Budget season has a particular rhythm inside a pharmaceutical company. Commercial arrives with revenue against forecast. Market access arrives with formulary wins and reimbursement decisions. R&D arrives with trial milestones, enrolment rates, regulatory submissions. Each of them opens with a number, and the conversation that follows is about what the number means.
Then Medical Affairs presents. And the presentation contains meetings held, advisory boards convened, congress interactions, materials shared, HCPs reached. Every figure is accurate. Every figure was hard-won. And none of them answers the question the room is actually asking, which is whether any of it worked.
This is not a failure of effort or of rigour. Medical Affairs teams are among the most scientifically capable groups in the industry, and they have spent two decades building measurement frameworks. The gap persists anyway, and it persists for a structural reason worth stating plainly: the unit of value in Medical Affairs is a conversation, and conversations have never been countable in any way that survives scrutiny.
Activity metrics survived because they were the only option
It is easy to be dismissive about meeting counts. It is more useful to ask why they became the standard in the first place.
The answer is that they were measurable. When an MSL had a two-hour scientific exchange with a haematologist about bleeding risk in a specific patient population, the only things that could be reliably captured afterwards were that the meeting happened, roughly how long it took, and whatever the MSL chose to write in a free-text field before the next appointment. The substance of the exchange — which evidence was discussed, which objections were raised, what the clinician actually took away — existed only in the memory of two people, one of whom does not work for the company.
Faced with that, counting the meetings was not laziness. It was the honest maximum. An organisation cannot report on what it cannot observe, and for most of the history of the function, the scientific content of field interactions was genuinely unobservable at scale.
The problem is that a measure adopted out of necessity eventually gets treated as a measure chosen on merit. Activity counts stopped being a proxy and started being the target. And once a proxy becomes a target, it starts shaping behaviour rather than describing it.
What a volume metric quietly rewards
Consider two interactions in the same week.
In the first, an MSL meets a broadly supportive clinician who already prescribes within the therapeutic area, agrees with the data, and has no outstanding questions. The meeting is pleasant, runs to time, and produces a positive impression on both sides. It is logged.
In the second, the MSL meets a sceptical academic who challenges the trial design, questions the applicability of the primary endpoint to her patient population, raises a comparator the company would rather not discuss, and leaves unconvinced. The conversation runs long, is uncomfortable, and produces three specific evidence requests the company did not know existed. It is also logged.
On any activity metric, these two interactions are identical. On any measure of what actually happened, they are not remotely comparable. The second one moved something. It surfaced an evidence gap, tested the strength of the scientific position, and generated a follow-up that could change what gets published next year. The first one confirmed a state that already existed.
A metric that cannot distinguish between them does not merely fail to capture value. Over time, it shifts effort toward whichever conversation is easier to have, because that is the one that fits more comfortably into a target. Nobody decides this. It is what happens when the only available measure counts occurrences. How an activity target reshapes a diary is worth following on its own: why 340 meetings was never a measure.
Borrowing from Commercial makes the problem worse
The most common response to this gap is to import a measurement framework from elsewhere in the business, usually Commercial. Prescribing data is plentiful, well-structured, and already flowing. Attaching Medical activity to prescribing movement produces a number quickly.
It also produces three problems, and they compound.
The first is causal. Prescribing behaviour is influenced by formulary status, guideline updates, competitor activity, sales effort, health-system economics and clinical experience. Attributing movement in that variable to a scientific conversation requires assumptions that will not survive a serious challenge, and everyone in the room knows it.
The second is that it measures the wrong thing even when the attribution holds. A scientific exchange that correctly leads a clinician to conclude that a therapy is not appropriate for her patient population is a successful medical interaction. Measured commercially, it is a failure. Any framework that scores that outcome as a loss is pointed away from the function's actual purpose.
The third is the one that matters most in the long run. Medical Affairs derives its standing — internally, and with the clinicians it engages — from being non-promotional. A function measured on commercial outcomes has, in a structural sense, become a promotional function, whatever its intent. The measurement framework is not a neutral reporting layer. It defines what the function is for.
So the choice as usually presented is between a metric that says nothing and a metric that says the wrong thing. Neither is acceptable, and the persistence of that choice is why the problem has stayed unsolved for so long.
What would have to be true
Set aside what is currently possible and ask what an honest measure of scientific engagement would actually require.
It would have to measure the conversation, not the person conducting it. An MSL who draws hard questions from a demanding clinician is doing the job well, not badly, and any measure that punishes difficulty will be quietly gamed within a quarter.
It would have to be derived from the organisation's own approved scientific content, rather than from an externally imposed target. What counts as a good discussion of bleeding risk depends on what the evidence base actually supports — which is a question the medical team has already answered when it approved the material. What such a measure should score, and what it should never count, is a question in its own right: what scientific depth should mean.
It would have to capture what came back, not only what went out. Coverage of planned scientific points is half the picture. The other half is what the clinician contributed: unmet needs, evidence gaps, treatment barriers, questions nobody anticipated.
It would have to work when nothing is recorded. A meaningful proportion of scientific interactions will always be undocumented by audio, whether through preference, setting, or local rules. A measure that only functions with a recording is a measure of consent rates.
And it would have to compound. A single interaction is a data point. The same clinician across six interactions, with knowledge tracked by product and by scientific topic, is an asset — and it is the only asset in Medical Affairs that grows more valuable with time rather than less.
None of those requirements is exotic. What made them unreachable was the observability problem: for most of the function's history, the conversation itself was simply not available as data.
The constraint has lifted
That is the part that has changed, and it has changed recently enough that most measurement frameworks in use were designed before it did.
Scientific conversations can now be captured, with consent, and analysed for substance rather than merely transcribed. Talking points can be generated from approved content rather than invented as targets. What the clinician raised can be structured rather than lost in a free-text box. Knowledge can be tracked as a position that moves, rather than assumed from attendance.
Which means the honest framing of where Medical Affairs stands is no longer that impact is unmeasurable. It is that impact has been unmeasured, for a good reason that no longer applies.
That distinction matters at budget season. A function that cannot measure its value is asking for trust. A function that can is making an argument. The second position is considerably stronger, and it is now available. What that argument looks like across a quarter is a review built on what the science did.
The question worth sitting with is not whether your team is delivering scientific value. It almost certainly is. The question is what your function would be able to argue for — headcount, evidence generation, a seat in a decision it is currently briefed on afterwards — if it could evidence what it already delivers.
RocketMSL was built to close that gap. See how the platform measures scientific engagement →
