Key takeaways
- →Memory loss after training is steepest in the first day, so the first review matters more than the three that follow it.
- →Four 45-minute sessions spread across a month will out-retain one six-hour day, even though the single day is far easier to book.
- →End-of-session feedback measures how the training felt, not what anyone kept; only a delayed check tells you whether it worked.
- →Every successful retrieval flattens the next curve, which is why three short quizzes beat one long recap slide.
- →If you truly only get the room once, spend the final ten minutes manufacturing the follow-up rather than covering more content.
The forgetting curve is the reason a training session that felt excellent on the day leaves almost nothing behind by the end of the month. Ebbinghaus's memory experiments showed that newly learned material falls away fastest in the first hours and days, then flattens out, which means the gap between your session and the first time anyone uses the material is doing more damage than the quality of your slides. Any serious forgetting curve training plan therefore starts with spacing, not with content.
That implication is uncomfortable for anyone who has to book rooms. If you have a day of training budget, you will get more retention from four 45-minute sessions spread over a month than from one six-hour marathon, even though the marathon is simpler to schedule and feels more substantial to the person paying for it.
This article covers what the curve does and does not claim, how to build a spacing arc around a real session, what to do when you only get the room once, and how to measure decay instead of assuming it.
What the forgetting curve actually claims
Hermann Ebbinghaus ran the original experiments on himself in the 1880s, memorising lists of nonsense syllables and re-testing at intervals. He found that the proportion he could still recall dropped sharply within the first hours, more gently over the following days, and then levelled off. The result is a curve that falls hard and then flattens, not a straight line.
Two caveats matter before you build a training programme on it. Nonsense syllables are not a product roadmap or a safeguarding policy, and meaningful material that hooks into what someone already knows decays more slowly. The specific percentages people quote in slide decks come from one man testing one kind of material, and should be treated as illustration rather than measurement.
What survives the caveats is the shape, and the shape is what you plan against. Loss is front-loaded. Anything you do in the first 24 to 48 hours has more leverage than anything you do a fortnight later. And each time someone successfully pulls the material back out of memory, the next curve starts higher and falls more slowly.
Where the losses actually happen
Map the decay against your calendar rather than against an abstract axis. The picture below is the pattern most workplace trainers see when they test the same five questions at intervals: strong recall while the material is still fresh, a hard fall overnight, and a long shallow tail that is much easier to hold up than to rebuild.
End of session
Recall looks excellent, but the material is still sitting in working memory and half the room can see your last slide. This is the least informative moment to test anything.
Next morning
The steepest part of the fall has already happened. A five-question check here recovers more than the same check run a week later.
One week
Detail is gone; structure often survives. People remember that there were four steps, not what step three was.
One month
Without a retrieval touch, what is left is a vague sense of the topic. With three touches behind it, the curve has flattened enough to be usable on real work.
Why the room's reaction tells you nothing
Ask people at half past four whether the session was useful and most will say yes. They are answering a question about how the last few hours felt: whether you were clear, whether the examples landed, whether they were bored. None of that predicts what they will be able to do in a fortnight.
The trap has a name in learning research: fluency. While material is in front of you, processing it feels effortless, and the brain reads that ease as evidence of knowing. Re-reading a handout produces the same false signal. The only reliable test of whether something is retrievable is to try to retrieve it without the source in view.
So keep your end-of-session sheet if you like it, but stop treating it as an outcome measure. It tells you whether people would come back. A quiz on day seven tells you whether they learned anything.
The three levers that flatten the curve
You cannot make people forget more slowly by explaining more clearly. Clarity affects how much gets encoded in the first place, which matters, but the slope after the session is governed by what happens between sessions. Three levers do almost all the work.
Spacing: the gaps do the work
Distributed practice beats massed practice for durable memory. The same total study time, broken up and spread out, produces better long-term recall than one continuous block, which is exactly why the two-day intensive is a poor format for anything you need in six months. Gaps feel like inefficiency and are actually the active ingredient.
Retrieval: pulling beats pushing
A review slide pushes the material back at people. A question makes them pull it out. Retrieval strengthens the memory far more than re-exposure does, and the effort of a partly failed attempt is part of why it works. If a follow-up touch is a recap email, you have chosen the weaker option; make it three questions instead.
