In six days, 1,330 people in a global immunology kickoff answered 301,000 questions. That's more than 200 answers per person. In programs like this, many people play against each other in the gaps between sessions, on the coach, after dinner, and the pull of a rematch does more than any reminder email.
That's what we mean when we say gamification can be a measurement instrument. Our thesis is blunt: for a launch, the point of a game is not points or badges, it's getting an honest, complete reading of what the whole field force knows, in one week, and then fixing it. A leaderboard bolted onto an e-learning course won't do that. A well designed battle will. Here's why, what the numbers look like across our case library, and where the approach has limits.
Most gamification in pharma is decoration on a course
Gamification has a reputation problem, and it's mostly earned. In a lot of organisations it means the same compulsory module with a points counter in the corner and a badge at the end. Completion goes up a little. Nobody learns much more. Senior sales leaders, quite reasonably, file it under "nice for engagement" and move on.
The trouble is that this kind of gamification borrows the surface of games (scores, levels, trophies) and ignores the mechanics that actually make people play: a real opponent, a short and intense format, the pull of a rematch. Without those, you get an LMS in fancy dress.
It also borrows the weakest habit of many LMS set-ups: measuring completion. For a launch, completion tells you almost nothing. A rep who clicked through the product module on the train has "completed" it. Whether they can explain the primary endpoint to a sceptical specialist in week three is a different question entirely, and it's the only one the brand team really cares about.
So the design question isn't "how do we make training more fun?" It's "how do we get almost everyone to answer hundreds of real questions, with real effort, in the same few days, so we can see what they know?" Fun is the means. Data is the end.
What a battle actually is
An Atrivity battle is built from one-minute challenges: five questions in sixty seconds, played against a real colleague. You challenge someone, you both answer the same five questions, the faster and more accurate player wins. Then you want another go. A battle runs these challenges across the whole field force in a synchronised window, usually a few days.
Simple as it looks, it works because it passes five tests that a typical course fails.
- A real opponent. Playing against a named colleague, not a machine, is what brings people back. Losing to the rep in the next territory is surprisingly motivating.
- A synchronised window. Everyone plays in the same days, on the same content, so the snapshot is fair and comparable across teams and countries.
- Mass participation. Participation is high, whether the program is an expected part of the year or opt-in, so the data describes the field rather than the diligent few.
- Volume. Hundreds of answers per person in a few days, enough to see each individual's gaps, not just a team average. Repeated answers from the same person on the same question are not independent measurements, though: they show one person's trajectory, not hundreds of separate readings.
- Data exhaust. Every answer is recorded, so the game produces a readiness map as a by-product of people enjoying themselves.
The sixty-second limit matters more than it looks. In the field, a rep has seconds, not minutes, to recall the right number or the right distinction in front of a physician. A timed format tests retrieval under mild pressure, which is closer to a real call than an untimed quiz you can pause to check the visual aid.
The timer has a cost, and it is worth naming. A wrong answer under time pressure can reflect reading speed, working in a second language, or an accessibility need, not a lack of knowledge. We run content in each player's own language where the client has approved it, and we read first-attempt errors with that in mind: the more telling signal is whether the same person gets it right on later attempts. If a group is consistently disadvantaged by the clock, give it untimed practice and do not rank it on speed.
High participation, with real effort, is the whole trick
The most useful property of a battle is not that it is fun. It is that almost everybody plays, and plays to win.
In a global MedTech wound care program in our case library, which ran in June 2026 with reinforcement to the end of August, 787 people were invited across 25 countries and 3 channels, and participation reached 93%. To be precise about what that shows: this was the company's annual readiness initiative, an expected part of the year, not voluntary play. The final reinforcement wave was opt-in, and 124 people chose to take part. So the 93% is a measurement taken on almost the entire population, in the same window, not proof that people would have joined unprompted.
Why does effort matter so much? When people click through something they resent, you measure compliance. When they are trying to beat a colleague, they are trying to get the answers right, so the errors are more likely to be honest ones.
When people play to win, their mistakes are honest. That's what makes the data worth having.
High participation can reduce selection bias, but the missing participants still matter. If only your top performers complete the optional module, your "readiness" figure is a flattering picture of the people who needed it least. With 93% playing, the reading includes the rep in the remote territory who rarely opens email.
What the numbers look like at scale
Back to that immunology kickoff. It ran four franchise games at the same time, 404 questions in total, over 6 days. The 1,330 players produced 301,000 answers, which identified 15,238 knowledge gaps. By the end of the program, 53% of those gaps were verified closed, with franchises ranging from 47 to 58%.
That spread across franchises is itself useful. Same company, same week, same format, and yet one franchise closed noticeably more of its gaps than another. A course completion report would have shown four green ticks. The battle data showed the brand leads exactly where to look.
