Picture a regional sales director in Latin America in the spring of 2021. She can tell you exactly how many video calls her reps made that quarter. What she cannot tell you is whether her reps even know what the training said a good virtual call looks like. The program we describe here ran in 2021.
That was the problem a pharma company's Latin American field force brought to us: the medical visit had moved to a screen, and the call skills built for a doctor's office did not transfer on their own. The company wanted to know, person by person, who had absorbed the principles of its virtual selling model and who had not. Over three days, 725 reps answered 161,000 questions, and 73% of the knowledge gaps they revealed were verified closed before the battle ended. Be clear about what that measures: knowledge of the principles, tested with multiple-choice questions. It does not tell you who can hold a doctor's attention on video. That takes role play or observed calls, and we come back to it below.
The visit moved to video. The skills did not come with it.
By early 2021 remote engagement was no longer an emergency measure. Many doctors had grown used to it, some preferred it, and field forces across the region were expected to deliver a full detail through a laptop camera.
The trouble is that a virtual visit is a different craft. In person, a rep reads the room: posture, the glance at the clock, the moment a specialist leans in. On video, most of that signal disappears, and what remains is compressed into a small rectangle with bad lighting. Presence depends on things reps had never been trained on: camera framing, sound, what the background says about you, how you use your voice and hands when half your body is out of shot.
The company had already run a first chapter of its program on the future role of the rep. This second chapter went deeper into the practical side of the virtual visit, from preparing the platform to the psychology of influence. The content was the company's own training model: its rules on communication, doctor types and influence came from its internal curriculum, drawing on popular frameworks such as the classic persuasion principles and the ethos, pathos, logos triad. We tested whether reps knew that model. We did not validate the model itself, and some of its conventions (for example the share of an interaction it assigns to data) are teaching rules of thumb, not measured findings. The syllabus covered:
- setting up the platform and camera, and making a first impression on screen;
- voice projection, hands and body language;
- storytelling through ethos, pathos and logos, and the elevator pitch;
- reading different types of doctor;
- the classic laws of influence, from authority and consensus to scarcity, contrast and the foot in the door.
The stakes were not abstract. Influence and body language are among the hardest things to audit in a field force. A manager on a ride-along sees one rep on one day. Call counts alone do not show how well reps apply the model. So the company had invested in virtual selling training without any way of knowing whether its principles had even been understood, and by whom. Knowing the principles is the first layer. Applying them on camera is a second layer, and it needs a different kind of evidence.
What we designed, and why a three-day battle
We built one game in Spanish for the whole regional field force: 74 questions across five areas, played as a battle in a synchronised three-day window, with notifications pulling reps back in across the three days.
Why a battle rather than a course or an always-on library? Three reasons.
First, speed of diagnosis. A synchronised window means the whole field force plays at the same time, so the company gets a complete picture of what 725 people know about the model in days, not a slow trickle over a quarter.
Second, repetition with a reason. Reps play one-minute challenges against real colleagues, five questions in sixty seconds. Losing to a peer on a question about camera framing is a far stronger prompt to find the right answer than a red cross on a quiz. Over the three days, the average rep answered 222 questions, which means each question came back several times. That is what turns a quiz into retrieval practice.
Third, a rigorous definition of success. Every gap was tracked individually: one rep, wrong on one question at first attempt. It only counted as closed when that same rep answered it correctly later, verified in the platform. That rule filters out most first-attempt noise (one later right answer does not rule out a lucky guess, which is why repeated attempts matter), and it produces a named list of who still has which gap when the window closes.
There was also a practical reason to keep the window short. Reps were already spending long days on camera. A program that asked for an hour of e-learning would have competed with their calls. A one-minute challenge fits between two of them, and three days is long enough for repetition to work without turning into another standing obligation.
The questions were designed around the misconceptions a trained rep might plausibly hold, not trivia. The wrong options were the tempting ones, the answers someone would give if they half remembered the training.
What happened: 14,868 gaps found, 10,789 verified closed
The 725 reps generated 14,868 knowledge gaps at first attempt. That number is not a failure. It is the point of the exercise: every one of those was a place where a rep would have walked into a virtual visit with the wrong idea, and nobody would have known.
By the end of the three days, 10,789 of them had been verified closed, 73% of the total, meaning the same rep later answered that question correctly within the assessment. Readiness, measured as the share of correct answers, rose from 72% on first attempts to 92% on last attempts. The remaining 4,079 open gaps went to the company as a follow-up list for coaching, each tied to a named rep and a specific topic.
