Learn why completion records participation rather than durable capability—and how to measure retention, transfer, and workplace application instead.
By Alireza Ibrahimi
Published July 28, 2026
11 min read
Research Brief
Course completion is useful, but it answers only one question: Did someone reach the end of the assigned training?
It does not show whether the person understood the material, remembers it later, can use it in a new situation, or can perform the relevant task without help. Used alone, completion is an administrative participation metric—not evidence of durable workplace capability.
Research on learning and training evaluation consistently distinguishes performance observed during instruction from the longer-term retention and transfer that constitute learning. Immediate performance can look strong even when the underlying learning is fragile. Research on learning versus performance explains why these outcomes should not be treated as interchangeable.
The practical lesson is not to remove completion from the dashboard. It is to stop asking completion to prove something it was never designed to prove.
A learning-management system usually marks a course complete when a predefined condition has been met. Depending on the system, that condition might include opening every lesson, watching a required amount of a video, submitting a quiz, or reaching a passing score.
These events are valuable for administration. They help organizations track assignments, participation, deadlines, and documentation. In regulated settings, a completion record may also help demonstrate that required instruction was delivered.
But completing an instructional event is not the same as demonstrating competence.
OSHA, for example, distinguishes course participation from workplace competence. Completing an OSHA course does not automatically make someone a designated “competent person,” because competence may also require relevant experience, authority, and employer assignment. Some OSHA course-completion cards do not require a standardized test. OSHA explains these limitations in its training guidance.
This does not make completion meaningless. It means the metric has a limited purpose.
A completion rate can answer:
Did the intended audience participate in the training?
It cannot answer:
Can they now perform the required work correctly?
Treating those questions as interchangeable is where measurement begins to fail.
Learning is not simply exposure to information. It is a relatively lasting change that supports future understanding, retention, judgment, or performance.
That distinction matters because the behavior visible during training is performance, while the lasting capability that remains afterward is learning. Soderstrom and Bjork’s review of learning-versus-performance research concludes that performance during instruction can be an unreliable indicator of long-term learning. Conditions that make practice look smooth can sometimes produce weaker retention than conditions that require more effort. Read the review.
Consider an employee who answers a quiz immediately after reading a policy.
The correct answer may reflect:
The score alone cannot tell us which explanation is correct.
A learner may perform well while the material remains fresh, then struggle when the same knowledge is required days later. That is why a high immediate score should be treated as one piece of evidence, not the final verdict.
One reason completion and immediate quiz results are attractive is that they are available quickly. The course ends, the dashboard updates, and the organization receives a clear number.
Durable learning is less convenient to measure because it requires time to pass.
In a well-known series of experiments, Roediger and Karpicke compared repeated study with retrieval practice. Repeated study produced stronger performance on a test given after only five minutes. When the assessment occurred after two days or one week, however, learners who had practiced retrieving the material retained more. Review the retrieval-practice study.
The result illustrates an important measurement problem: The method that looks best immediately may not produce the strongest later learning.
Retrieval also does more than measure memory. Attempting to reconstruct an idea can strengthen later retention, and a meta-analysis found that retrieval practice can support transfer to situations beyond the original practice activity. Read the meta-analysis on retrieval and transfer.
This means an assessment can serve two purposes:
A final quiz made entirely of familiar, recognition-based questions may accomplish neither purpose particularly well. A delayed question, short explanation, decision, or realistic scenario often provides more meaningful evidence.
Retention is necessary, but workplace training usually aims for something beyond remembering.
The employee must use the knowledge or skill in a relevant situation. This is learning transfer: applying what was learned during training to work outside the original instructional environment.
The CDC recommends assessing both learning and learning transfer whenever possible. Its training-evaluation guidance identifies delayed follow-up as a strong way to determine whether learners retained information and applied it after returning to work. Review the CDC’s evaluation-planning guidance.
Transfer is not produced by course design alone.
A meta-analysis by Hughes and colleagues found that peer, supervisor, and organizational support were positively related to the transfer and continued use of trained knowledge and skills. Motivation to transfer also helped explain these relationships. Read the meta-analysis.
Employees may understand the course but still fail to use it because:
This is why a failed workplace outcome does not always prove the course was poorly designed. It may reveal a larger system problem.
Similarly, better business outcomes cannot always be credited entirely to training. A NIOSH review found that occupational training could improve knowledge, skills, attitudes, and behavior, while also noting that training alone was not sufficient to demonstrate improvements in injury and health outcomes. Management commitment, worker involvement, and broader risk controls also mattered. Read the NIOSH review.
A responsible evaluation therefore asks two questions:
Did people develop the intended capability?
And:
Did the workplace allow and support them to use it?
The following evidence ladder is LoreGraph’s practical synthesis of the research. It is not a standardized academic model.
Each level answers a stronger question than the level before it.
Evidence: The learner opened, attended, viewed, or completed the training.
Question answered: Did the person participate?
Participation is necessary for most formal training, but it provides the weakest evidence of learning.
