A scenario we see often. The brand team spends the best part of a year building the launch story: a message house with three pillars, in a deliberate order, each backed by a specific data point and approved through MLR. It is presented at the kickoff, printed into the visual aid and loaded into the enablement portal.
Six weeks later, a regional manager rides along with one of her reps. The rep opens with the third pillar, blends the second with a line from last year's product, and skips the first entirely because "doctors don't want to hear about that". Nobody in marketing knows. Nobody in sales thinks anything is wrong.
That gap between the plan and the call is where many launches quietly lose ground. Our argument here is that it can be seen, week by week and country by country, well before sales data can say anything useful about it, and that closing it needs Marketing and Sales to look at the same evidence rather than at each other. We'll cover why messages change on the way to the field, how to measure uptake directly, where AI assistants genuinely help, and how to run a single weekly loop between the two teams.
Launch drift now has a name, and a timetable
PharmExec recently gave this problem a useful label: launch drift, defined as the gap between the go-to-market strategy and how it is actually executed in the field. The root causes it lists will sound familiar to anyone who has sat in a launch review: gaps between strategy and execution, variability in field execution, belief barriers among HCPs and friction with payers.
The more practical part of that analysis is timing. It argues that drift should be monitored from week one, that early trends become visible around weeks four to six, and that they are typically confirmed around month three. That's a narrow window. By the time a trend is confirmed, the first cycle has run, first impressions with key specialists have formed and the field has settled into habits.
Why does early matter so much? McKinsey's analysis of 210 launches between 2003 and 2009, published in 2014, found that about two thirds of new drugs missed the pre-launch consensus sales expectations in their first year. And first-year trajectories tend to stick. Of the products that underperformed in year one, 78% still underperformed in year two, and most first-year underperformers remained below expectations in later years. Outperformers were stickier too: 65% continued to outperform in year two, and 53% of those in year three. Whatever else that says, it says the first months carry a lot of weight.
Messages rarely get lost on the way to the field. They get edited, by people with good intentions and no feedback.
Messages don't disappear, they get edited
It's tempting to describe the problem as reps "not using the materials". That's rarely what's happening. What happens is that each step between the brand plan and the physician makes small, reasonable edits.
The kickoff compresses a year of strategy into a few sessions. Regional meetings compress it again. Country teams adapt the story to local guidelines, access conditions and competitors, which is necessary and right, but each adaptation is also a chance for emphasis to shift. Distributor and channel-partner teams often receive a shorter version still, sometimes secondhand.
Then individual reps do what experienced professionals do. They keep what worked last time, drop what felt awkward in the first few calls and fill gaps with the story they already know. After a few weeks the "message" in the field is a personal blend of the new story, the previous product and whatever a KOL said at the last meeting.
Three conditions make this worse in almost every launch we've worked on:
- the sheer volume of material in the first weeks, far more than anyone can absorb;
- people joining mid-launch, who never attended the kickoff and learn the story from colleagues;
- no feedback loop, so Marketing has no view of which messages landed and which didn't, country by country.
The third one is the root. Without feedback, every edit is invisible, and every team assumes the problem is somewhere else.
The usual signals arrive late or say little
Brand teams do have signals. They're just slow or indirect.
Sales data comes in with a lag, and by then it mixes message problems with access, supply, competitor activity and a dozen other factors. CRM call notes record that a call happened and which materials were shown, but the free text is thin and rarely says what was actually said. Market research with physicians is valuable but expensive, periodic and a step removed from the field force itself. Ride-along reports are rich but patchy, and they depend heavily on which manager wrote them.
None of these tells you, in week three, that most reps in two countries can't correctly state the patient population, or that a third of the field mixes up the primary and secondary endpoints. That's the kind of fact that is easy to fix if you know it and very costly if you don't.
Make message uptake measurable, message by message
The direct approach is simple to describe. Take each key message in the brand plan and turn it into questions that the field answers on their phones, in short rounds, over the first weeks of the launch.
A message such as "the product is for patients who remain uncontrolled on current therapy" becomes a question about which patient profile fits the indication, with wrong options that reflect how reps typically drift. A proof point becomes a question about the exact result. An objection handler becomes a question about the approved response, including one option that sounds right but makes a claim reps may not make.