Consequence: put it on a real task inside a week
The strongest retention lever available to most trainers is not a learning technique at all. It is arranging for someone to have to use the material on live work within seven days: a real customer call using the new objection handling, a real ticket triaged with the new rubric. Use creates its own retrieval, repeatedly, with consequences attached.
A default schedule you can defend to a sponsor
When someone asks why the training needs four calendar entries instead of one, give them numbers rather than theory. These are the structural defaults we recommend as a starting point; adjust the intervals to your material, but keep the first touch inside a day.
Building a four-touch training arc
Cut the content until three things are load-bearing
Before you can space anything you have to decide what deserves to survive. Write down the three things a participant must still be able to do in a month. Everything else in the deck is context, and context does not need a retrieval schedule.
Write the questions before you write the session
Draft the five checks you will run the next morning first, then build the session so that those five answers are unavoidable. This is the single change that most improves a training design, because it forces you to teach toward something testable rather than toward a running time.
Book the follow-ups in the same invitation
Three ten-minute slots at day one, day seven and day thirty, sent with the original calendar invite. Follow-ups that are arranged after the session almost never happen, because the enthusiasm that would have booked them has already decayed too.
Run the day-one check cold
No recap first, no notes open. If you re-present the material before you test it, you measure reading speed rather than memory. Ten minutes is enough: five questions, then three minutes on whichever one the group got wrong.
Change the format at each touch
Multiple choice on day one, a short scenario on day seven, a live discussion prompt at thirty days. Varying the retrieval context makes the memory more flexible and stops people learning the question rather than the answer.
Retire what holds and re-teach what does not
By the third touch you will have two or three items that everyone gets right and one that keeps failing. Drop the first group and rebuild the second from scratch, because a concept that fails three retrievals was not encoded properly on the day.
What to put in each follow-up
Top Tips
- Open with the question, not with the recap. Even fifteen seconds of struggle before the answer appears does more for retention than a perfectly worded reminder.
- Mix in one question from the previous touch so that older material keeps getting revisited rather than being reviewed once and abandoned.
- Show the distribution of answers, not just the correct one. Seeing that half the room chose the same wrong option is the moment the misconception surfaces and can be addressed.
- Keep the whole touch under ten minutes. Follow-ups die of ambition, not of indifference; a ten-minute slot survives a busy week and a forty-minute one does not.
- Include one question that cannot be answered from memory alone, so people practise applying the material rather than reciting it.
Before anyone leaves the room
The end of the session is the only moment when you have everyone's attention and everyone's diary in the same place. Use it to build the scaffolding for the next month rather than to summarise what you have already said.
- ✓The three load-bearing points are written down somewhere participants can find them
- ✓Day one, day seven and day thirty checks are already in everyone's calendar
- ✓Each person has named one real task where they will use the material this week
- ✓You have a baseline score from an in-session quiz to compare later results against
- ✓The recap that goes out contains questions, not just a summary of the slides
- ○Managers know which three behaviours to look for over the next fortnight
When you only get the room once
Plenty of training happens at a conference, an onboarding week or a supplier visit where a follow-up session is genuinely impossible. The curve does not care about your constraints, so you have to compress the spacing inside the session itself.
Space within the day. Teach a block, move to something else, then come back and test the first block forty minutes later without warning. Interleaving two or three strands rather than finishing one before starting the next produces worse performance during the session and better retention afterwards, which is exactly the trade-off you want and exactly the one participants will complain about.
Then spend the last ten minutes manufacturing a follow-up you will not be present for. Have each person write two questions they think a colleague should be able to answer in a fortnight, collect them, and send them out as a quiz on day seven. Nobody needs to book a room for that, and it converts a one-off into a two-touch arc.
Measure the decay instead of assuming it
You cannot improve a curve you have never plotted. The cheapest useful measurement is to run the same five questions three times and watch the scores: in session, at day seven, at day thirty. The absolute numbers matter less than the gradient between them.
As a rough working rule, a drop of more than about a fifth between day one and day thirty means the spacing is too loose or the material was never encoded properly. A flat line across the three points usually means your questions are too easy rather than that your training is brilliant, so make them harder before you celebrate.
