

Two points from the training research matter for this decision. First, satisfaction surveys are not effectiveness evidence. The 2017 Journal of Applied Psychology review notes that trainee reactions are not necessarily a strong predictor of learning.⁶ A course rated highly by participants may still leave them unable to perform the task. Second, the MHRA examples show inspectors looking for exactly this gap, including an operator signed off as trained in raw material assessment on the basis of a read-and-understand questionnaire with no practical assessment.³

What This Means for the Platform

Effectiveness design has a direct effect on LMS requirements, and it is the area where platforms differ most. Before selection, confirm you know which of these your curricula will need:

• Knowledge checks with configurable pass marks, question pools, and limits on retakes.
• Observed performance records, signed by a named assessor who is themselves qualified, with the assessor’s qualification visible in the record.
• Qualification status that is separate from document completion, can expire, and can be suspended (for example, after a deviation) without deleting history.
• Periodic requalification on a schedule set per task, not one schedule for everything.

If the curricula do not yet use these record types, a platform that supports them may still be configured only for read-and-understand, because that is what was migrated. This is the clearest case where fixing the curriculum first changes what you buy and how you configure it.

Problem 5: Training Matrices Nobody Owns

The training matrix is the table, in a spreadsheet or in the LMS, that says which roles need which training. At many companies it has an administrator but no owner. The LMS administrator maintains it, changes it on request, and runs the reports. Nobody in the business is accountable for whether it is right.

Signs the Matrix Has No Owner

• Requests to change the matrix come by email to the LMS administrator, who makes the change without a recorded approval.
• Nobody can say when a role’s curriculum was last reviewed as a whole.
• The matrix in the LMS and the matrix in a department’s spreadsheet do not agree, and nobody knows which is current.
• New equipment, methods, or sites are added to operations months before they appear in the matrix.

The 2023 PharmTech article states the ownership principle simply: “The business function should own what an employee needs to be trained on, and the training department handles the execution.”⁴ EU GMP Chapter 2 adds an approval expectation: training programs should be available and approved by the head of Production or the head of Quality Control, as appropriate.⁵ An unowned matrix meets neither.

Split Ownership by Decision

Ownership works best when each decision has one owner. A practical split looks like this:

The LMS administrator still does a great deal of the work, but they configure decisions made by others. That separation matters at inspection, because the answer to “who decided this operator needed this training” is a named function head with an approval record, not “the system assigned it.”

A Short Governance Routine

Ownership needs a routine to stay real. A quarterly curriculum review per department, lasting an hour, covers four questions: what tasks were added or retired, which documents were added or retired, which assignments are most often overdue and why, and which deviations or CAPAs in the quarter point to a training gap. The output is an approved change list for the LMS administrator. The MHRA finding that a site had no system for confirming that everyone who required training had been trained is what an unowned matrix looks like from the outside.³

Scoping LMS Selection and Migration

Once the curricula are sound, the platform work becomes much better defined. The requirements come from the fixed curricula, the migration scope shrinks, and the integration design follows from the ownership model. This is also where outside LMS consulting for GxP training tends to help most, because most internal teams run an LMS selection rarely and a training history migration even less often.

Write Requirements From the Fixed Curricula

The most common selection mistake is to start from a vendor’s feature list or an analyst grid and score products against it. A better start is the curriculum design itself. Each decision made under the five problems implies a requirement:

The last row deserves emphasis. Inspectors ask about the past: who performed a step on a batch made last year, and were they qualified on the version in effect that day. Many systems report current status well and historical status poorly. Test it with your own data during the demo, not with the vendor’s sample records.

Content Standards and Portability

Most GxP training content is documents, but e-learning modules, videos, and assessments are growing. Ask how the platform handles packaged content and learning records. The Experience API (xAPI), now published as IEEE 9274.1.1-2023, defines a JSON data model and a web service for tracking and accessing learner experience data.¹¹ Support for open standards like this does not guarantee you can move records later, but it makes the exit easier to plan. Exit terms and record export are covered in more depth in our earlier article on QMS, DMS, and LMS integration.

Supplier Assessment and Validation Scope

Treat the LMS supplier as a GxP supplier. The training records it holds are evidence of compliance with 211.25 and EU Chapter 2, and its electronic signatures are relied on as signatures. Scope the supplier assessment, the quality agreement, and the validation effort around the functions that matter most: assignment rules, signature execution, qualification status, audit trail, and point-in-time reporting. Less critical functions, such as dashboards and notification templates, need less testing.

Training History Migration

Training history is the record set that most often goes wrong in an LMS migration. EU GMP Annex 11 says that when data are transferred to another format or system, “validation should include checks that data are not altered in value and/or meaning during this migration process.”¹² For training records, meaning is the hard part. A completion record that says “SOP-0412, completed March 3” has lost its meaning if it no longer shows which version was trained, by what method, who assessed it, and what the signature meant.

Scope the migration around the fields that carry meaning:

• Document identifier and version for each completion.
• Training type and evidence type (acknowledgment, knowledge check result, observed assessment).
• Trainer or assessor, where relevant, and their qualification at the time.
• Signature manifestation: printed name, date and time, and meaning of the signature.
• Qualification status history, including expirations, suspensions, and restorations.
• Waivers and exemptions, with the recorded rationale.
• Records for contractors and former employees, which are easy to drop and often needed.

Expect the new system to reject some legacy records. A 2013 Learning Guild article on LMS implementation warned that “your new LMS may enforce stricter data integrity rules than your legacy system does,” using the example of courses whose end dates came before their start dates.¹³ Decide how to handle each class of rejected record before the migration run, and document the decision, rather than fixing records one by one during cutover.

