About two thirds of new drugs missed their pre-launch consensus sales expectations in their first year. That is the headline finding of McKinsey's analysis of 210 launches between 2003 and 2009, published in 2014. The cohort is historical, but the uncomfortable part still resonates: of the launches that start behind, 78% are still behind in year two, and 70% of those are still behind in year three.
The first twelve months are not a warm-up. They are the launch. Many things shape year one, from access and pricing to competition and the data itself. Of the ones a commercial team controls, the one most launch teams still do not measure is what each person in the field actually knows, question by question, from week one. Measure it and you get three things: an early warning that arrives before the sales data, a precise list of what to fix, and a way to check that it was fixed. The rest of this piece is what that looks like across 100 programs, and what to put in place in the ninety days either side of launch.
Year one is the verdict, not the rehearsal
Launch plans are usually written as if the first year were a ramp. Uptake is expected to be slow, formulary wins take time, the specialists who matter will come round once the data settles in. The McKinsey numbers, for that 2003 to 2009 cohort, describe something harsher than a ramp. They describe a groove.
Look at the other side of the same research. Of the launches that outperform in year one, 65% keep outperforming in year two, and 53% of those are still ahead in year three. Momentum compounds in both directions. A launch that starts strong earns early prescribers, early advocacy and early payer confidence, and each of those makes the next quarter easier. A launch that starts weak spends year two explaining year one.
That matters more now than it did a decade ago. IQVIA estimates that 73 novel active substances launched globally in 2025, above the average of the past ten years, and its Launch Excellence work puts roughly $200 billion of revenue at risk from loss of exclusivity by 2030. Companies need more launches, and they need them to land well consistently, not occasionally. There is less room for a slow first year than there used to be.
So the useful question is not "how do we recover a weak launch?" It is: what decides which groove a launch falls into, and which of those things can we still influence in the weeks that count?
Launch drift starts in the field, not in the forecast
A July 2026 analysis in PharmExec gives the problem a useful name. Launch drift is the gap between the go-to-market strategy and how it is actually executed in the field. The strategy on the slide and the conversation in the clinic slowly come apart, and nobody notices until the numbers do.
Among the root causes it lists are strategy-execution gaps, variability in field execution, HCP belief barriers and payer friction. Payer friction is largely outside the rep's hands. The other three are not. They all come back to a very ordinary moment.
Picture week three of a launch. A rep is in front of a specialist who has read the headline results and is mildly sceptical. The specialist asks one precise question about the primary endpoint in the subgroup they treat most. The rep has ninety seconds. Either they give the answer exactly as it appears in the approved materials, calmly, without hedging, or they round it, soften it, or reach for a line from the competitor's story that got lodged in their head during the kickoff.
That moment happens thousands of times in the first quarter. Together, those moments are a large part of the launch.
PharmExec's advice on timing is worth taking seriously too: monitor from week one, expect early trends to show at weeks four to six, and look for confirmation around month three. That is a short window. If the first reliable signal of drift arrives at week six, and your first reliable signal of field knowledge arrives at the end-of-year certification, you are measuring one of the inputs long after the outcome.
If you only learn what the field knows at the end-of-year certification, you are measuring one of the inputs long after the outcome.
Your launch dashboard measures effort, not knowledge
Ask most launch teams how ready the field is and you will get three numbers. Attendance at the kickoff. Completion of the e-learning modules. The pass rate on the final test.
All three can look excellent while the knowledge that matters is missing. Attendance tells you people were in the room. Completion tells you they clicked through. A pass mark tells you that, on the day, across a set of questions, the average was good enough. None of them tells you that the rep in the scene above will give the right number for the right subgroup when it counts.
There is also a timing problem. Murre and Dros's 2015 replication of Ebbinghaus, a single-participant relearning experiment with nonsense syllables, illustrates forgetting in a specific experimental task. It says nothing precise about how much of a product story a rep retains, and we would not stretch it that far. The direction is enough. What was learned at the kickoff on Tuesday is not what will be said in the clinic three weeks later unless something in between brings it back.
Across 100 programs, the gaps are specific, numerous and invisible
We have been measuring that second question for a decade. Across 100 programs run with field forces between 2016 and 2026, in 39 organisations, our case library holds about 9 million answers. In them we found 794,130 knowledge gaps at first attempt.
A gap, in our definition, is deliberately narrow: one person, wrong on one question, the first time they see it. It is not a low team average or a weak module. It is a specific fact that a specific person does not have. And a gap only counts as closed when that same person answers the same question correctly later. Across the library, 345,366 of those gaps were verified closed, about 43%.
Three things stand out for launches.
- The spread is wide. Closure ranged from 18% to 80% across programs. Across the dataset, 345,366 of 794,130 first-attempt gaps were verified closed, about 43%. We cannot attribute that spread to any single cause, but in our experience design choices account for a good part of it, and the same field force can end up very ready or barely ready depending on how the program is built.
