Product knowledge and topic knowledge: why one score per HCP is not enough

Product knowledge is what a healthcare professional understands about a specific product's own evidence. Topic knowledge is what they understand about the science around it — the mechanism, the disease area, the comparative landscape. The two move independently, and a system that averages them into a single score per HCP discards the only detail that would have told you what to discuss next.

Most systems that record HCP understanding record one thing. A tier, a score, a segment, a label that sorts a list. It is convenient for reporting and close to useless for planning, because the question it answers — how well does this clinician know us — is not a question anyone acts on.

This piece is about the shape of that record at a single point in time. The separate argument for why it should persist across MSLs and build up over years belongs to why a knowledge position should persist and compound and is not repeated here. Both are part of the measurement problem this sits inside.

Why one score per HCP is the wrong shape

Averaging removes variance, and in this case the variance is the entire signal. A clinician who understands a mechanism thoroughly and has never engaged with the comparative data is in a completely different position from one who has skimmed both, and a single number puts them in the same place.

Three things go wrong, and they are worth separating because they have different fixes.

It is not reversible. From a composite score you cannot recover which parts are strong. Two clinicians at the same point on the scale may need conversations that go in opposite directions, and nothing in the number distinguishes them. Information has been destroyed, not summarised.

It moves too slowly to give feedback. A number aggregating many topics barely shifts after one good conversation. Anyone trying to use it to see whether their approach is working will see nothing for a quarter, conclude the measure is inert, and stop looking at it. They will be right.

It invites a ranking. One number per person is a league table waiting to be built, and it will get built, because that is what a single sortable score is for. HCP segmentation done this way slides quickly from planning into targeting. A league table of clinicians is the wrong artefact for a medical function to be producing, whichever direction it is sorted in.

The fix is not a better single score. It is more than one axis.

What product knowledge actually measures

Product knowledge is understanding of a specific product's own evidence. The trial that supports it, the endpoints it was measured on, the population it was studied in, the safety profile, where it sits in current guidance.

It is specific and it does not transfer. A clinician who knows one product in a class in detail may know very little about another in the same class, and that is a normal state rather than a gap to be alarmed about.

The important property is this: product knowledge is not something a clinician ever states, and not something that can usefully be asked about directly. Nobody announces how well they understand a product. What they do is demonstrate it — in how they discuss the trial design, in the questions they ask about the population, in whether they reach for the comparator data unprompted.

Which means product knowledge is better inferred from what was demonstrated than asserted as a rating. A judgement typed into a field after a meeting is one person's impression on one day. A position built from what the clinician actually engaged with across a set of related conversations is evidence, and it can be shown to whoever inherits the account.

There is a failure mode specific to this axis and it is worth designing for. An inferred position is only as complete as the conversations it was built from — strong on the ground that happened to be covered, silent on the rest. A record has to be able to say not established rather than defaulting to low, because those two mean entirely different things and only one of them is a finding.

What topic knowledge measures instead

Topic knowledge is understanding of a scientific topic independent of any product. Mechanism, disease biology, the state of the evidence base, the comparative landscape, how the treatment pathway is actually organised.

It has three properties that make it the more useful of the two axes on its own.

It is durable. It survives a product being withdrawn, a guideline changing, or a clinician moving institution. It is a fact about the person rather than about the relationship.

It is portable across a class. Topic understanding built in one conversation informs how the clinician will read every product in that space, including the ones they have not encountered yet.

It is observable. Clinicians talk about science readily and rarely talk about products in the abstract, so a conversation naturally produces evidence about topic knowledge and only indirectly about product knowledge. This is why the two axes are not equally easy to populate, and why a system that treats them as symmetric will always have a thinner product axis.

There is also a sequencing point that falls out of it. Topic knowledge is where a conversation has to start. A discussion of where a product sits with someone who has not engaged with the underlying science does not fail loudly — it produces polite agreement, high coverage and nothing coming back, which is the least informative meeting there is. What a depth measure should contain is built to make exactly that pattern visible.

One thing topic knowledge is not is agreement. A clinician who knows an area extremely well may be the most sceptical person on the list, and should be. Conflating knowledge with advocacy is a separate and serious failure, argued properly elsewhere.

Four combinations, and what each one means for the next conversation

This is where two axes earn their keep. Each combination implies a different conversation, and two of them imply going backwards.

High topic, high product. The informed clinician. Most of the remaining value runs in the other direction — what they can contribute rather than what can be explained to them. Meetings here should be built around questions, and they are where the genuinely useful insights come from. Arriving with foundational material is the fastest way to waste the slot and some credibility with it.

High topic, low product. The most common position and the most actionable. They know the science and have not engaged with your evidence specifically. They do not need the disease-area background; they need the trial design, the population, and how it sits against what they already know well. The conversation starts several steps in, and beginning at step one is the single most common misjudgement in this quadrant.

Low topic, high product. Uncommon, and worth looking into rather than celebrating. It usually means exposure through one narrow channel — a congress session, a colleague, a piece of promotional contact — that delivered product facts without the science underneath them. This position is fragile. It does not survive a sharp question from a well-informed peer. The right conversation goes backwards, to mechanism and evidence base, even though it feels like covering ground already won.

Low topic, low product. Early, and there is no shortcut through it. Attempting to discuss positioning here produces agreement that means nothing and no usable insight. The work is the science, in order, and the honest planning assumption is a sequence of conversations rather than one meeting with more slides in it.

The reason this is usable and a single score is not: quadrants two and three produce a similar mid-scale composite and need opposite approaches. Averaging them is not a simplification. It is the deletion of the one thing that would have told you which conversation to have.

Why this matters more for specialists than generalists

The gap between the two axes widens with specialisation, which means the cost of collapsing them is not evenly distributed.

A clinician seeing a broad range of conditions will tend to have roughly comparable depth across the topics and products within any one of them. There is not much variance for an average to destroy, and a single score does less damage.

A specialist is the opposite case. They may have read the primary literature, followed the field for years, possibly been an investigator, and they will have deep topic knowledge alongside highly uneven product knowledge across a class — thorough on two, passing on a third, nothing at all on the newest. A single number describes none of those states.

This is also the population where getting it wrong is most expensive. Pitching foundational science at someone who has published on it does lasting damage to credibility, and it is among the most common complaints clinicians make about field medical interactions. The information needed to avoid it exists in the record. A composite score removes it.

The practical consequence is straightforward. The more specialised the audience, the more the record has to carry two axes rather than one, and the less any single segmentation tier is worth.

RocketMSL tracks understanding separately by product and by scientific topic, built from what the clinician demonstrated rather than from a rating entered after the meeting. See how understanding is tracked per product and per topic →