Training That Sticks
Everyone attended. 94% passed the test. Two months later the exact thing you trained for happens anyway. That isn’t a problem with your people or your trainer — it is a property of memory that has been measured since 1885, and most of it is fixable. A free 28-minute course for whoever ends up responsible for training: five short lessons, a test after each, and every number in it sourced.
If you take one thing from this page
The single change with the largest effect isn’t a better session. It’s coming back to the material after part of it has already been forgotten. The best gap grows with how long you need people to remember, but more slowly than it does (Cepeda et al., 2008):
- Need it to last a week? Come back after a day.
- A month? After about ten days.
- Two months, or a year? After three weeks either way — the gap stops growing long before the target does.
- Not sure? Wait longer rather than sooner. Too short a gap costs you a lot; one that’s too long costs noticeably less.
Each return can be five to seven questions, unannounced. The point isn’t to measure — it’s to make people retrieve the material, which is what actually strengthens it.
Why the training disappears
The usual response to forgetting is a more thorough session: more hours, more slides, more examples. It doesn’t work, and the reason is mechanical. The forgetting curve starts the moment the session ends, and its shape doesn’t depend on how many hours came before it. An eight-hour session is not remembered eight times better than a two-hour one a month later.
Two groups spent the same time with the same text. One re-read it four times; the other read it once and then spent three periods trying to recall it. Five minutes later the re-readers were ahead, 83% to 71%. A week later it had flipped: about 40% versus about 60% (Roediger & Karpicke, 2006).
Re-reading feels more reliable precisely because the score right afterward is higher. That feeling is familiarity — your eye recognizing the phrasing — and a week later it is gone. Which is why a test isn’t a checkpoint that follows the learning. It is part of the learning, and usually the strongest part.
And why nobody notices
About 90% of organizations evaluate training by asking whether people liked it (ATD, 2016). The link between a “liked it” rating and how much anyone actually learned is a correlation of 0.02 — effectively none (Alliger et al., 1997). Behavior, the level where the point of the training lives, gets evaluated for about a third of programs. What gets measured is what’s easy to measure, and that is rarely what matters.
What you’ll be able to do afterward
- Say why a session is forgotten without blaming the people who attended
- Put a schedule of reviews around an existing course — with the gaps the research actually supports
- Tell a question that measures understanding from one that measures recognition of a slide
- Write a wrong option that is somebody’s real mistake instead of obvious filler
- Set a passing score by the cost of an error rather than by how strict it feels
- Recognize the four levels of evaluation, and name which one your company actually stops at
- Pick a measurement of behavior you can afford, and take the baseline before you need it
- Turn a compliance course that exists for the record into one that also works
What’s inside
Where the training goes
The forgetting curve, what Ebbinghaus actually measured, and why length is the wrong lever.
A test is not a checkpoint
The testing effect, and how long to wait before coming back to the material.
The question that measures understanding
Five item-writing rules, how many questions, and what passing score.
How to tell whether it worked
Kirkpatrick’s four levels, why level one means nothing, and three ways to measure behavior.
Compliance without the cynicism
What kills mandatory training, and a setup that runs with no budget and no platform.
A six-question test follows each lesson. The pass mark is 70% and attempts are not limited — deliberately: lesson 2 is about why a retake is the most useful part of the process, and it would be odd to teach that and then ration it. The last page of the course is the full reference list.
Questions people ask
- Who is this for?
- Whoever ends up responsible for training — HR, L&D, or a manager who was handed the task. It assumes no background in instructional design and no budget.
- Does it promise that my people will remember the material?
- No, and it says so in the first lesson. It explains why a one-off session is forgotten and shows which parts of that are fixable. What you get is the ability to tell the difference — not a guarantee about anyone else’s memory.
- Where do the numbers come from?
- Every figure in the course carries its source in brackets, and the last page lists all twelve papers in full. Where the research doesn’t settle something — and there are two such places — the course says so instead of rounding it into a claim.
- Do I need an account?
- No. You can go through the whole course without signing up. An email address is asked for at the end, and only if you want the certificate.
- Is it really free?
- Yes, for anyone taking it. Cletica makes money from organizations that run courses for their own staff and need the records — not from learners.
- Can I run this for my team?
- Yes. The course can be assigned to a list of people, with per-person progress, test scores, time spent on each lesson, and a certificate for everyone who passes.
Related
The running example in this course is security training — people pass the phishing test and click the link two months later. Security Awareness for Employees is the course on the other side of that: the one your people take. Reading it after this one is a fair way to check whether it does what this course says a course should.
Running this for a team?
The same course can be assigned to a list of people, with per-person progress, test scores, time spent on each lesson, and a certificate for everyone who passes. Built on Cletica, which is also where you can build your own courses, surveys and tests — including the spaced follow-up tests this course argues for.