Keep the question set stable across cohorts. Once you have run the same checks with three groups you have something far more useful than a satisfaction score: a comparable picture of which parts of your content reliably survive and which parts evaporate no matter who delivers them.
Three habits that quietly help
Test before you teach
A two-minute quiz at the start on material nobody has seen yet feels pointless and is not. Guessing and being wrong primes attention for the correct answer when it arrives.
Make the first touch social
A day-one check that a whole team does together, with the distribution visible, gets far higher completion than an individual assignment nobody can see anyone else doing.
Keep a question bank, not a slide deck
Decks are rewritten for every audience and thrown away. A bank of tested questions with known difficulty is the asset that makes your next training cheaper to run.
Four ways facilitators break their own spacing
Top Tips
- Sending a recap document instead of a question set. It feels like a follow-up, generates no retrieval, and mostly produces the fluency illusion in written form.
- Scheduling the first touch a fortnight later because that was the next free slot. By then the steepest part of the curve has done its work and you are rebuilding rather than reinforcing.
- Making the follow-up optional and voluntary. Attendance collapses to the people who least need it, which then makes your retention data look artificially healthy.
- Changing the material between touches. If day seven asks about a revised framework, you are teaching something new rather than strengthening something old, and the curve restarts.
- Attaching scores to appraisal. The moment a retention check has career consequences, people revise for it specifically, and you lose the honest signal about what the training actually left behind.
Running the follow-ups without adding meetings
The practical obstacle to spacing is rarely disagreement about the science. It is that three extra sessions per training course is a lot of calendar, and the person who has to arrange them has other work.
This is where a live session tool earns its place. In Session Flo you can build the day-one, day-seven and day-thirty checks as separate quiz activities on the same event, run each one as a five-minute item at the top of a meeting that was happening anyway, and compare the score distributions across the three without exporting anything. The recap goes out automatically, so the follow-up is not dependent on you remembering to write it.
None of that changes the curve. It just removes the administrative friction that stops most teams from ever testing whether their training survived contact with the following month.
Frequently asked questions
Is the forgetting curve still considered valid research?
The general shape has held up well: rapid early loss followed by a flattening tail, with each successful review reducing the rate of decay. What has not held up is the precision of the original numbers. Ebbinghaus tested one person on nonsense syllables, and meaningful material that connects to existing knowledge decays more slowly. Treat the curve as a planning shape rather than a formula you can compute against.
How soon after training should the first review happen?
Inside 24 hours if you possibly can, and the next working morning is usually the practical version of that. The steepest part of the decay happens in the first day, so a check run then recovers material while it is still partly available. A review at two weeks is not useless, but it is doing reconstruction work rather than reinforcement, and it takes longer for less benefit.
Does spaced repetition work for skills as well as facts?
Yes, though the retrieval looks different. For factual material a quiz question is the natural retrieval. For a skill, the equivalent is a short scenario or a live attempt with feedback: handle this objection, triage this ticket, run this first minute of the conversation. The principle is unchanged, which is that spaced attempts with feedback beat one long block of practice.
How many follow-up touches are enough?
Three is the usual sweet spot for a half-day of workplace training: one at a day, one at a week, one at a month. The first two do most of the work. Beyond three, returns fall off quickly unless the material is genuinely critical, in which case a fourth touch at ninety days is worth booking. More important than the count is that the intervals expand rather than repeat.
What if people complain that the quizzes feel like school?
Usually the objection is about stakes rather than about testing. Keep the checks unscored against individuals, show the group distribution instead of a name-by-name table, and spend the discussion on the question most people got wrong rather than on who got it right. Framed as a diagnostic for the trainer rather than an exam for the participant, resistance tends to disappear within two rounds.
Keep reading
- why retrieval practice beats re-reading — The mechanism behind the follow-up quizzes recommended here.
- run a training session that sticks — How to design the original session so there is something worth spacing.
- write quiz questions that actually teach — Question design for the day-one and day-seven checks.
- turn a recap into a retrieval touch — Making the post-session summary do real work.
- Session Flo pricing — Running spaced follow-up quizzes across a month.