Novo and Lilly pay hundreds of professionals to talk about their GLP-1s. I followed each one, year by year, in the file that logs them.
In short
Ozempic has influencers, and most of them see patients. They are doctors and other healthcare professionals who get paid by Novo Nordisk or Eli Lilly to give talks and consulting about their drugs. In the United States the law requires every one of those payments to be disclosed in a registry anyone can download, and that makes it possible to answer with data a question that in any other industry would stay pure speculation: how long does that relationship last?
In the previous case I counted the money: USD 180.2 million disclosed for the GLP-1 drugs for diabetes and obesity between 2021 and 2025, with Lilly concentrating and Novo spreading. And I left a commitment in writing: I was going to measure how long those speakers last, and if more than 30% of the program changed from one year to the next, I was going to admit that Lilly's concentrated model was fragile. This case keeps that commitment: I followed Lilly's 657 speakers and Novo's 1,139, one by one, across the five years.
Nobody reached that 30%. The interesting part is why: retention is bought. Under USD 5,000 a year, a Lilly speaker stays into the next year 16.9% of the time; above 25,000, almost everybody stays at both houses (96%, grouping the high bands). Lilly retains more in the aggregate because it parks its money in the band where nobody leaves.
I close with what this gives a commercial team: how much you have to pay for a speaker to stay, why almost nobody leaves for the competition, and what I will watch when CMS publishes next year's data.
Retention is bought. Under USD 5,000 a year a speaker does not last long at either house; above 25,000 almost nobody leaves either one. Lilly does not retain better: it bought the permanence.
Every number in this case comes from counting professionals in CMS Open Payments (the public registry of industry payments to healthcare professionals run by the federal government), with the same rules as the previous case plus five new decisions, written down before implementing them in the decision log. The two that matter most: you belong to a year's speaker program if you received at least one speaker or consulting payment for a GLP-1 that year, and the spend bands compare the two companies at equal investment per head. The eight attacks I ran against these findings are published in the repository, failures included.
The window opens in 2021 with a strange scene: Lilly shows up with 178 GLP-1 speakers as if the program had been born that year. The pre-history says otherwise: 93.3% were already collecting speaker fees from Lilly between 2017 and 2020, for other house drugs (at Novo, 74.1%). The 2021 GLP-1 program was a redeployment of speakers already on the payroll.
Genuine recruiting came later, and fast: of Lilly's 2021 entrants, 93.3% had prior history with the house; of its 2025 entrants, only 17.8%. Eight out of ten new Lilly speakers in 2025 were genuinely new faces. The 2017-2020 years enter this case only as this control: no figure from that era holds up a chart or a headline.
Turnover is measured on each year's members, which are fewer than the running total: between 178 and 357 speakers per year at Lilly, between 510 and 690 at Novo. On those cohorts, Lilly's annual turnover moved between 10.0% and 25.3%; Novo's, between 9.7% and 28.4%. Neither company passed 30% in any year, however you count who belongs to the program. The one nuance: counting only speakers at USD 5,000 or more, Novo would have reached 44.3% in the year of its cut.
The three-year picture says the same: of Lilly's 2021 cohort, 62.9% was still active in 2024, and of its 2022 cohort, 78.6% in 2025 (Novo: 51.7% and 62.8%). Of Lilly's 657, 105 collected in all five years; of Novo's 1,139, 318.
Up to here it looks like Lilly manages its relationships better. Splitting by spend takes that reading apart.
At the same annual spend-per-head band, Novo retains the same or better than Lilly under USD 25,000: 47.4% against 16.9% in the lowest band, 85.3% against 69.2% in the 5-to-25k band. Above 25,000 the question dissolves: almost everybody stays, at both.
Lilly's aggregate edge (82.2% retention against 79.2%) comes from composition: six out of ten of its professional-years live in the bands of 25k or more, against fewer than three out of ten at Novo, which keeps most of its people where retention is lower. The specialty mix explains another part: within each professional profile Lilly retains somewhat more, but standardizing the mix cuts the gap in half. And the result does not depend on where I put the cutoffs: I tried moving them to half and to double, and comparing each company against its own distribution, and the ordering did not change. At both houses retention climbs with the money up to 25k; from there on, paying more changes almost nothing (the detail of the tests is in the repository).
Someone leaving the GLP-1 program almost never crosses the street: rival signings never passed 15 people in a year, and counting in the most generous way possible they reach 33 of Lilly's 172 exits in the whole window. The dominant exit is out of speaker work altogether, and at Lilly reassignment to other house drugs also weighs.
Novo's exodus has two chapters. The second has a visible cause: the 26.9% turnover belongs to the year its program was dismantled, when its speaker spending fell 69%. The first (28.4%) happened while the budget was still rising, and stands as the piece the budget explanation does not cover. When Novo rebuilt the program (a 140% increase in 2025), turnover fell to 9.7%: its fragility was a budget decision, and it left with the budget.
The whole picture fits in one sentence: an Ozempic influencer lasts as long as their budget line. Under 5,000 dollars a year the relationship cuts itself; above 25,000 it almost always continues; and when it ends, almost nobody crosses to the rival: the typical exit is to stop getting paid to talk, because the house stopped paying.
Data up to here. What follows is my opinion.
For whoever builds or audits a speaker program, this case leaves a price list for permanence. The step that buys retention sits between 5 and 25 thousand dollars a year: below that, the relationship is disposable for both companies; above 25k, paying more buys almost no extra permanence, because there is no turnover left to remove. And the classic fear of the concentrated program (the rival taking your expensive voices) does not show up in five years of the file: the exit to the rival is the rarest of the three.
The reading has a limit I prefer to declare: keeping a speaker and gaining something from keeping them are two different claims, and the second is not in this database. The next cut is free and already has a date: when CMS publishes Program Year 2026, Lilly's expensive cohort (the people collecting 25k or more a year today) will show whether the bought permanence holds another year or the circle comes apart.
The repository with the code, the decisions, the eight attacks and the previous case is published in full. If you build or audit speaker programs and you see something badly framed, write to me: that is the kind of error I would rather fix with someone who knows the ground.
This is independent work, made with public sources. I have no commercial relationship, and never had one, with Novo Nordisk, Eli Lilly or any of the companies mentioned.
If you read this far, something about the problem caught your interest. I like discussing these decisions with people who live them: write me what you would do differently, or let's book a call.
See all projects