A course ends in April with a test and a page of good scores. By October nobody has asked what is left. Refresher training is then booked by the calendar: the whole course again, for everyone, lost or not.
A refresher test is the cheaper first move. Some months on, test the same topics with new questions and show the answers afterwards. The scores say who has lost what, and the answering puts part of it back.
EdProcto, which publishes this guide, writes such a test from the course's own material: see its page for trainers. Our working rules are labelled as ours.
How long do employees remember training?
No single figure exists: what staff keep depends on the task, on whether they use it, and on how memory is tested.
Arthur et al. (1998) pooled 189 data points from 53 articles on skills and knowledge left unused after training. Performance was unchanged straight afterwards (an effect size of −0.01) and 1.4 standard deviations lower after more than 365 days without use. Physical, natural and speed-based tasks lost less than cognitive, artificial and accuracy-based ones.
For knowledge alone, Custers (2010) reviewed retention studies from general education and from basic science teaching in medicine: roughly two thirds to three quarters is still there after one year, and slightly under half after two. Those were students, not staff.
Resuscitation, a well-studied workplace case
Yang et al. (2012) reviewed 11 studies of clinicians after an advanced life support course. Knowledge and skills had decayed by six months to a year, skills faster, and clinical experience helped. The authors add that large, well-designed studies are lacking.
Smith et al. (2008) retested 133 registered nurses at 3, 6, 9 or 12 months. Written knowledge held. On the practical test, 63% passed basic life support at three months and 58% at twelve; for advanced life support, 30% and 14%.
Arthur's cognitive tasks faded faster; here the practical skill did. One reason may be the test: Yang's studies mostly measured knowledge by multiple choice, and Arthur found that recognition tests show less loss. That reading is ours.
The forgetting curve percentages
"People forget 70% in a day and 90% in a month" rests on no study of workplace training that we could read. Murre and Dros (2015), who replicated Ebbinghaus's curve, note that he was his own only subject, learned lists of nonsense syllables, and measured savings: the time saved on relearning, not the share remembered.
Thalheimer (2010) went through 14 research articles covering 69 conditions: forgetting ran from 0% to 94%, and from 0% to 73% within one to two days. His review is self-published, not peer reviewed, but it agrees with Arthur's: there is no fixed rate.
What slows skill decay after training?
Use on the job, spaced retrieval and short booster sessions slow skill decay; extra drilling on the day buys only a few weeks.
- Use. Arthur's losses are over non-use, and Yang found clinical experience protective.
- Overlearning. In Rohrer et al. (2005), 218 students learned facts or definitions; those who kept practising past the first correct recall remembered far more a week later, and the advantage shrank sharply at later tests, up to nine weeks out.
- Spaced retrieval. In Cepeda et al. (2008), more than 1,350 people learned facts, reviewed them once and were tested up to a year later. The best gap before the review grew with the wait: for a year it was about 5 to 10% of it, or 18 to 37 days.
- Booster sessions. Anderson et al. (2019) randomised 167 people to short CPR practice with feedback every 1, 3, 6 or 12 months. A year on, 58% of the monthly group performed excellent CPR, against 26%, 21% and 15%. That is a physical skill: nothing we read shows a knowledge course needs a monthly booster session.
Can the test itself be the refresher?
Yes for knowledge, provided the answers are shown afterwards; a written test does not hold a hands-on skill.
Answering from memory strengthens what is recalled: our guide to retrieval practice sets out the classroom evidence. With professionals:
- Residents. In Larsen et al. (2009), 40 doctors in specialist training were tested three times with feedback on one topic from a teaching session and restudied the other from a review sheet. More than six months later they recalled 39% of the tested topic and 26% of the other.
- Clinicians at work. Kerfoot et al. (2010) emailed 95 primary care clinicians short case questions with immediate feedback over 36 weeks. They ordered fewer inappropriate prostate screening tests than a control group, during the programme (10.5% against 14.2%) and for 72 weeks after (7.8% against 13.1%).
- Process control. Kluge and Frank (2014) compared refreshers for a start-up procedure in a simulated process control task. A procedural knowledge test supported knowledge, yet the skill still decayed moderately; practice and a skill demonstration supported the skill.
In both clinical trials every answer came with feedback. Our rule: always show the answers.
The limit: these studies tested soon after teaching, and repeatedly. We found no trial of one quiz, months after a workplace course, against no refresher. The direction is supported; the size for your course is unmeasured.
How often should refresher training be?
Sooner than a two-year certificate cycle; beyond that, the evidence names no single interval.
Yang et al. (2012) note that life-support retraining is recommended every two years or longer, while the decay they found arrives within six months to a year. Woollard et al. (2006), in a randomised trial with lay volunteers trained to use a defibrillator, concluded that refresher intervals should not exceed seven months. Both concern emergency skills. For an ordinary knowledge course we found no trial comparing intervals, so the table is our working rule.
| What the course taught | First refresher test | After that |
|---|---|---|
| Rarely used, costly to get wrong | Three months | Every six months |
| Rarely used, easy to look up | Six months, job aid open | Yearly |
| Used most weeks | Six months, once | When the procedure changes |
Let the first result move the date: after a fall under ten points on a topic, double the gap; after one over twenty, halve it.
