One of the huge pharmaceutical companies employed Bayser Consulting for development of product launch strategy. Simulation modeling was applied for reconstruction of interactions between the company, doctors and patients.
Objective
Identify the optimal promotional strategy for an upcoming product launch and the corresponding sales. In particular, what is the ideal breakout between Direct-to-consumer advertising (DTC) and Personal Selling? Address specific questions such as:
- Does the client need a Contract Sales Organization (CSO) and if so, how large and for how long?
- How should the client reshuffle the current field sales plan to best free up primary position details?
Solution
The model included:
- Probabilistic generation of medical prescriptions as the result of the meeting of the physician (exposed to reps and Key opinion leaders (KOL)) and the patient (exposed to DTC and co-pay assistance), and the transformation of the prescription into drug consumption and subsequent refills.
- Animated visual display (using AnyLogic) of the interactions between patients, physicians, and representatives, the outcome of those interactions, and the resulting sales. This is an excellent way to grasp the essence of the problem and generate penetrating questions.
- An "Analysis of Misses" feature could speed up the search for optimal allocation by suggesting how the current promotional strategy needed to be altered to maximize sales.

Figure 1. Screenshot of the Agent Based Model
Why Agent-based Modeling?
- The modelers needed to simulate the patient, the physician, and the sales representative as distinct entities since they were interested in their interactions and how promotional exposure altered their behaviors over time.
- The modelers wanted to study group dynamics and, as such, needed to differentiate between members of the group. To that end, personality traits needed to be assigned to individuals following, say, a Gaussian distribution.
- The modelers needed to simulate various promotional strategies that concurrently affected sales at different levels. In particular, they needed to model the lifecycle of the drug starting from the patient falling sick to the consumption of the prescribed drug.
- The modelers needed to peer into the workings of the group practice: how it embraced the drug, what bottlenecks impeded adoption, and what corrective measures could be taken to maximize drug uptake.
It is the need to take into account these factors that determined the consultants’ choice of AnyLogic, which is powerful at Agent Based Modeling.
Figure 2. Types of interactions between Agents
Outcome
The results of the project were:
- Identification of the optimal promotional strategy and assessment of various candidate strategies.
- Recommendations regarding the CSO and how to reshuffle the product portfolio accordingly.
- Uptake curve of the drug and corresponding sales forecast.
- Understanding of the various decision points, and how they interact with each other towards overall sales.
- Insightful questions that the analysts would not have thought of in the absence of this model.
Note on Emergent Behavior
The emergent behavior of a group is the behavior a group exhibits because individuals make different choices than what they would, if they were not a part of the group. What's more, the choices of an individual impact choices other members of the group make.

Figure 3. Emergent behavior of ants
The foraging path the ant colony ends up taking with depends on the initial decisions individual ants make. Likewise, the behavior of the group practice depends on and evolves with the decisions of individual physicians.
Conclusion
The Agent-Based Modeling module of AnyLogic is the platform of choice as:
- it allowed the modelers to answer all the questions they posed,
- it provided a close-up view of the dynamics of the group practice, and
- it was ideal for studying emergent behavior.
The model can be scaled up so the allocation question can be broadened to include payer rebates for improving access, in addition to DTC and Personal Selling.