Why Most Training Wears Off in 30 Days — And What to Do About It

2026-04-15 · 8 min read · OneRange Team

The forgetting curve is well-documented and mostly ignored by corporate training. How spaced practice fixes retention — and why AI training makes it operationally possible

There's a graph from 1885, originally drawn by a German psychologist named Hermann Ebbinghaus, that shows how quickly people forget what they've learned. Within a day, most of it is gone. Within a month, almost all of it is. The graph hasn't changed. Corporate training, mostly, has not adapted.

What is the forgetting curve?

The forgetting curve describes how retention of newly learned material drops sharply in the hours and days after learning, then continues to decline more slowly. Ebbinghaus mapped it on himself with nonsense syllables; the effect has been replicated across a century of learning research with real material and real learners. The practical shape is what matters: the steepest losses happen immediately, which means the window to intervene is much earlier than most training calendars assume.

The forgetting curve in plain terms

Without reinforcement, learners retain roughly 20-30% of new material a month after a single training session. With well-timed practice, that number can climb above 80%. The technique is called spaced repetition, and it's been validated in education research for over a century.

Two things make spacing work. The gap itself forces retrieval to be effortful, and effortful retrieval is what consolidates memory — which is why re-reading notes feels productive and does almost nothing. And the interval schedule matches the shape of the curve: reinforce just before the material would have faded, and each round buys a longer window than the last.

Why don't companies do it?

Spaced practice is operationally annoying. It requires sending learners back to short, targeted exercises at intervals — three days, ten days, thirty days — and tracking what each learner specifically struggled with. That's a logistical problem at scale, which is why most programs don't bother.

There's a second reason, and it's structural. A training program measured on completion has no reason to care what happens on day 31. The metric closes when the course closes. Nothing in the reporting distinguishes a team that retained the material from one that forgot it, so nothing in the budget or the calendar accounts for decay. It's the completion-rate trap in another costume: what you can't see, you don't fund.

What does a spacing schedule actually look like?

Not complicated, and not a full second course. A workable pattern for a single high-stakes skill:

• Day 3 — one short retrieval exercise on the two or three concepts most likely to be misapplied
• Day 10 — a scenario that requires combining those concepts rather than reciting them
• Day 30 — a brief assessment that also serves as the retention measurement
• Day 60+ — resurface only what that specific learner got wrong

The last line is where most manual programs collapse. Personalizing reinforcement to each learner's specific gaps means tracking, per person, per concept, what was missed and when — which is spreadsheet work that scales badly and gets abandoned by the second cycle.

What changes with AI training

• The system knows which concepts each learner missed and can resurface them automatically
• Reinforcement happens in two-minute chat exchanges, not 45-minute course replays
• Managers see retention data, not just completion data
• Skill scores degrade and recover in ways that match how people actually learn

That third point reframes what a skill record is. A certificate is a permanent claim about a temporary state — which is exactly why certification and proof of skill diverge over time. A skill score that decays without practice and recovers with it is simply a more honest instrument, and it makes decay visible early enough to do something about.

Isn't this just nagging people?

It is if you do it badly. Reinforcement that ignores what someone already demonstrated — sending the whole cohort the same day-30 refresher regardless of performance — reads as busywork, and people opt out. Reinforcement that targets the specific things a person got wrong reads as useful, because it is. The difference is entirely in whether the system knows what each learner missed.

Keep the touches genuinely short. Two minutes of retrieval beats twenty minutes of replay, both for retention and for whether anyone actually does it.

Where to start

Pick one high-stakes skill — one where forgetting carries a real cost. Add a single round of spaced reinforcement at the 30-day mark. Measure retention with a short scenario assessment at 60 days. The improvement will be obvious enough to justify expanding the model.

Measure the control group too. Run the same 60-day assessment on a cohort that got the training but no reinforcement, and the gap between the two groups is your business case — far more persuasive than citing Ebbinghaus at a budget meeting. If you're already instrumented for real-time learning outcomes, you'll see the divergence well before day 60.

OneRange Vero handles the tracking side of this: because every session is interactive and measured, the system knows what each person struggled with and can bring it back at the right interval, without anyone maintaining a reinforcement spreadsheet.

Tags: Skill Development, L&D, Analytics

FAQ

Frequently asked questions

How much training do people actually forget?

Without reinforcement, retention of new material a month after a single session is commonly in the 20-30% range; with well-timed spaced practice it can exceed 80%. The exact numbers vary by material and learner, but the shape — steep early loss, dramatically improved by spacing — is one of the most replicated findings in learning research.

What are the best intervals for spaced repetition at work?

Expanding intervals work well: roughly day 3, day 10, and day 30, then longer gaps for anything still shaky. The principle matters more than the precise days — reinforce before the material fades, and lengthen the gap each time a learner succeeds.

Does spaced practice work for skills, or just facts?

Both, but the exercise has to match. Facts respond to retrieval practice; skills respond to scenario practice that requires applying judgment. The spacing schedule is the same — what changes is that the reinforcement asks the learner to do the thing rather than recall it.