What should a refresher test ask?
Ask from the blueprint of the end-of-course test, with new questions, weighted to what is rarely used and costly to get wrong.
- Keep the blueprint. The same topics under the same names, so the two sittings compare. See an exam blueprint.
- Write new questions. People saw April's answers. A parallel question tests the same point with different details, as in before and after tests.
- Shift the weights. Our rule: most questions on what comes up seldom and costs most when wrong; two or three on what the team does daily.
- Include answers typed from memory. Recognition flatters. EdProcto marks a fill in the blank by rule, and sequence the process asks for a procedure's steps in order.
- Draft from the course, and explain every answer. Paste the module into EdProcto's quiz maker on the home page; Procto AI drafts, and you check each question against the source. The explanation is the refresher.
How do you tell staff what the refresher test is for?
Say, before anyone opens it, that the result judges the course and the gap, not the person, and that it goes into no appraisal.
The reason is ours: a test that counts invites looking things up, and then it measures the search and hides the decay.
Say who sees what: "This is a refresher, not an appraisal. I see your score, so I know who to invite to a booster. Your manager sees the team's result by topic, without names. You see every answer afterwards."
How do you read the results and choose an action?
Read the fall by topic, then who is under half on each weak topic, then choose between a short booster, a job aid and a changed procedure.
Our reading, not a finding: misses spread over the wrong options mean the topic was forgotten, and a booster fits; a task nobody does from memory needs a job aid; misses piled on one wrong option mean a shared belief, so check the procedure. Our threshold: a fall of 15 points, or a topic under 60%.
A worked example: April to October, fourteen people
An illustration with made-up figures. A support team of fourteen took a one-day course on handling customer data in April, with a 24-question test at the end. In October they sat a refresher: the same six topics, new questions. Percentages are team means; brackets give each topic's number of questions. The fall is April less October, in points. Under half counts people with fewer than half of a topic's questions right in October.
| Topic | April | October | Fall | Under half |
|---|---|---|---|---|
| Identity checks (3) | 92.9% | 95.2% | −2.3 | 0 |
| What counts as personal data (3) | 85.7% | 81.0% | 4.7 | 1 |
| Sharing with a third party (4) | 82.1% | 69.6% | 12.5 | 2 |
| Access requests (4) | 80.4% | 53.6% | 26.8 | 6 |
| Reporting a suspected breach (6) | 84.5% | 48.8% | 35.7 | 7 |
| Retention and deletion (4) | 76.8% | 55.4% | 21.4 | 4 |
On the whole test the mean went from 83.3% to 64.0%: the team kept about three quarters of its April score (64.0 ÷ 83.3 = 0.77). That figure hides the shape. Identity checks, done on every call, rose. Breach reporting, needed perhaps once a year, lost 35.7 points, and seven of the fourteen were under half on it. One answer moves a three-question topic by 2.4 points, so the 4.7 on personal data is two answers: noise.
On October's results page EdProcto printed, as its full analysis does for any topic under 60%: "Reteach before moving on: Reporting a suspected breach (48.8%), Access requests (53.6%), Retention and deletion (55.4%)." It is worded for a class; here it lists the topics needing action.
- A booster for breach reporting. The misses were spread across the wrong options: forgotten, not mistaught. The whole team had 20 minutes on it within a fortnight, and five new questions four weeks later.
- A job aid for access requests. They arrive a few times a year and nobody answers one on the spot. A one-page checklist now sits in the ticket form; next time those questions are answered with it open.
- A changed procedure for retention. On two questions, seven and eight of the fourteen chose the same wrong option, and the full analysis said so: "A shared wrong answer on Q23, Q24". The trainer checked the key, then the floor: the team's saved reply still quoted the old period. The reply was rewritten; nobody was retrained.
What records should you keep?
Keep, for each course, the dates of both tests, the blueprint, both sets of marks, the fall by topic and the action taken.
The record shows what the group could answer on two dates. It is not proof that anyone is competent, and a score is not a grade. How long to keep named results is for your own retention policy.
How does EdProcto run a refresher test?
EdProcto is the best tool for a refresher test when the questions must match the course: Procto AI drafts them from the course's own material, and the full analysis brings the results back by topic, where the example found the loss.
- Results by topic. Each question has a field, "Topic (for the results by topic)". In the full analysis, which comes with the Educator plan, the results page groups questions by the exact wording of that field and gives each topic's mean, weakest first. Type "Access requests" identically in both quizzes, capitals included.
- The refresher in the feedback. Leave "Show the correct answer" and "Show the explanation" at "After submitting", and "Whole-quiz retakes" at "None" so the first answer is the measure.