The same article suggested that legacy data could be archived and, if needed, retrieved by asking IT to restore a backup and run a report. That is reasonable for many industries. It is not enough for GxP training history, because inspectors ask about past qualification during the inspection, not weeks later. Annex 11 also says that archived data should be checked for accessibility, readability, and integrity, and that if the system changes, “the ability to retrieve the data should be ensured and tested.”¹²

That leaves two workable options. The first is to migrate full history into the new LMS as read-only historical records. The second is to migrate current qualification status and a defined recent window, and move older history to a validated, searchable archive with tested retrieval. Retention periods should follow the records the training supports, such as batch records and study records, not a general HR retention rule. Either option works if the point-in-time question can be answered quickly. The choice usually depends on how cleanly the legacy data map, which is one more reason to clean the curricula first: fewer active assignments means fewer records to map and test.

Part 11 Electronic Signatures

Training sign-offs in an LMS are electronic signatures, and the new system’s signature controls are part of what you validate. The relevant parts of 21 CFR Part 11 are short:

• Signature manifestation (11.50). Signed records must show the printed name of the signer, the date and time of signing, and the meaning of the signature, “such as review, approval, responsibility, or authorship.”¹⁴ In an LMS, the meaning is often vague. Define distinct meanings for an employee’s acknowledgment, an assessor’s sign-off, and a QA approval of a curriculum.
• Signature and record linking (11.70). Signatures must be linked to their records so they cannot be “excised, copied, or otherwise transferred to falsify an electronic record by ordinary means.”¹⁵
• General requirements and components (11.100 and 11.200). Each electronic signature must be unique to one individual and never reused or reassigned, the organization must verify identity before issuing one, and signatures not based on biometrics need at least two distinct components, such as an identification code and a password.¹⁶

Migration raises a specific question: what happens to historical signatures? They cannot be re-executed in the new system, and they should not appear as if they were. The usual approach is to migrate each signature as part of the historical record, carrying its manifestation fields (name, date and time, meaning) and the source system, and to keep the legacy audit trail in the archive so the original link between signature and record can be shown. Agree this approach with QA and record it in the migration plan before any data moves. If the new LMS uses single sign-on for signature execution, our earlier article on single sign-on and Part 11 covers how to meet the Part 11 component requirements with it.

Integration With the QMS and Document System

Our earlier article on QMS, DMS, and LMS integration covers whether the QMS, document system, and LMS should be one platform and how to design the interfaces between them. For curriculum purposes, three data flows matter most:

1. Document revision to training assignment. The document system passes the new version, the training impact class, and the affected audience. Without the last two fields, the integration simply automates the retrain-everyone problem.
2. CAPA to training. A retraining action created from a CAPA carries the CAPA reference, the content, and the named people, and closes back to the CAPA when complete.
3. Deviation investigation to training history. Investigators can see, for the people involved, what they were qualified on and when, without asking the LMS administrator for a report.

Integration With the HR System

The HR system is the source for who works at the company, in what job, at which site, and under which manager. Most assignment errors after go-live trace back to this feed. A standard such as SCIM, the IETF’s System for Cross-domain Identity Management protocol (RFC 7644), gives a common way to provision and update user identities across systems, including cloud services.¹⁷ Whatever the mechanism, scope these events explicitly:

• Joiners. A new hire’s role assignment triggers the right curriculum before their first GxP task, not after.
• Movers. A change of job, site, or department adds the new tasks and removes the old ones, keeping history. This is where curricula grow without limit if no one designs the removal.
• Leavers. Accounts are disabled and signatures can no longer be executed, but records are retained for the full retention period.
• People outside HR. Contractors, temporary staff, and service providers often are not in the HR system at all. They need a defined route into the LMS, and an owner.

The fragile part is not the interface; it is the mapping table from HR job codes to LMS roles. If roles are built from tasks, as described under Problem 1, this table must be owned and reviewed like any other part of the matrix, because a single wrong mapping assigns the wrong curriculum to every person who holds that job code.

Clinical Staff and Trial Training

For companies with clinical operations, the LMS often holds GCP training as well. ICH E6(R3) section 2.3.2 says that trial-related training for people assisting in a trial should correspond to what they need for their delegated trial activities that go beyond their usual training and experience.¹⁸ That is the same task-based principle as Problem 1, applied to trials, and it argues against assigning every clinical SOP to every clinical employee. If the LMS will also train site staff or vendors, scope external user access, identity verification, and record ownership during selection rather than after.

A Sequenced Plan for Curriculum and Platform Work

The order of work matters more than the speed. The plan below puts the curriculum work first, in the current system, so that selection and migration are based on curricula that are already correct.

Steps 1 through 3 can run while the business case for a new platform is still being written. In some cases they solve enough of the problem that the platform change can wait, which is a legitimate outcome. Where the platform change does go ahead, the curriculum work gives the project a defined scope, a baseline to measure against, and a much smaller set of records to migrate.

Conclusion

Training compliance problems usually look like software problems because the software is where people see them: the overdue list, the slow report, the reminder email nobody reads. Underneath, they are curriculum problems. Roles that do not match tasks, read-and-understand used for everything, revisions that retrain everyone, effectiveness left to individual courses, and matrices without owners will produce the same results in any LMS. Fix them first, in the system you already have, and the platform decision becomes simpler, the requirements become specific, and the migration carries fewer and cleaner records. Then scope the platform work around what inspectors ask: who was qualified, on which version, on which date, and who decided they needed it.

Sakara Digital works with pharma and biotech organizations on quality systems and the training programs that support them, including curriculum redesign, LMS selection, and training history migration. If you are planning an LMS change and want an independent view on whether the curricula are ready to move, we are happy to have that conversation.

For Further Reading