Picture the Monday after a national cycle meeting. Two days in a hotel ballroom, a new visual aid, a refreshed POA, a medical update on the latest data and a workshop on the three objections everyone expects. By Thursday, a rep is in a specialist's office and the physician asks about the primary endpoint in one subgroup. The rep knows they saw that slide. They cannot quite remember what it said.
Nobody in that room did anything wrong. The training was well made, the content was approved, and the rep was paying attention. The problem is the design. Most sales training in Life Sciences is built as an event, and memory does not work in events. What makes training stick is retrieval, spaced out over weeks, measured person by person, with coaching aimed at the errors that repetition will not fix. This post explains why, and what that looks like in practice for a field force.
Unrehearsed material fades fast, and most training plans ignore it
In 1885 Hermann Ebbinghaus measured how quickly he forgot lists of nonsense syllables. In 2015, Jaap Murre and Joeri Dros replicated the experiment and published it in PLOS One. Like the original, it was a relearning experiment with a single participant. Measured as savings (how much less effort it takes to relearn a list), savings after 31 days were down to about 4%.
A fair objection is that nonsense syllables are not a clinical story. A rep's knowledge of their own product is meaningful, connected and used every day, so it should decay more slowly. That is true, and we are careful about transferring a laboratory finding like this one. It does not mean reps forget 96% of a cycle meeting in a month, and it gives no rate for clinical content at all. What each source supports is narrower: Murre and Dros illustrate forgetting in their experimental task, and Dunlosky and colleagues review the evidence for practice testing and distributed practice. In our experience the details are what slip first, and in pharma the details are the job: the dosing in renal impairment, the exact wording of the indication, the number that belongs to the secondary endpoint and not the primary.
The stakes are rising too. IQVIA's Launch Excellence research estimated about $200 billion of revenue at risk from loss of exclusivity by 2030, and argued that companies will need an unprecedented number of launches that perform consistently well. The FDA's drug centre approved 46 novel drugs in 2025. More launches means more new content landing on the same field forces, and less time between one learning event and the next.
Three habits that quietly make training fade
We have run many programs with field forces, and the same habits come up again and again. None of them are foolish. They are just built around the calendar of the training team rather than the memory of the rep.
- Everything is packed into the kickoff. The biggest learning dose arrives on the day with the most competition for attention, and then nothing follows it.
- Completion is treated as learning. The LMS shows 98% of reps finished the module. It says nothing about whether they could answer the question a KOL will ask next month.
- Rereading stands in for practice. Reps are sent the slide deck again, or a summary PDF, and asked to "review it before visits". It feels productive. It isn't very.
That last one is backed by one of the most useful reviews in learning science. In 2013, John Dunlosky and colleagues assessed ten common study techniques in Psychological Science in the Public Interest. Two came out as high utility across the learners and materials studied: practice testing and distributed practice. Rereading, highlighting and summarising came out as low utility, despite being what most people actually do.
The rep who rereads the deck feels ready. The rep who has to answer the question cold finds out whether they are.
Retrieval beats review, and the gap grows with time
The clearest single experiment on this is Henry Roediger and Jeffrey Karpicke's 2006 paper in Psychological Science, "Test-enhanced learning". Students read prose passages and then either restudied them or took tests on them. Five minutes later, the restudy group looked better. Two days and a week later, the testing group remembered more, and the advantage grew with the delay. The restudy group also felt more confident than their results justified.
Translate that to a field force and it looks a lot like the cycle meeting problem. Reviewing the material produces a comfortable sense of readiness. Being asked to retrieve it, and being told when you are wrong, feels harder and, in these studies, worked better. One caution: this research was done with students and study materials. It supports the design principle; it does not validate any particular vendor's commercial results, ours included.
Retrieval is not the same as a quiz at the end of a module, though. Three conditions matter:
- It has to be spaced. One test right after training mostly measures short-term memory. The value comes from coming back to the same point days and weeks later.
- It has to be short. A rep between two visits has a few minutes, not forty. Five questions in a minute can be done in a car park.
- It has to give feedback on the specific mistake. Knowing you scored 7 out of 10 does nothing. Knowing you mixed up the primary and secondary endpoints, and seeing the correct one, does.
Measure gaps per person, not scores per team
Here is where most training measurement falls down. A team average of 82% tells a sales director almost nothing useful. It hides the rep who has the dosing wrong, and the six who share the same misconception about a competitor.
