How MindSet Marketing Solutions Works
Predictive Modeling
Speedy, accurate analytics made for today.
The most important asset companies possess is their knowledge of what has happened in the past. Companies can now leverage their historical, customer and/or third party data and factor in current medical encounter information to better target their DTC marketing campaigns and develop predictive models. Predictive modeling is a proven technology that's been used successfully for decades by the largest financial and retail organizations to discover and target their most likely prospective customers. However, the cost and lengthy turnaround time of traditional modeling techniques has made it prohibitive for most companies- until now.
Our modeling software, modelMEDICXTM, buids models "on the fly" from the source data inputs that have undergone extensive quality analysis. They are built with patented genetic algorithms that enhance productivity by automating the tedious, time-consuming data analysis process and, by reducing the total model development cycle time from weeks or months to days. This allows you to discover breakthrough opportunities and leverage hundreds or even thousands of data attributes in analytics models, reducing the manual process of selecting variables and guessing at your outcomes.
modelMEDICXTM
modelMEDICXTM is a state-of the-art advanced analytic software system designed to build more accurate predictive models for customer acquisition, cross-sell and retention programs. The first step is to load the data into modelMEDICXTM. The software can clean and sample the data, or an analyst can do it manually. Next, a dependent variable is selected and model building begins. Over several hours (usually overnight), the software creates and tests thousands of potential models, using the most “fit” ones to breed or create new generations of models. The output of the modeling process produces a “score” for each of more than 37 million zip+4's in the U.S., which allows us to find the very best prospective targets for a variety of health-related areas including medical conditions, prescription drugs, O-T-C medications, insurance coverage, co-pay tiers, medication compliance and persistency and more.
Using this approach, we can leverage hundreds or even thousands of data attributes in analytical models, reducing the manual process of selecting variables and increases the likelihood that subtle predictors will be found, resulting in models that can outperform regression-based approaches by 15% or more. This gives you more time to test new ideas and discover breakthrough opportunities because:
- We deliver 10%-15% better predictive accuracy versus traditional regression-based techniques on critical success metrics such as response rates or retention yields
- We reduce model development time by 50-75% through automated data attributes
- We explore thousands of data attributes to provide better insight
- We eliminate manual data preparation
- We update models on a daily or weekly basis to reflect the most recent customer or third party data
- Predictions are based on genetic algorithm technology that can find subtle patterns of customer behavior by allowing you to work with 100% of the available information rather than just a subset of the available information
What Is a Genetic Algorithm?
Inspired by the theories of Charles Darwin, genetic algorithms are a new way to accelerate the building of predictive models using a large number of data attributes. This approach begins with the system creating a random set of models. The models then “compete” like animals in a natural environment with the strongest or, in this case, the most predictive, surviving. Models that survive each generation then “breed” and “mutate” just like natural organisms. Each succeeding generation competes on fitness and the best (most fit) models move on. After many generations (but taking only a few hours to process), the model with the greatest predictive power emerges.
Why Use It
modelMEDICXTM runs on 100% Java and uses familiar tree and tab navigation This process continues for thousands of generations and millions of candidate models until a “king” emerges with the best predictive power. Our analysts fine-tune the model and make final adjustments using an understanding of the business problem. If necessary, the software can run additional generations with different parameters. Finally, this model can be applied to the business problem through automated scoring procedures.
New more accurate predictions are built from:
- Patient level disease information to predict neighborhood disease prevalence
- Specific patient and consumer behaviors to estimate potential
- Seasonal events to predict patient and consumer purchasing cycles
- Psychographic information to predict patient and consumer choice
Providing Better Applications For:
- Direct marketing
- Disease, Rx, and OTC affinity scoring
- Medication compliance management
- Health plan fair share using DTC pull thru
- Cross-sell/up-sell promotion
