Picture a medical delegate in Abidjan or Douala walking into a paediatric clinic with ten minutes of the doctor's time. The conversation is not about price or promotion. It is about science: how protein quality differs between sources, what human milk oligosaccharides do in an infant's gut, how the composition of breast milk changes as lactation goes on. If the delegate gets that science wrong, the doctor notices, and the delegate's credibility with that clinic is gone for months. The program in this case ran from July 2024 to May 2026.
Now add one more fact. The team that has to hold that science turns over. New delegates and nutrition advisors join every year, across a region of many countries and two working languages. A single training program, however good, would be out of date in its audience within twelve months.
This case is about how an infant nutrition business covering West and Central Africa answered that problem. Instead of one certification, it ran the same science four times over 22 months, with 2,006 participant places across editions, and verified the closure of 61% of the 26,822 knowledge gaps it found within the assessment. The more useful part of the story is where the other 39% sat, and why an average can hide the questions that matter most.
A scientific portfolio, a rotating team and a high bar
The client was the infant nutrition arm of a global nutrition company, with a regional team of medical delegates and nutrition advisors working across West and Central Africa in French and English. Their customers are health professionals: paediatricians, general practitioners, midwives, pharmacists. The portfolio includes infant formula and a range sold through the pharmacy and medical channel.
In this category, the conversation with a health professional is expected to be scientific first. Breastfeeding comes before any product discussion, and the rules on how breast-milk substitutes are presented are strict. A delegate who cannot explain the science credibly has little else to offer.
Three pressures shaped the brief:
- The science is genuinely hard. HMO families and their function, microbiota and immune development, protein and nutrition, breast milk composition: these are topics where a half understanding is easy to fall into and hard to spot.
- The team rotates. People who were certified last year may have left, and new joiners need to reach the same level quickly.
- The region is wide. Many countries, two languages, and a field that spends most of its time on the road rather than in a training room.
The client's question was not "can we train people once?" It was "can we keep a changing team certified on the same science, and prove it?"
What we designed: one curriculum, run as a recurring certification
The answer was not a bigger program. It was the same program, run on a schedule.
Science first, portfolio last
The curriculum was built in blocks, in a deliberate order: nutrition and protein, HMOs, microbiota and immunity, breastfeeding and breast milk, and only then the pharmaceutical and infant formula range. A smaller set of other modules completed it. In total, 369 questions were in play across the program.
The order was a statement about what the delegate is for. If the science is solid, the product conversation follows naturally and stays within what can be said. If the science is weak, no amount of product knowledge makes up for it.
Twice a year, in short windows
Four editions ran between July 2024 and May 2026, roughly twice a year. Each edition was a battle in a synchronised window of 5 to 14 days, with notifications to bring people back in. Delegates played one-minute challenges against real colleagues across the region, in French or English.
Short windows had a practical reason. A field team on the road will make time for a focused competition that lasts a week or two. It will not make time for a curriculum that drags on for a quarter.
The same blocks, every edition
This was the key design choice. The same science blocks came back in each edition. For people who stayed, it meant meeting the same science again at a roughly six month rhythm. We chose that rhythm to fit the team's calendar and turnover, not because six months is the right interval for everyone; a team with faster turnover or faster-changing content would need a shorter one. For new joiners, each edition was a full certification on the same standard as their colleagues. And for the client, it made the trend readable: the same topics, measured on comparable questions, edition after edition.
Throughout, a gap was defined strictly: one participant, wrong on one question, at the first attempt. It counted as closed only when that same participant answered it right on a later attempt, verified in the platform.
What happened: 61% closed, with every block and every edition in a narrow band
Across the four editions, there were 2,006 participant places. That figure counts each person once per edition they played, so people who stayed on the team are counted more than once; the records we used for this case do not split repeat participants from new joiners. The largest single edition, the second one in 2024, had 596. Together they recorded about 448,000 answers.
At first attempt, the program identified 26,822 knowledge gaps. Of those, 16,384 were verified closed on a later attempt: 61%. Readiness, comparing first and last attempts, rose from 87% to 95%. The remaining 10,438 gaps were flagged for follow-up coaching.
The consistency is what stood out. By block, closure ran from 58% to 67%:
- Nutrition and protein: 6,040 gaps, 58% closed.
- HMOs: 5,415 gaps, 67% closed.
- Microbiota and immunity: 5,234 gaps, 60% closed.
- Breastfeeding and breast milk: 4,846 gaps, 61% closed.
- Pharmaceutical and infant formula range: 4,008 gaps, 58% closed.
- Other content blocks: 1,279 gaps, 67% closed.
