Case Study
Predictive Modeling & Virtual Prototyping
How a personal care manufacturer simulated consumer success before entering the lab
The Situation
A personal care manufacturer specializing in shaving products was struggling with the high costs and extended timelines associated with physical consumer studies for its men's 5-blade refillable razors. Seeking to maximize the value of historical consumer and sensory data collected across 14 products in 2018 and 2019, the company needed a way to simulate consumer success in-silico. The goal was to build a machine learning model with enhanced predictive capabilities to identify sensory drivers of liking and predict performance before moving to physical production.
How Does Predictive Modeling Reduce Development Timelines?
In traditional R&D cycles, the path from prototype to market is linear and iterative, often requiring multiple rounds of physical sampling and consumer validation. Predictive modeling transforms this into a parallel process. By identifying the specific sensory drivers of “Overall Liking,” researchers can adjust product variables within a digital environment to identify the optimal configuration for success.
Methodology: The “Playable Model”
Multi-Source Data Alignment
The process begins by aligning historical datasets:
- ●Sensory Descriptive Data: Standardized attributes and sensory profiles identified for the product category.
- ●Consumer Targets: Historical liking scores trained at the individual consumer level, allowing for tailored predictions and consumer segmentation.
Exploratory Landscape Analysis
Before a model is trained, a detailed Exploratory Data Analysis (EDA) is conducted to understand the full sensory landscape. This step determines the best modeling approach and evaluates the feasibility of incorporating additional data from different product categories (e.g., 3-blade disposable razors) or international markets to strengthen the model.
Interactive Simulation Dashboards
The final model is deployed within a “Playable” interactive dashboard that accepts standardized Excel format inputs and outputs branded PowerPoint slides. This interface allows developers to:
- ●Adjust Input Variables: Manually manipulate sensory profiles of new or virtual products to explore how changes affect consumer performance.
- ●Instant Prediction: Predict consumer liking for optimized prototypes based purely on their sensory profile.
- ●Compare to Benchmarks: Visualize how a virtual prototype sits within the competitive landscape compared to the benchmarks used to train the model.
Strategic Impact on Product Development
| Traditional Prototyping | Virtual Prototyping (Predictive) |
|---|---|
| Requires physical production of every iteration. | Conducts simulations and virtual prototyping digitally. |
| High cost per consumer test iteration. | High upfront value that compounds with every simulation. |
| Insights limited to the specific products tested. | Uncovers the underlying drivers of liking across the category. |
| Reactive: “Did they like this prototype?” | Proactive: “What sensory profile will maximize consumer liking?” |
Impact: Strategic Product Optimization
Quantifying the Value of “In Silico” Research
The integration of predictive modeling into the sensory workflow yields primary organizational benefits:
Technical Infrastructure
Aigora implements these systems using a modern, maintainable technology stack that ensures the model remains a living asset:
Backend Logic: The core machine learning model and statistical calculations are written in the R programming language.
Frontend Dashboard: The interactive user interface is built using a combination of R, Shiny, HTML, JavaScript, CSS, and Angular Framework component-based strategies.
Deployment: The application is coordinated through platforms like Shinyapps or DockerHub, scaled to support the core product development team securely.
Ready to Transition Your Product Development to Living Infrastructure?
See how Aigora's predictive modeling tools can simulate consumer responses and optimize product development before a single prototype is produced.