The MedTech readiness program used a different rhythm: a 4-day battle followed by 9 weeks of reinforcement. It recorded 15,758 gaps and closed 57% of them.
At the level of the whole case library (100 programs, 39 organisations, 2016 to 2026), we recorded about 9 million answers and identified 794,130 gaps at first attempt. Of those, 345,366 were verified closed, about 43%. Closure across programs ranged from 18% to 80%.
That range is, in our view, the argument for design. The game mechanic is the same everywhere. What differs is how it's used: when it runs, what follows it, whether anyone acts on the results. We cannot isolate each factor, but those are the ones we see move with closure.
A gap only counts as closed when the same person gets it right later
A battle finds gaps. It doesn't fix them by itself. The measurement is only worth something if there's a loop behind it, and the loop needs a strict definition.
We count a gap as one person getting one question wrong at the first attempt. It closes only when that same person answers that same question correctly later. Not the team average rising, not a different rep getting it right. The same individual, the same item.
The loop runs in three steps. The battle identifies the gaps. Gap Fixing then sends each person the explanation for their own mistakes, not a generic recap of the whole module. Then later waves check whether the fix held. For critical items (the claims, safety messages and differentiators that really matter) we set a stricter bar: a critical gap counts as closed only at 5 of 5 later check questions on the same concept. One later right answer does not rule out a lucky guess; the stricter bar makes it much less likely that luck closes a critical gap, though no rule removes it entirely. And a closed gap is a verified correction within the assessment, not proof of how someone performs in a call.
Timing adds a lot. The 16 programs in our library tied to a business moment (a launch, kickoff, certification, congress or cycle meeting) closed 55% of their gaps, against 41% for the other 84. That is an observed association, not proof that the moment did it. Our reading is that people study harder when the knowledge is about to be used, and a launch is the most concrete "about to be used" there is.
The best example in the library is a respiratory launch in Japan: 257 players, two waves (disease first, then product), 32 days, 80,000 answers and 4,503 gaps, of which 80% were closed. Readiness went from 75 to 95%. The two-wave design meant the product story was introduced after the disease story, which we think helped, though the program alone cannot prove it.
Content alone moves far less than competition
If the game is just a delivery vehicle, you'd expect content to do the heavy lifting and the competition to be optional garnish. Our data points the other way.
In a consumer health program for the pharmacy channel, 1,421 staff joined an always-on platform over 139 days, covering 5 brand cycles. Closure was concentrated in the competitions. Content on its own, available and well produced, moved far less.
That fits what we know about memory. Murre and Dros's 2015 replication of Ebbinghaus, a relearning experiment with nonsense syllables, shows that unrehearsed material fades fast and that spacing helps. Reading something once is fighting that with one hand. Being asked to retrieve it, repeatedly, with a colleague waiting for your answer, is a very different exercise.
It's worth being honest that this isn't a clean experiment. Pharmacy staff who join competitions may already be more engaged. But the pattern shows up often enough across programs that we design around it: content prepares, competition consolidates.
A fair objection: isn't this just a quiz with a timer?
A fair objection is that none of this is deep learning. Five questions in sixty seconds tests recall, not the ability to hold a nuanced scientific conversation with a KOL. Real selling is about judgement, not multiple choice.
We agree with half of that. A battle doesn't replace coaching, role play or a good ride-along. It isn't meant to. What it does is tell you, before those expensive activities, which facts the field can't reliably retrieve. You can't have a nuanced conversation about a trial if you can't remember its primary endpoint. The battle clears away the basics so coaching can focus on judgement.
And the battle is also very good at showing where coaching is needed. In one in five of the example questions we tracked, fewer than 20% of the people who were wrong corrected it; some of those were never retested, so not all are proven misconceptions. Where the wrong answer keeps coming back, it usually is one. In the immunology kickoff, a mechanism misconception held by 125 reps barely moved however many times it came up. More rounds are unlikely to fix that. A manager who knows exactly which reps hold it, and sits down with them for ten minutes, has a much better chance.
So the honest positioning is this: the battle measures and fixes the knowable, and it hands managers a precise, short list of what needs a human conversation.
What to ask before your next kickoff
If a battle is on the table for your next launch or kickoff, the questions worth asking in the planning meeting are not about badges or prizes.
- Will almost everyone play, in the same window, so the reading is fair across countries?
- Will each person receive the fix for their own mistakes, and will we check that it held?
- Who will act on the misconceptions that repetition doesn't shift?
That's how we design battles at Atrivity: one-minute challenges against real colleagues, a synchronised window, Gap Fixing for each person's own errors, and readiness indexes by team, country and topic that sales and medical leaders can actually use. Fun enough to play, rigorous enough to trust.