First published in PM360 online.

Today’s pharma leaders are under pressure: Decision cycles are tightening, stakes are rising and the market is constantly changing. Decision-making is no longer siloed at each stage of drug development — from early clinical trials through commercialization and beyond.

Insights functions within pharma are going through a rebuild phase, redefining the impact they have in the decision-making cycle. Existing primary market research was not designed for this degree of complexity or pace on its own. Yet many organizations continue to depend on static, point-in-time insights, which lack connection to the broader context or other decision-critical data. Relying too heavily on these insights carries risks and can lead to ineffective approaches.

Artificial intelligence is changing this paradigm. Advances in data integration, simulation and AI-enabled research workflows are making it possible to build persistent market research data foundations grounded in real evidence — capturing not just what stakeholders say at a single moment but how attitudes, behaviors and decision drivers evolve over time. Rather than treating each study as a discrete event, a cumulative learning system can be built to reflect shifting market dynamics.¹ Rooted in AI, decision-grade digital twins have emerged as a promising development for insight generation. Built on longitudinal evidence-based data and designed to continuously update as new signals emerge, a digital twin moves beyond static reporting to enable scenario exploration and strategic testing. Used well, it augments traditional market research with a more dynamic layer of decision support to help teams understand not only where the market is today but how it may respond as conditions change.

What is a digital twin?

Traditional market research helps answer a familiar question: What do stakeholders think? A digital twin aims to add a second, two-fold question: What are they likely to do next and how might that change under different conditions?

In pharma, that distinction matters because stakeholder behavior is rarely static. Physicians, patients, payers, and other decision-makers respond to new data, clinical experience, competitive actions, affordability pressures and changing treatment contexts. A static snapshot can still be valuable, but it does not always capture how decisions may evolve.

A digital twin is therefore best understood not as a single synthetic output but as an always-learning decision framework. It can combine market research inputs with other relevant signals such as real-world data, competitive intelligence and behavioral indicators to create a more current and structured representation of the market. Over time, each study and each new signal can strengthen the system, allowing one wave of research to inform the next. As digital twins mature, they become behavior anchored and predictive, informing strategic choices in uncertain conditions.

A decision-grade digital twin provides hypotheses and support for key inputs into the asset or brand strategy. First, it can help teams understand near-current market conditions rather than relying on older, static snapshots. Second, it can support prediction by estimating how different stakeholder groups may respond as circumstances change. Third, it can support simulation by allowing teams to test scenarios before acting on them.

Real-world application: from insight to action

Practical examples illustrate how a digital twin can improve outcomes.

Global segmentation studies identify distinct physician personas based on behavior, such as innovation-driven prescribers or guideline-oriented clinicians. Traditionally, this approach provides support and creative inspiration for positioning and messaging of new products, but the reaction of those personas to such marketing approaches is unknown (unless more market research is undertaken). In addition, the characterization of the market can quickly become dated as competitor and regulatory activity shift behaviors. With a digital twin, these personas can be built into independent agents that represent each group of stakeholders. As the twin evolves and learns from market inputs, it is possible to explore how the personas will respond to different scenarios, such as the release of new clinical data, a pricing change or a revised positioning strategy, allowing teams to have more dynamic interactions with their key customer personas.

This approach drives more forward-looking, strategic and focused decision-making. Teams can also test messaging concepts across segments before fielding research, refine positioning based on predicted responses, optimize study design, and prioritize strategies before engaging with physicians.

Building trust in a digital twin

For a digital twin to reliably support asset decision-making, it must meet a high standard of rigor. Its data sets must be continuously tested against real-world data, updated with fresh inputs and supported by robust governance and quality controls.

In order to be considered “decision-grade,” a digital twin should consistently demonstrate five capabilities:

• Predictive validity: the ability to forecast future outcomes, not just replicate past data.
• Scenario realism: consistent behavior under changing market conditions.
• Continuous refresh: internal mechanisms with a consistent cadence to update models and detect drift as the market evolves.
• Bias mitigation: safeguards and audits to ensure appropriate representation and avoid amplifying historical distortions.
• Data privacy and provenance: clear documentation of inputs, assumptions and methodologies that resist individual identification.

Building a reliable digital twin requires more than technological capability. Beyond the depth of data required for the build, only human experts can provide the deep therapeutic knowledge, market research expertise and understanding of market complexity that are also essential to success.

Where insights leaders can start with a digital twin

For organizations beginning this journey, the front end of the market research design process is a good place to start. Teams can use a digital twin to optimize study design, refine survey questions, prioritize concepts and identify knowledge gaps before research begins. This approach will challenge insights teams to focus on asking the right questions, not asking all questions. These types of situations can deliver immediate efficiency gains with relatively low risk.

From there, companies can move into more advanced applications, such as:

• Enhancing interpretation of research findings.
• Extending insights through post-study interrogation.
• Monitoring market signals to detect ongoing shifts in behavior.

Over time, the value of having twin compounds. As the digital twin is continuously enriched with new data and insights, it becomes more accurate, predictive, and integral to decision-making. This stems from what’s been called the flywheel effect — each use strengthens the system and amplifies its impact. At the same time, attention must be paid to the inputs: feeding the twin with new secondary data, competitive intelligence and primary market research helps preserve its ability to reflect the external environment it is built to simulate. In contrast, relying too heavily on internal brand plans can cause the twin to mirror those perspectives rather than the outside world.

Importantly, this is not about replacing existing market research. It is about augmenting it — adding a layer of intelligence that improves speed, quality, and confidence in decision-making.

From insight to foresight: gaining an edge

The ability to anticipate change and have forward-looking evidence to act on it is becoming a defining competitive advantage in the commercialization of drugs. Digital twins offer a path forward, but this space is still maturing. Equipped to make faster, more confident decisions, insights teams that begin building and operationalizing these capabilities now will hold a leading edge into the future.

References

1. Pasculli G, Virgolin M, Myles P, Vidovszky A, Fisher C, Baisin E, Mourby M, Pappalardo F, D’Amico S, Torchia M, Chebykin A, Carbone V, Emili L, Roeshammar D. Synthetic data in healthcare and drug development: definitions, regulatory frameworks, issues. CPT Pharmacometrics Syst Pharmacol. 2025 May;14(5). Available from: https://ascpt.onlinelibrary.wiley.com/doi/10.1002/psp4.70021.