- The gaps are not rare. Across programs with gap fixing, the average is about 16 critical gaps per rep per year. Sixteen facts that matter, per person, that many LMS set-ups, built around completions, do not show.
- The gaps cluster. The same wrong answer tends to be held by many people at once, often because it was taught ambiguously or because it sounds more plausible than the truth.
That last point is where launch drift and field knowledge meet. When fifty reps share one wrong belief about a mechanism or an indication, you do not have fifty individual lapses. You have one strategic message that did not land, repeated fifty times a day.
A launch is the best moment you will ever have to close gaps
Here is the good news. In our data, programs tied to a business moment closed more of their gaps.
In our case library, 16 programs were built around a specific moment: a launch, a kickoff, a certification, a congress or a cycle meeting. Those programs closed 55% of their gaps, verified person by person. The other 84 programs, the always-on general knowledge ones, closed 41%. That is an observed association, not proof that a deadline causes closure: the moment-based programs also differed in content, audience and design.
Our reading is that a launch gives people a reason to care that has nothing to do with training. Managers are paying attention, the content is new, and there is a date after which being wrong has a visible cost. For once, the field pulls for the information.
One example from the library makes the point. A respiratory launch in Japan ran with 257 players in two waves, first the disease area and then the product, over 32 days. It recorded around 80,000 answers and identified 4,503 gaps. Eighty percent of them were verified closed within the assessment, and the readiness index moved from 75 to 95%. The design choice we would single out was the sequencing: nobody was asked to master the product story until the disease story underneath it was solid. We cannot prove it was the sequencing that did it, only that the program was built that way.
A launch is not the moment to accept an average closure rate. It is a moment when a better than average result is realistic, if you measure.
A fair objection: launches are won on access and relationships
A fair objection is that none of this is what really moves year one. Launches are won on access, on formulary position, on medical education, on the KOLs who shape opinion, and on relationships that reps have built over years. Knowledge, the argument goes, is table stakes. Every rep gets trained. What separates launches is everything else.
We agree with most of that. Access is decisive, and a well-informed rep cannot sell a product the payer has not listed. But the objection assumes that "every rep gets trained" means "every rep knows". That is exactly the assumption our data says is wrong. Gaps are held by people who have, on paper, been trained, and many of them sit on the facts that matter most: claims, safety, the differentiators the whole positioning rests on.
Relationships also cut the other way. A specialist who trusts a rep will forgive a lot, but not a confident wrong answer on the one number they care about. Trust built over years can be spent in a single call.
So knowledge is not the whole launch. It is the part of the launch that is most controllable, most measurable and, in our experience, least measured.
What to put in place in the ninety days around launch
None of this requires rebuilding the launch plan. It requires adding a measurement layer to it, and acting on what it shows. In practice that means a handful of concrete moves.
- Diagnose before the kickoff. Run a short diagnostic challenge on the launch content two or three weeks before the plenary, so the kickoff agenda spends its time on the topics where the gaps actually are, not the ones the brand team is proudest of.
- Flag what is critical. Not every question is equal. Mark the ones where a wrong answer is a credibility, compliance or safety risk, and weight readiness around those.
- Track readiness like a commercial KPI. Readiness by country, team and topic, reviewed weekly in the first three months, on the same page as the early uptake indicators. If PharmExec is right that early trends appear at weeks four to six, you want the knowledge signal on the table at the same time.
- Close gaps individually. A gap is closed when that person answers correctly later, not when the whole team has seen the slide again. Send each person the fix for their own mistakes, and check again.
- Hand managers the misconceptions. Some wrong beliefs do not move with repetition. Give first-line managers the short list of people and topics before their next ride-along, so the coaching conversation happens where it is needed.
Launch teams that do well tend to know what their field knows
We have run programs for launches that went brilliantly and for launches that struggled, and we are careful not to claim that field knowledge explains everything. It does not. Our case library measures knowledge, not sales, so it cannot tell you which launches won. What we can say from experience is that the launch teams we have seen do well tend to know, early and precisely, what their field knows, and treat it as something to manage rather than something to assume. Knowing the approved answer is also not the same as delivering it well in a call; that needs ride-alongs and call observation.
That is the reason we built Atrivity the way we did: one-minute challenges that reps play against real colleagues on their phones, battles in a synchronised window around the kickoff or the cycle meeting, and Gap Fixing that sends each person the fix for their own mistakes and only counts a critical gap as closed after repeated correct answers. Leaders see readiness by team, country and topic from the first week, not the first certification.
Whatever tool you use, take one question into your next launch meeting: if the first real signal of drift shows up at week six, when will we see the first real signal of what our field knows? If the honest answer is "later than that", you have found the gap worth closing first.