Progressive Insurance

Progressive Insurance, the largest motor vehicle insurer in the United States with over $55 billion in annual revenue, faced two compounding challenges. Its marketing campaigns relied on static creative assets that could not adapt to individual audience segments at scale, capping conversion performance. Simultaneously, Progressive had accumulated over 20 years and 14 billion miles of telematics driving data — yet its legacy analytics infrastructure could not translate that data into timely predictive models. Business units waited months for insights, and the data science team could not keep pace with growing demand across churn, fraud, billing, and risk use cases.

Progressive pursued two parallel AI initiatives. For marketing, the company partnered with Claritas and MMA Global to deploy Claritas' AI Creative Optimization platform, using Generative AI to produce 120 distinct synthetic audio ad variations. The system analyzed 6 million impressions to determine which of the 96 activated variants resonated with specific audience segments, then continuously reoptimized delivery in real time. For analytics, Progressive adopted H2O.ai's open-source machine learning platform alongside H2O Driverless AI, an automated ML solution handling feature engineering, model validation, hyperparameter tuning, and interpretability. This gave the data science team the ability to build and deploy predictive models — covering fraud detection, customer churn, billing anomalies, and threat analysis — across multiple business units without expanding headcount.

The GenAI advertising campaign produced measurable, multi-dimensional improvements:

• 197% lift in campaign performance comparing AI-on versus AI-off conditions
• 52% lift in total conversions driven by GenAI-optimized ads
• 300% more users reached by GenAI ads compared to non-AI ad variants
• Campaign timeline reduced from a typical three months to under eight weeks

On the analytics side, H2O.ai enabled Progressive's data team to produce models faster and address more business problems across more units with the same team size — a meaningful throughput multiplier without additional hiring.

• Creative personalization at scale requires more than content generation — a real-time data activation layer must select and serve the right variant per audience segment or the volume of variations provides no advantage.
• Legacy analytics infrastructure sets a hard ceiling on insight velocity; automated ML platforms like H2O Driverless AI can multiply a data science team's output without proportional headcount growth.
• Decades of behavioral and telematics data only become assets when paired with ML pipelines capable of operationalizing predictions across multiple business units simultaneously.
• Pilot scope matters: Progressive tested across 6 million impressions before activating 96 of 120 variants, using empirical signal rather than assumptions to guide rollout.