Launching a small branded drug into a category dominated by cheap generics is one of the harder commercial problems in pharma. The total prescriber universe is enormous; the rep team you can afford to deploy is small; and the physicians who write the most prescriptions in the category aren't necessarily the ones most likely to write your drug. The targeting question matters in a way it doesn't for established blockbusters.
For a specialty pharma client launching two cardiovascular drugs into adjacent categories, the question was concrete: out of nearly a quarter-million prescribers, who were the 20,000 to 25,000 the field team should actually call on — and where geographically did those targets concentrate enough to justify a territory?
From the Universe Down to a Target List
The starting point was every prescriber who had written at least one relevant prescription in either category over the prior six months — about 247,000 physicians. That's far too many to cover, and most of them weren't going to be commercially relevant regardless. The first pass narrowed by category-level prescribing volume, keeping anyone in the top five deciles for either drug class. That cut the candidate set to roughly 36,500.
From there, the second pass scored each candidate on a weighted physician valuation: prescribing volume in each category, weighted toward branded volume rather than generic; the physician's history of prescribing the legacy branded product in the category; and a multiplier reflecting expected branded share given their patient mix. Apply guardrails, sort, and the top of the list became the actionable target universe of about 25,000 physicians.
The Insight That Mattered: Expected vs. Actual Branded Share
The weighted volume score is the obvious half of the algorithm. The non-obvious half — and the more useful one — was the branded-share multiplier. Within any prescribing category, payer plan mix shapes how much of a given physician's volume should be branded. A physician whose patients are mostly on plans where the brand has a strong formulary position will, all else equal, write more branded prescriptions than a physician whose patients are mostly on plans where the brand is restricted. That's "expected" branded share — calculated from each physician's payer mix and the brand's actual position across plans.
What the data showed was that expected branded share correlated reasonably well with actual branded share — but with a wide tail of physicians prescribing well below their expected level. Those physicians weren't writing less branded volume because their patients couldn't get the drug. They were writing less because of habit, rep coverage gaps, or plain inertia. The gap between expected and actual was the real opportunity.
A physician's prescribing pattern reflects two things: what their patients can get covered, and what the physician chooses to prescribe within that constraint. Expected branded share captures the first. The gap between expected and actual captures the second — and it's where field activity actually moves the number.
starting universe
two filtering passes
before any rep planning
From a Target List to a Field Plan
A target list isn't a coverage plan. The question of which physicians to call on has to translate into where to deploy reps, how big each territory is, and which markets get covered first. We aggregated physician-level scores to candidate territories — built from zip code groupings within each market — and ranked metros by total target potential. The top ten markets accounted for a disproportionate share of the actionable opportunity. Those became wave one. Markets just outside the top ten became candidates for wave two as the launch progressed.
For physicians in markets outside the covered territories, the same scoring identified the highest-potential prescribers for non-personal promotion — direct mail, digital outreach, conference activity. The targeting framework didn't just decide who got a rep visit; it decided what form of contact each high-value physician received.
How the Weights Evolve
One subtlety worth flagging: the algorithm's weights aren't static. Pre-launch, the targeting score weighted both products evenly because both audiences mattered for territory design even though only one product was being detailed initially. Post-launch, as the second product approached its own launch, the weights shifted to give that product more pull on the targeting math. The same scoring framework, recalibrated periodically, kept the field plan aligned with where the business was actually going.
Why This Generalizes
The methodology applies well past pharma. Any business with a large universe of potential customers, finite sales coverage, and observable proxies for both volume and conversion potential faces the same problem: how do you turn a mailing list into a target list? The pieces are consistent. Build a candidate score from observable behavior. Layer on a multiplier reflecting structural opportunity — whether that's payer mix in pharma, firmographic potential in B2B, or category penetration in consumer. Aggregate up to coverage units. Recalibrate as the market evolves. The hard part isn't the math. It's having the discipline to keep the score honest as commercial reality shifts.
Building a targeting model for a sales force, a launch, or an account-coverage decision? Say hello.