Commercial pharma teams have more physician data than ever before. What most don't have is a clean way to keep teams and systems current. Segments get rebuilt once a quarter, sometimes just twice a year. Physicians change every week; a first prescription, a reimbursement question, a shifted priority. By the time a campaign reaches them, it's often talking to someone who's already moved on.
That gap between how fast physicians move and how often commercial teams update their view of them was the focus of a recent webinar hosted by ODAIA & Fierce Pharma, featuring Bradley Birenbaum of GSK and moderated by Frank Fascinato of Treviso Consulting.
An Advocate, Treated Like a Stranger
Frank opened by sharing this story. Just before launch, a physician received an email explaining that a newly approved therapy was finally reimbursed. It read like an introduction she didn't need. Having authored the clinical trial behind it, she was intimately familiar with it. She had been an outspoken advocate for the product from the start. The message was crafted for someone new to the product who didn't have her history with it. It had treated her like just another name on a target list.
In moderator Frank Fascinato's words:
"There was a disconnect between the data and reality. The physician had already moved on. The campaign hadn't."
Frank Fascinato, Treviso Consulting
Their goal was recognizing that physicians rarely move at the same pace, even when a typical segmentation model insists they should. Two clinicians who looked identical a month ago may need entirely different things today. One still wants confidence in the evidence, the other is stuck on patient access, with no single message that bridges both. Treat them as though they occupy the same stage, and even a well-produced campaign starts to feel generic.
Sometimes what a physician needs is a piece of content. Sometimes it's a tool, or simply a person showing up at the right moment. Mistaking one for another is its own version of the timing problem: the right idea, delivered in a form that doesn't fit the moment. It's a small moment, but it's the whole problem in miniature — a system with enough information to know better, and no mechanism to act on what it knew in time.
The Ceiling Nobody Could Reach
Segmentation still gives teams a practical way to organize the market. It exists because evaluating physicians one at a time used to be nearly impossible to do at scale. Sorting HCPs into groups turned an unruly universe into something a team could actually act on. Segmentation's weakness is time: physicians continue moving while the categories built around them remain fixed. That's the ceiling Bradley Birenbaum called out when he described what his team is actually chasing:
"The aspirational problem we're trying to solve is how do we get to N equals one HCP targeting. There's no way that I can have that team get to N equals one targeting without a lot of help."
Bradley Birenbaum, Director, HCP Marketing, GSK
Segmentation doesn't just misjudge who a physician is today, it can also miss the exact moment they moved. A physician who already wrote a first script for a product doesn't need more evidence to build confidence. They need help with what comes next: staying covered, staying reimbursed, keeping the patient on therapy. Serve the wrong one, and the physician is left to solve the next problem alone.
Jackie Markle, VP of Pharma Technical Strategy at ODAIA, has seen exactly where that gap lands:
"If they've already written for the first time and we're still serving them content to build confidence, but they're not getting the message about access and affordability, we're doing them an injustice. Then they're out scrambling to help their patient."
Jackie Markle, VP, Pharma Technical Strategy, ODAIA
This is the challenge underneath every brand team running HCP engagement at scale. Digital channels made that lag worse. A banner click or an email open doesn't say much on its own, and the signals that actually correlate with a physician advancing are easy to misread as noise.
Coordinating channels turned out to be the easier problem. Coordinating an honest read of what's actually stopping a physician from moving forward is the harder one, and it's the one that decides whether the next message lands as relevant or generic.
The clearest analogy for this might be a picnic. Planning one means checking the weather twice, once when you decide to go and again the morning you leave, because conditions change in between. Commercial planning works the same way. A campaign built months in advance reflects a moment in time, and physicians rarely wait around for it to catch up. The question is what does it actually take to close that gap inside a real brand team's day-to-day workflow.
From Rep Intelligence to Marketing Intelligence
The best sales reps have always been able to walk into a healthcare professional’s office and say the right thing, because they knew their history with the brand. The question the panel kept circling: is that same precision possible at the marketing level, across every channel, for every physician, without adding headcount?