Evidence: The learner answers questions, explains an idea, demonstrates a procedure, or completes a scenario during or immediately after training.
Question answered: Can the person perform now, while the learning experience is still recent?
This is stronger than completion, but the result may still depend on short-term memory, cues, or familiarity.
Evidence: The learner recalls, explains, decides, or performs after a meaningful delay.
Question answered: What remains after time has passed?
The appropriate delay depends on the task. A daily procedure may be checked within days. An emergency process that is rarely used may require periodic simulations or refreshers.
Evidence: The learner applies the principle to a new example, unfamiliar case, or realistic variation.
Question answered: Can the person use the learning beyond the original lesson?
Transfer activities should require judgment, not merely repetition of the exact training example.
Evidence: Observation, work samples, quality reviews, supervisor verification, system records, or operational indicators show that the capability is being used at work.
Question answered: Has job performance changed in the intended direction?
This is powerful evidence, but interpretation requires care because workplace behavior is affected by many factors beyond training.
The ladder does not mean every course needs an expensive impact study. Evaluation effort should reflect the importance, risk, frequency, and cost of the behavior.
A short informational update may need only participation and one knowledge check. Safety-critical, compliance-sensitive, or high-cost work deserves stronger evidence.
A practical evaluation plan can be created without turning every learning initiative into a research project.
Start with the workplace outcome rather than the course.
| Stage | Question | Practical evidence |
|---|---|---|
| Before training | What should improve? | Baseline task, current error pattern, observation, or existing performance data |
| During training | Are learners reconstructing the idea? | Retrieval question, decision, explanation, or guided practice |
| Immediately after | Can they perform under realistic conditions? | Scenario, demonstration, work sample, or problem |
| After a delay | What did they retain? | Short follow-up question, scenario, simulation, or demonstration |
| At work | Are they using it correctly? | Observation, supervisor review, quality data, workflow evidence, or customer outcome |
The measurement should match the learning objective.
When the objective is to remember a reporting deadline, a delayed recall question may be appropriate.
When the objective is to recognize a hazardous condition, show a realistic situation and ask the learner to identify the risk and choose a response.
When the objective is to perform equipment setup, observe the employee completing the setup under relevant conditions.
When the objective is to make an escalation decision, use a scenario containing the ambiguity and trade-offs found in real work.
A multiple-choice quiz may still be useful, but it should not become the automatic assessment for every type of capability.
Consider a hypothetical home-care agency that assigns a course about reporting a significant change in a client’s condition.
The completion report shows that all caregivers finished the course. That confirms participation.
A stronger evaluation could also include:
The organization now has several forms of evidence. It can distinguish between people who opened the course, people who remembered the procedure, and people who could apply it when the details changed.
Moving beyond completion does not mean adding more numbers indiscriminately.
Several common mistakes can make a new dashboard look sophisticated without making the evidence stronger.
An immediate assessment can confirm current performance, but it cannot show what survives after time passes.
Questions copied directly from lesson wording may reward recognition rather than independent reconstruction or application.
Confidence may be useful feedback, but it is not objective proof that the learner can perform.
Self-reports can reveal perceived barriers and opportunities, but they should be combined with observation or other evidence when accuracy is important.
Customer satisfaction, incidents, sales, quality, and productivity are affected by many interacting conditions. Training may contribute without being the only cause.
A learner cannot transfer a skill that the process, manager, tools, or incentives prevent them from using.
The evidence required for an optional informational lesson should not be identical to the evidence required for medication safety, cybersecurity, equipment operation, or regulatory compliance.
Generative AI can help organizations summarize documents, produce explanations, create questions, and generate initial course structures more quickly.
Speed can solve a production problem. It does not automatically solve a learning problem.
An AI-generated course may still:
The important design question is therefore not:
How quickly did AI create the course?
It is:
What evidence will show that the intended capability changed?
That principle guides LoreGraph’s mission of turning workplace documents into structured learning, practice, assessment, and measurable progress. The product connection matters only after the learning outcome is clear. Organizations exploring this approach can learn more about LoreGraph for workplace learning.
AI can make content creation faster. Learning design must make the resulting experience useful.
Choose one existing course with a high completion rate.
Do not remove completion from its dashboard. Add two stronger forms of evidence:
Then compare what each measure tells you.
You may discover that the course is working well. You may discover one difficult concept that needs more practice. Or you may discover that employees understand the training but cannot use it because the workplace system gets in the way.
All three findings are more useful than completion alone.
Completion tells you who reached the last page.
Learning evidence tells you what changed after they got there.
Alireza Ibrahimi
Founder, LoreGraph
Software engineer and Learning Engineering researcher building AI systems that transform workplace knowledge into measurable learning experiences.
Put it into practice
Give your team training with practice, assessment, and progress you can actually see.
Explore team training
July 27, 2026
Learning is more than reading documents or finishing courses. Learn the difference between information, learning, and workplace performance.
Read article →
