Q&A: Why Your Pharmaceutical Training Might Not Be Working

Mark Lee, Industry Assistant Professor of People Analytics in NYU Tandon’s Technology, Management and Innovation Department, has spent more than a year building a case that could reshape how the FDA thinks about pharmaceutical training.

Before joining the Tandon faculty full time in 2025, Lee was Head of Research, Analytics and Business Development at UL Solutions ComplianceWire®, where he first noticed something troubling in the data behind how life sciences companies train their workforce, and decided to do something about it.

He shared the problem he unearthed, the solution he is pursuing, and how it relates to his current Tandon work.

Q: What's the core problem you're trying to solve?

I explain it this way: not just anybody can make ibuprofen for the commercial marketplace. The FDA requires that companies that sell the drug qualify their employees before they work on a manufacturing line or in a chemistry lab.

Unfortunately, the main method companies use to meet the FDA regulations is to the Standard Operating Procedure (SOP). Over 70% of pharmaceutical training consists of these SOPs. Companies are asking people to read a large number of documents out of context, and expecting them to go make ibuprofen. That, to me, is a very real risk. Not everybody retains information for very long from things they read. It is a very well established finding in the learning literature that people don’t retain information from a single reading of a document. Test them a few days later and they have forgotten.  

Part of the problem is that the underlying regulation is old. The FDA's training requirement, 21 CFR 211.25, was published in 1979, when personal computers weren't widely used and none of today's methods of training-content delivery existed. Since documents are required by regulation, they're easy to distribute and track to meet the letter of the law, but that kind of reading-and-understanding activity is widely considered a check-the-box exercise, not an effective one.

Q: How did you first identify this as a problem?

I have access to a database covering approximately 22% of the people on Earth who make pharmaceuticals and medical devices. When I analyzed the data, I found the majority of documents people are handed as training are open for less than two minutes. So it's not just a poor training method, it's not even being used. Fifty-four percent of the training given to pharmaceutical employees isn't even looked at.

Q: What did you do with that finding?

That, along with a few other data points, led me to say this is unacceptable. I worked with the Association for GxP Excellence, a nonprofit, to draft a proposed guidance and submitted it to the FDA's docket.

The core idea is that SOPs should stay as the controlled reference, but stop being used as the training itself. Training should shift to role-based instruction with real competency assessment — on-the-job training, simulations, e-learning — measured by demonstrated skill, not just a signature. If adopted, it would represent the FDA's interpretation of how training should be done. I've been on this for 14 months, and just got it submitted formally.

Q: What happens next?

Ideally, the next big step is that the FDA publishes the proposed guidance as a draft for public comment. Industry then weighs in — some will complain because it's more work, some suggest changes — and the FDA adopts some feedback and disregards some before publishing a final guidance.

We gave this proposal to the FDA's Office of Pharmaceutical Quality first and addressed their feedback before submitting, so I'm hopeful that shortens the road to public comment and adoption.  

The other factor in favor of the guidance is that when this transition has been implemented in real companies, the cost savings are enormous. Very conservatively, by making training right the first time, we can save billions of dollars by improving employee onboarding time, improving product quality, and reducing risk to patients.

Q: How does this connect to what you teach in the classroom at Tandon?

I'm a people analytics guy. I work on the measurement of work, its valuation, and the human factors involved. This is an example of analyzing big data to uncover a problem tied to a regulation written in 1978, when the word processor was high technology, that's persisted simply because that's how it's always been done. I'm a human factors engineer: I study what people do, what technology is capable of, and I look for improvements.

I use the pharmaceutical industry as a case study in my classes all the time, and these examples have been invaluable. The lesson for students: passively reading material someone else generated for you leads to low retention, you have to work through information to really learn and retain it. This is such an important lesson for students who are using AI in their studies. Answers to homework, assignments, and tests that are easily retrieved from an LLM will soon be forgotten. If you want to truly learn something, you need to work at it!