Within a week of the first round, the brand team can see three things it has probably never seen before:
- which messages are known across the field, and which are not;
- which messages are actively confused, where many people pick the same wrong answer;
- where each country, team and channel stands on each message.
The second one is the most useful. A message nobody knows is a communication problem. A message that most people consistently get wrong in the same way is a signal that something in the materials, the training or the local adaptation is pushing people toward the wrong story. That's something you can fix in the next round of materials, not the next plan cycle.
In effect, this is a knowledge survey of your own field force: fast, repeatable and specific enough to act on. It tells you what people know and confuse, not what they say in the call.
A fair objection is that knowing a message is not the same as delivering it. A rep can answer every question correctly and still open the call with the wrong pillar. That's true, and we'd never claim knowledge data replaces observation. But the logic only runs one way. A rep who doesn't know the approved patient profile can't be expected to deliver it. Knowledge measurement tells you where delivery is unlikely to be right, which is exactly where managers should spend their ride-along time first.
Support the moment of the conversation
Measurement shows where the gap is. It doesn't help the rep standing outside a clinic who has just been asked something they can't answer.
This is where AI assistants genuinely earn their place, as long as they're built the right way. Ours is Atrivity Assistant. It answers field questions from approved content only. Before a visit, a rep can check the approved answer to a likely objection or refresh the safety messages. After a visit, they can ask the question the specialist raised that they couldn't answer on the spot, and get the approved answer rather than a guess from a colleague.
The bigger value, for Marketing, is what flows back. Every question the field asks is a signal about what they're unsure of. If dozens of reps in one country ask how the product compares with a particular competitor, that's worth checking as a possible message gap. It is not direct evidence of which messages reps use or avoid in front of physicians: a rep may ask because a doctor raised it, because they're preparing, or out of curiosity. Treat query patterns as a prompt to test with a question round and a few ride-alongs, then decide whether the next reinforcement round or the materials need to change.
The conditions that make AI safe in a regulated launch
We're cautious here, and we think buyers should be too. An assistant that improvises around the label is worse than no assistant. The conditions we'd insist on are straightforward:
- content generated only from approved documents, never from the open web;
- subject matter experts approve the source material and training questions before release; the assistant follows the agreed validation and escalation process;
- conflicts between source documents are flagged for a human, not silently resolved;
- data hosted in the EU, with each client's content isolated from every other client's.
Those conditions matter to MLR and to IT, and they should. They also make the output more useful, because a rep who trusts the answer will use it in front of a physician.
One loop for both teams
The real change isn't a tool. It's what happens when Marketing and Sales look at the same readiness data every week.
Without shared evidence, launch reviews tend to follow a familiar script. Marketing says the field isn't using the materials. Sales says the materials don't fit what doctors are asking. Both are partly right, and neither can prove it. With shared evidence, the conversation becomes specific: the field in these two countries can't state these three messages correctly, this objection is generating most of the questions to the assistant, and this misconception is concentrated in the teams that joined after the kickoff. That's a problem you can fix in a week.
What makes this work is cadence, not sophistication. A small review every week beats a big review every quarter, because drift compounds. A slightly wrong story told to a specialist in week two is a first impression that the correct story has to overcome in week eight.
It also gives each team something it has wanted for years. Marketing gets evidence of what the field actually knows and asks, not what it reports. Sales gets materials and reinforcement that respond to the questions the field raises. Medical gets early warning of knowledge that is drifting away from the label. L&D gets a curriculum it can update from data rather than opinion.
Start with one question at the next launch review
If you're preparing a launch, or already a few weeks in, try asking this at the next cross-functional review: for each of our key messages, what share of the field could state it correctly today, in each country, and how do we know?
If the room goes quiet, you've found your first action. Turn the key messages into questions, run a short round next week and put the results in front of both teams at the same meeting.
That's the loop we build at Atrivity, with one-minute challenges that measure message uptake, readiness data by team, country and topic, and Atrivity Assistant, which answers the field from approved content and feeds its questions back to the brand team. The plan and the field will never be perfectly aligned. But the gap between them doesn't have to be invisible.