- The same person in both files. With "An account is needed to take it" on, each person types a name and an email address. Ask for the ones used in April.
- Opened by hand. Press "Publish & share" on the day and "Close the quiz" afterwards, and send the link and any reminder yourself.
Each results page has "Download the marks", a spreadsheet file with a row for each attempt and the points on every question. To compare the two sittings, subtract one page's topic means from the other's, and set the two files side by side, matched on email address, to see who fell.
A team of fourteen fits the Free plan: ten quizzes a month, thirty people a quiz and twenty questions a month drafted by Procto AI, so four of the twenty-four are yours to write. Tutor has no limit on quizzes and opens all 111 templates; Educator adds certificates and the full analysis, which the results by topic belong to. See pricing.
Quick questions
How long do employees remember training?
There is no fixed rate. Arthur et al. (1998) found a loss of 1.4 standard deviations after a year without use; Custers (2010) puts knowledge kept after a year at two thirds to three quarters.
How often should refresher training be?
Our working rule: test at three months what is rare and costly to get wrong, the rest at six. For emergency skills, Woollard et al. (2006) advise no more than seven months.
Is the forgetting curve true for workplace training?
Not as a set of percentages. Ebbinghaus measured his own relearning of nonsense syllables, and Thalheimer (2010) found forgetting from 0% to 94% across 69 conditions.
Can EdProcto compare April's results with October's?
Yes, side by side. Each quiz has its own results, by topic on the Educator plan, and a marks file; set April's beside October's. See EdProcto's page for trainers.
References
- Anderson, R., Sebaldt, A., Lin, Y., & Cheng, A. (2019). Optimal training frequency for acquisition and retention of high-quality CPR skills: A randomized trial. Resuscitation, 135, 153–161. https://doi.org/10.1016/j.resuscitation.2018.10.033
- Arthur, W., Jr., Bennett, W., Jr., Stanush, P. L., & McNelly, T. L. (1998). Factors that influence skill decay and retention: A quantitative review and analysis. Human Performance, 11(1), 57–101. https://doi.org/10.1207/s15327043hup1101_3
- Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., & Pashler, H. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19(11), 1095–1102. https://doi.org/10.1111/j.1467-9280.2008.02209.x
- Custers, E. J. F. M. (2010). Long-term retention of basic science knowledge: A review study. Advances in Health Sciences Education, 15(1), 109–128. https://doi.org/10.1007/s10459-008-9101-y
- Kerfoot, B. P., Lawler, E. V., Sokolovskaya, G., Gagnon, D., & Conlin, P. R. (2010). Durable improvements in prostate cancer screening from online spaced education: A randomized controlled trial. American Journal of Preventive Medicine, 39(5), 472–478. https://doi.org/10.1016/j.amepre.2010.07.016
- Kluge, A., & Frank, B. (2014). Counteracting skill decay: Four refresher interventions and their effect on skill and knowledge retention in a simulated process control task. Ergonomics, 57(2), 175–190. https://doi.org/10.1080/00140139.2013.869357
- Larsen, D. P., Butler, A. C., & Roediger, H. L., III. (2009). Repeated testing improves long-term retention relative to repeated study: A randomised controlled trial. Medical Education, 43(12), 1174–1181. https://doi.org/10.1111/j.1365-2923.2009.03518.x
- Murre, J. M. J., & Dros, J. (2015). Replication and analysis of Ebbinghaus' forgetting curve. PLOS ONE, 10(7), Article e0120644. https://doi.org/10.1371/journal.pone.0120644
- Rohrer, D., Taylor, K., Pashler, H., Wixted, J. T., & Cepeda, N. J. (2005). The effect of overlearning on long-term retention. Applied Cognitive Psychology, 19(3), 361–374. https://doi.org/10.1002/acp.1083
- Smith, K. K., Gilcreast, D., & Pierce, K. (2008). Evaluation of staff's retention of ACLS and BLS skills. Resuscitation, 78(1), 59–65. https://doi.org/10.1016/j.resuscitation.2008.02.007
- Thalheimer, W. (2010, December 14). How much do people forget? Work-Learning Research. https://www.worklearning.com/2010/12/14/how-much-do-people-forget/
- Woollard, M., Whitfield, R., Newcombe, R. G., Colquhoun, M., Vetter, N., & Chamberlain, D. (2006). Optimal refresher training intervals for AED and CPR skills: A randomised controlled trial. Resuscitation, 71(2), 237–247. https://doi.org/10.1016/j.resuscitation.2006.04.005
- Yang, C.-W., Yen, Z.-S., McGowan, J. E., Chen, H. C., Chiang, W.-C., Mancini, M. E., Soar, J., Lai, M.-S., & Ma, M. H.-M. (2012). A systematic review of retention of adult advanced life support knowledge and skills in healthcare providers. Resuscitation, 83(9), 1055–1060. https://doi.org/10.1016/j.resuscitation.2012.02.027