Birenbaum's answer, in practice, was to start narrow. He piloted the model in the rep channel first, where the data is simplest and the feedback loop is fastest, and proved the model was moving the numbers that mattered before expanding. From there, email is usually the easiest next step – cheap to test, simple to measure. Noisier channels, like banner ads, come later, once there's enough data to trust the signal.
The strategy stayed the same. What changed was who did the work of getting there. Instead of a manager asking analytics to build a target list and waiting weeks for the handoff, the model does that work continuously, inside guardrails Birenbaum set. His team spends less time building lists and more time watching what's actually moving the needle: which messages are earning attention, and where a segment needs content that doesn't exist yet.
That shift, from managing campaigns to managing a system that keeps adjusting, is what ODAIA calls Marketing Intelligence: giving brand teams the same one-to-one precision reps have always had, without asking anyone to work weekends to get there.
No Black Boxes
None of this works if a brand team can't trust it. Birenbaum's conditions for trusting any AI system were about accountability more than capability. The system had to optimize for the strategy he'd already set. It had to run on data he could evaluate himself. And it could never operate as a black box he couldn't question.
"I don't trust an AI model that can't explain it. If it's right, great — I don't know why it's right. If it's wrong, what am I going to do when my boss asks me why it's wrong? Well, I trusted the model."
Bradley Birenbaum, GSK
In a regulated industry, a recommendation nobody can explain becomes a liability the moment someone asks why.
What surprised him was how little infrastructure the shift actually required. It started with a spreadsheet mapping which messages mattered at which stage of the journey, something his team now calls, half-jokingly, their Rosetta Stone:
"Oddly, I thought that was going to be a bigger lift, tagging all the data and getting everything ready. That took three hours. It actually wasn't that hard."
Bradley Birenbaum, GSK
A Different Question to Ask the Data
The lift itself turned out to be the easy part. Speed was the harder problem. Birenbaum described engagement data that could take 45 days to reach his team, for decisions that needed to be made in one. By the time a report landed, the physician who generated the signal might already be asking a different question, facing a different barrier, or prescribing differently. The data was still accurate. It just wasn't useful anymore.
That changed how he reads results now:
"Should I double down and let that continue, or should I go back and readjust my strategy to say, actually, we've been hunting in the wrong place?"
Bradley Birenbaum, GSK
He called the faster read "free market research." Campaign performance still shows how well a plan was executed. His question reaches underneath that result to examine whether the original understanding of the physician was accurate in the first place.
Back to the Advocate
Which brings the conversation back to where it started. The launch email contained correct information, but it reflected an outdated picture of its recipient. By the time it arrived, the physician had moved beyond the audience the campaign originally imagined. Her history with the therapy was already deep. The message had simply failed to keep up with her.
Birenbaum put a fitting cap on it later in the session: good data means a model shouldn't have to relearn the obvious. Feed it what you already know, and it stops making the mistake the launch email made.
Three things worth carrying out of this conversation:
- A segment is a shortcut, not a strategy. It let teams act at scale back when evaluating every physician individually wasn't possible. That constraint is gone. Segmentation hasn't caught up.
- Prove it where the feedback loop is fastest, then extend it. Birenbaum didn't launch across every channel at once. He started in the rep channel, saw a 10% lift in TRx per call, then carried what worked into email, then digital.
- Every recommendation has to answer "why this HCP, why now." The reasoning has to hold up for the marketer acting on it and the teams responsible for analytics, data governance and privacy.
Most brand teams are past the launch-day email mistake. The harder version of the same gap is still common: engagement data that arrives weeks after it mattered, a campaign still running on last quarter's read of a physician.
The Marketing Intelligence case study walks through what closing that gap looks like in practice.
Watch the full conversation for how GSK got started, and the mistakes they'd tell you to avoid.