The five areas add up to the 14,868 total, and closure was consistent across them, which tells you the format worked on knowledge about soft skills as well as on technical content:
- digital skills and the sales process: 3,702 gaps, about three in four closed (the highest of the five);
- the virtual visit itself (platform, prioritisation, framing, appearance, common errors): 3,068 gaps, 72% closed;
- storytelling and voice: 2,914 gaps, 73% closed;
- reading the customer (first impression, doctor typology, body language, hands): 2,910 gaps, 68% closed;
- laws of influence: 2,274 gaps, 73% closed.
The lowest closure was in reading the customer. That is not surprising. It is the area most dependent on judgement and least reducible to a rule.
Three days of play gave the company a named list of who knew the principles of a good virtual visit, and who did not. Whether they could apply them on camera was a separate question.
Three questions that show how the gaps behaved
The averages tell you the program worked. The individual questions tell you why.
The number that was overestimated, and then corrected
One question asked reps what share of an interaction the logos, the data and figures, carries in the storytelling model they had been taught (the model's own convention, not a research finding). At first attempt, 247 reps picked an inflated answer, doubling the weight of the data. It is a revealing mistake for a pharma field force: of course reps assume the data is most of the story. 360 people got the question wrong first time, and 322 of them, 89%, closed it. By the end, only 25 were still picking the inflated figure.
The technique that looked like the answer
Another question asked which technique belongs in prioritising a virtual visit. The correct answer was handling objections. At first attempt, 200 reps chose a stage or performance technique instead. That matters in practice: a rep who thinks the virtual visit is a performance will deliver a polished presentation and leave the doctor's real resistance unanswered. Of the 372 reps who were wrong first time, 316 closed the gap, 85%, and the performance option fell from 200 to 33.
The one that repetition only half fixed
A third question was about mood before a virtual call and what it actually changes. The correct idea is that a good mood helps you read the other person's behaviour more generously. Many reps confused it with motivating yourself or relaxing the atmosphere. Of the 371 people wrong at first attempt, 217 closed it, 58%, and 154 were still wrong at the end. That is exactly the kind of gap that needs a manager, not another round of questions: the rep has a coherent but wrong model, and the right answer feels counterintuitive until someone explains it.
Lessons we took from it
Five things stood out, and they have shaped how we design programs since.
- Knowledge about soft skills can be measured if you ask about decisions, not definitions. The questions that worked best put the rep in a situation and asked what to do. The ones built on recall of terms were easier and taught less. Neither kind shows the rep can do it on a live call.
- Most gaps close with well-designed repetition. Nearly three in four did here, in three days. That frees managers from having to coach everything, so they can focus on what repetition cannot fix.
- Some misconceptions do not move by repetition alone. The mood question closed barely more than half. When a wrong answer rests on a plausible belief, the fix is a conversation. The value of the data is that it tells you which conversation, and with whom.
- The open gap list is the most valuable output. The closure rate goes in the report. The 4,079 named open gaps go to the managers, and that is where the program keeps paying after the battle ends.
- Engagement comes from the format, not the reminders. Notifications helped bring reps back, but an average of 222 answers per person in three days is not something you get by nagging. People came back because a colleague had challenged them and they wanted to win the next round.
What we would do next
If we ran this program again today, we would keep the format and make four changes.
We would pair the knowledge battle with observed practice: a short recorded role play or a manager's review of a real virtual call, scored against a simple rubric (see the box). Knowledge results and observed skill are two different kinds of evidence, and a report should show them side by side, never merge them into one score.
We would put a short diagnostic before the training, not only after, so the company could see which principles the field already knew and spend training time only where it was needed.
We would schedule a second, shorter battle four to six weeks later on the questions that closed least, to check retention. Three days shows the right answers were reached. It does not show they lasted.
We would also give first-line managers a simple view of their own team's open gaps with a suggested coaching point for each, ideally before their next joint review of virtual calls. The data existed. Getting it into managers' hands faster would have multiplied its value.
Virtual visits are now simply part of how field forces work, alongside face-to-face calls. The question this program answered in 2021 still applies to any new model a field force is asked to adopt: first "do they know it?", then, with a different method, "can they do it?" If you are asking the first question before your next cycle meeting, it is the kind of problem Atrivity was built to answer. The second one still needs a manager watching.