PANELYTICS
®
The Science of Superior Sampling
Experience the seamless precision of PANELYTICS, our proprietary method for superior sampling. With over 30+ years of expertise, including quality control heuristics, research on research (RoR), and analytics, we apply a test-learn-iterate system that includes:
01. Procuring and Curating Sample
-
Hand-pick sources/channels based on your specifications and the unique advantages of each.
-
Meticulously stratify samples to enhance representativeness.
02. Rules-Based Pre-Survey Quality Analysis
-
Implement a rules-based pre-survey quality check.
-
Use digital fingerprinting and metadata analysis to filter out fraudulent or unusable data.
03. AI / Machine Learning Survey Analytics
-
Utilize a continuous AI/Machine Learning classification system.
-
Identify potentially fraudulent or unusable data accurately.
PANELYTICS provides rigor and flexibility by utilizing Receiver Operating Characteristic curves, which indicate the trade-offs between sensitivity (true positives) and specificity (false positives). This supports optimal sample decision-making; you select the thresholds aligned with your risk tolerance.
METHODOLOGY
01. PROCURING AND CURATING SAMPLE
SAMPLE
BLENDING
We blend respondents from multiple panels and sources to enhance representation by eliminating or minimizing sample skewness and bias from membership recruitment methods.
Our sample comes from thoroughly vetted, profiled community assets that utilize transparent member recruitment and management practices.
RESPONSE
STRATIFICATION
We stratify the sample before sending invitations to our survey. Doing so improves representativeness as certain groups in the population respond more quickly to surveys, which leads to overrepresentation.
Stratifying or click balancing to Census informs any necessary balancing and reduces the need to apply large weights to the data.
CONTINUOUS IMPROVEMENT
Following data collection, critical quality insights feed into our continuous learning loop, incorporating ongoing RoR best practices and insights from our AI/machine learning QC models.
This test-learn-iterate system informs future sample blending and stratification, ensuring continuous improvement and enhanced data quality.
02. RULES BASED PRE-SURVEY CLASSIFICATION
Prior Participation in
Surveys Ecosystem
Presence of Suspicious
Technology
Known Fraud
Behavior
Pre-Screening Quality Heuristics
Implement a rules-based methodology for digital fingerprinting and pre-survey metadata analysis to detect and filter out fraudulent or unusable data.
03. AI / MACHINE LEARNING WITHIN-SURVEY ANALYTICS & MODELING
Training & Learning With External Language Data
Training & Learning With Internal Language Data
Training & Learning With Fingerprinting & Behavior
UNSTRUCTURED DATA CLASSIFICATION
DIGITAL METADATA & BEHAVIORAL CLASSIFICATION
Utilize a continuous AI/Machine Learning classification system to identify potentially fraudulent or unusable data accurately.