Working with a firm that uses state-of-the-art analytics isn’t just for research geeks.
The world of market research analytics is becoming more interesting. The burgeoning field of Behavior Economics is casting doubt on the validity of some traditional market research practices. Fortunately, selective use of new statistical techniques can address some of these concerns as well provide deeper insights into our data.
Advanced analytics are powerful because they reduce complex relationships to their essence, allowing us to uncover findings that would be impossible otherwise. But at Echo Cove we don’t inundate you with “coefficients” and the like (unless you want them), rather we interpret the statistics and use them to support our findings and recommendations. Furthermore, we do not believe in “black boxes.” We only use advanced analytics that that have been tested in academia and the field. Here are some of our commonly-used techniques and how we apply them:
Types of Advanced Techniques
- Shapley Value Analysis
- Data Fusion
- Anchored Max Diff
- Customized Discrete Choice
- Chi Square Residual Analysis
- Latent Class Modeling
- Maximum Difference Scaling
- Dual Choice Discrete Choice
- Profit or Share Optimization
- Tree Models
How and Where We Apply Advanced Techniques
Key Driver Analysis
This is analysis used in many types of studies to identify the “drivers” of outcomes such as brand image, loyalty, purchase and advocacy. Most research firms base their Key Driver Analysis on standard forms of multiple regression, often producing unreliable results because of problems like multicollinearity.
We use Shapely Value Analysis (originated in game theory) because it overcomes these problems by running iterative models with all combinations of drivers, producing more stable, accurate results.
We use Latent Class Modeling rather than K-Means and other traditional approaches because, among other advantages, it handles missing data and multiple types of data in one model and can incorporate weights.
In segmentation, we often want to create a short list of predictive questions to quickly assign individuals to segments (for future research or typing tools or sales activities) and/or assign all customers to segments based on transactional data. In both cases, we use Data Fusion (based on Nascent Linkage Maximization). It yields more accurate classification with fewer questions than standard approaches because it iteratively shifts borderline respondents between segments to achieve the most accurate model.
Product Development & Pricing
We typically use Customized Dual Choice Discrete Choice. In this approach, respondents answer two questions for each task: preference and likelihood to purchase. This has several advantages over Sawtooth and other “off the shelf” programs, such as producing more accurate results with smaller samples, the ability to better simulate real-life situations and stronger prediction of market behavior. We also:
- Model product feature/price scenarios to maximize margin, share and revenue, enabling informed strategic decision-making
- Segment on product feature/price preferences to ensure we are focusing on the most valuable customers and prospects
- Provide user-friendly simulation tools which allow our clients to quickly run scenarios as new ideas occur
New Qualitative “Add-ons”
Sometimes quantitative studies need more “color” or deeper exploration of key topics. In these cases we add qualitative techniques such as online interviews, in which select respondents are routed to an online interview in real time. This can be highly insightful because respondents with interesting attitudes and/or behaviors can be “cherry picked” for the online interviews.