By edition, closure ran from 58% to 71%. Closure rates varied between 58% and 67% by block and between 58% and 71% by edition. For a recurring certification with a rotating audience, that stability is what the client needed to see. It is worth being precise about what it shows, though. Similar closure rates on the same curriculum tell us each edition worked on its own terms, with whoever was on the team at the time. They do not prove that the people who played in 2024 retained the science into 2026, because the editions were not a matched cohort of the same people.
Inside the averages: three questions tell a different story
Averages are useful for a steering committee. They are less useful for a manager deciding what to coach. So we looked at specific questions, using the 596 participants of the largest edition.
Two of the three were true or false items, so a caveat first: on a two-option question, one later right answer does not rule out a lucky guess. We read these as corrections within the assessment, not as proof that the concept is understood.
The first was a true or false item on whether the stage of lactation affects the quantity of HMOs in breast milk. At first attempt, 61 people got it wrong. Repetition fixed 26 of them. Thirty-five were still wrong at the end: 43% closed.
The second was also about HMOs: whether they are fully digested in the digestive tract. Thirty-nine were wrong at first attempt, 17 corrected it, and 22 were still wrong. Again, less than half closed.
The third was practical: the amount of water to use per spoon when preparing a formula. Thirty-five were wrong at first. Twenty-seven fixed it through repetition, leaving eight: 77% closed.
The HMO block had the highest closure rate in the program, and it also held two of the most stubborn misconceptions.
Look at that contrast. The HMO block as a whole closed 67% of its gaps, the joint best in the program. Yet two of its core concepts resisted repetition, both times on questions the delegate is expected to explain to a health professional, not just repeat. Meanwhile, a practical preparation instruction, the kind of thing the delegate has to get exactly right when advising on formula use, closed well with repetition alone.
That pattern is common in scientific content. A fact with a single right answer (a volume, a dose, a schedule) tends to yield to practice. A concept that sits on top of a mental model (what HMOs are, what happens to them, why their quantity changes) does not, if the mental model is wrong. People keep choosing the same wrong answer, confidently, because it fits what they believe.
Five lessons from four editions
1. For a rotating team, recurrence is the strategy
A one-off certification certifies the people who were there. A recurring one measures the team, whoever is on it, against the same standard each time. Running the same science twice a year turned certification from an event into a routine, and new joiners went through the same path as everyone else.
2. Keep the blocks stable so the trend means something
Because the same blocks came back each edition, the client could compare like with like. A drop in one block in one edition would have been visible and actionable. Constantly changing the curriculum feels fresh, but it destroys the ability to see progress.
3. Put scientific understanding before product discussion
Putting breastfeeding, nutrition and microbiota ahead of the product range was not only good pedagogy. It matched what health professionals in this category expect, and it kept the product conversation grounded in what can responsibly be said.
4. Look below the block average
The best-performing block contained two of the most resistant misconceptions. If the client had only read block averages, it would have concluded HMOs were fine. Per-question tracking of the same people over time is what showed otherwise.
5. Some gaps need a conversation, not another question
These 22 unresolved gaps warrant follow-up assessment and an explanation from the medical team. They need a short, well prepared explanation from a medical colleague or a manager, ideally with the mechanism drawn out. That is why the 10,438 open gaps were flagged by person and topic for coaching, rather than simply rolled into the next edition. Six months is too long to leave a critical misconception with someone who is advising health professionals every week.
What we would do next
If we were planning the next cycle with this team, we would keep the recurring structure and the stable blocks, and change four things.
First, we would give managers and medical colleagues a short list of the concept-level misconceptions, by person, before each edition closes, with a brief explainer for each, and ask for the conversation within two weeks rather than at the next edition. The question data is precise enough to target those conversations to a handful of people per country.
Second, we would add a small number of new questions on the resistant concepts that approach the same idea from a different angle. If a person answers the true or false correctly but fails a scenario on the same concept, the misconception is still there.
Third, we would track new joiners and repeat participants separately, with a unique count of people across editions. The overall closure rate is stable, but the most useful signal for a rotating team is how quickly a new delegate reaches the level of an experienced one, and whether repeat participants hold their level from one edition to the next.
Fourth, we would not make new joiners wait for the next formal edition. Someone who joins a month after an edition closes would otherwise go five months without being tested. A short onboarding set drawn from the same blocks, played in their first weeks, would give them the same standard earlier and tell their manager where to start.
Why this format suits scientific selling
The program worked because it matched the way the job works. Delegates in this region hold a body of science that does not change much year to year, in front of health professionals who expect it to be right. What changes is the team. A recurring, short, competitive format with strict gap tracking gave the client a way to hold the standard steady while the people changed around it.
At Atrivity, this is the pattern we look for when a client describes a team that turns over and a curriculum that stays: battles in short synchronised windows, Gap Fixing that sends each person the fix for their own mistakes, and readiness by team, country and topic. The question we would leave with anyone running a similar team is simple: if half your delegates changed next year, how would you know that the new ones hold the science as well as the ones who left?