Summary
Statflo partnered with Arima to improve customer targeting for tablet cross-sell campaigns. By enriching limited first-party wireless usage data with Arima's privacy-safe Synthetic Society™, Statflo significantly improved its machine learning model accuracy, without using or sharing any PII, resulting in a 16% increase in sales and stronger campaign ROI.
The Challenge
Statflo's objective was to cross-sell tablets to existing customers while preserving its primary goal of plan renewals. However, the company faced several constraints:
- Limited first-party data: Wireless usage data alone lacked the depth needed to accurately identify high-propensity buyers.
- Model performance limitations: Existing ML models struggled with precision, recall, and explainability due to insufficient behavioral and demographic signals.
- High stakes targeting: Poor targeting risked wasting outreach on low-propensity customers and negatively impacting customer experience.
- Strict privacy requirements: Any solution needed to be fully privacy compliant, without sharing or exposing PII.
Statflo required a credible, privacy-safe data source to strengthen its models and improve campaign effectiveness.
Solution
Statflo integrated the Synthetic Society™ dataset with its first-party wireless usage data to enhance customer propensity modeling.
Key elements of the approach included:
- Synthetic data enrichment: From over 4,000 available attributes, Arima identified 20 high-impact variables most correlated with tablet purchases, including income, household composition, housing type, and buyer psychology.
- Improved model explainability: The enriched dataset allowed Statflo to better understand which factors influenced purchase behavior, increasing confidence in targeting decisions.
- Advanced ML modeling: Statflo applied XGBoost models to leverage the enriched data, significantly improving both precision and recall.
- Privacy-first architecture: No customer PII was shared or required, ensuring 100% compliance with privacy and data governance standards.
This approach enabled Statflo to build more accurate, explainable models and confidently tailor messaging for high-probability buyers.
Result
- 16% increase in tablet sales for Statflo's client
- 2–6x improvement in F1 score, driven by gains in both precision and recall
- Higher campaign ROI, with fewer messages sent to low-propensity customers
- Improved model explainability, uncovering insights such as higher tablet affinity among homeowners and families with children
- 100% privacy compliance, with no PII shared or exposed
As Doug Creighton, Data Science & Business Systems Lead at Statflo, noted:
"Arima's Synthetic Society database provides us with credible and impactful data signals and continually helps us to increase model accuracy so that we can give our clients better predictive guidance driving strong business results."