We measure something narrower and more demanding: a knowledge gap is one person wrong on one question at their first attempt, and it only counts as closed when that same person answers it correctly later. It is a strict definition, and it changes the conversation. Instead of "the team scored well", you get "these 44 people had this fact wrong, 26 of them have now fixed it, and these 18 have not".
Those figures are real. In a diagnostics program we ran from 2024 to 2025, a specialised sales team took part in short competitive capsules over 16 months. Across the program, half of the 6,621 gaps found at first attempt were verified closed within the assessment. On one question about the number of tests in a new assay pack, 44 people were wrong at first; repetition fixed 26 and left 18. A different pattern showed up in objection handling: those questions closed at 29%, far below the fact-based blocks. In that program, facts responded to repetition better than knowing the right response to an objection, which more often needed a conversation.
A very different program shows what good looks like when retrieval is anchored to a clear goal. An infant nutrition certification in Latin America, run in May 2025, asked a medical and nutrition team to demonstrate command of a single scientific subject before speaking with health professionals. It closed 73% of its gaps. We think one subject, studied deeply, with a certification at stake, gave people a reason to come back, though a single program can't prove that. And a closed gap there means a later correct answer in the assessment, not proof of what happens in a clinic.
Not every gap is a memory problem
This is the part most retention advice skips. Spaced retrieval is powerful for facts that have simply faded or never landed. It is much weaker on errors that are not forgetting at all, but belief.
Some wrong answers are sticky because they make sense to the person holding them. A rep who learned a mechanism of action slightly wrong at a previous company, or who absorbed a competitor's framing from a KOL, will often pick the same wrong answer again and again. Showing them the right answer a fourth time does not help much. They need someone to explain why their version is wrong.
So a well-designed program separates two kinds of gap:
- Gaps that close with repetition. Leave these to the system. Send the fix, bring the question back, check it again.
- Gaps that resist repetition. Escalate these to a person: a first-line manager on a ride-along, an MSL in a team call, a trainer at the next cycle meeting.
The second list is usually short, and it is gold. It tells a manager exactly what to talk about with exactly which rep, which is a far better use of a ride-along than a generic "how are things going".
Design training around the moments reps already care about
One more thing makes a big difference, and it is not a learning technique. It is timing.
Reps pay attention when something is at stake: a launch, a new indication, a congress where the key trial is being presented, a certification they need to pass. In our experience, programs tied to a moment like that get more second attempts than programs that arrive on a generic training calendar. That is an association we have observed, not a proven cause. The retrieval science is the same. The motivation to come back seems not to be.
That suggests a simple planning rule. Don't ask "when is the training slot?" Ask "what is the next moment in the field when this knowledge will be tested for real?", and work backwards. If a congress is in six weeks, start retrieval rounds on already approved background material; add new findings only after release and the required medical review. Finish with a short competition the week before, and use the gaps that remain to brief managers and MSLs before the booth opens.
It also changes what you do at the cycle meeting itself. If the facts are being learned and checked through short rounds before and after, the live session can be spent on the things that need people in a room: practising the conversation, working through the hard objections, agreeing how Marketing, Medical and the field will talk about a difficult subgroup. That is a far better use of two expensive days.
What sticky training looks like, in practice
Put the pieces together and a field training plan starts to look different from the usual one.
Before the meeting, a few short rounds show what the team already knows, so nobody sits through a module on something they have mastered. At the meeting, time goes to discussion and practice. After the meeting, one-minute challenges come back on the same key points over several weeks, each wrong answer triggering the specific fix from approved content. A few weeks in, the team competes in a short, time-boxed battle to see what has lasted. And at the end, the gaps that did not close go to managers by name, with the correct answer and the source.
This does not have to mean a bigger training budget, but it is not free. Someone has to write and review the questions through MLR, reps give a few minutes a week, and managers spend time on the coaching list. Mostly, it needs the budget and the effort to be spread over six weeks instead of two days. And it needs the organisation to accept a less flattering number: not "98% completed", but "these gaps are still open, and here is who is working on them".
At Atrivity this is how we build programs, from launches to annual certifications: short competitive retrieval, fixes for each person's own mistakes, and a clear line between what repetition closes and what a manager has to. If you are planning your next cycle meeting, try one question in the planning session: which three facts must every rep still have right six weeks later, and how will we know?