Insurtech · Telematics
Safe Driving Rate
~$8M
Additional margin per year
75K
Customer base
2
Data sources integrated
Q/Q
Quarterly roadmap cadence
Problem Definition
Posted speed limits leave margin on the table
The posted speed limit is not always the safe driving speed. We used speed as one telemetry input to Quanata's safe driving model, and knew we were leaving margin on the table due to a misaligned metric. When the posted limit says 65 but traffic consistently flows at 55, or when an access road speed differs from the adjacent freeway, the model was penalizing or rewarding drivers inaccurately — creating pricing error that directly impacted margin.
Research & Discovery
Finding contextual speed data at scale
In consultation with a data scientist and our data partnerships manager, I investigated alternative data sources to contextualize our telemetry signals from customers. We found a source that provides speed of traffic estimates at good resolution for most of the US. The investigation involved evaluating coverage, latency, cost, and accuracy tradeoffs across multiple vendors before identifying the provider that met our quality bar for production use.
Strategic Solution
Contextual when certain, posted when ambiguous
Collaborating with a data scientist, we modeled multiple ways of incorporating contextual speed data into our safe driving model, and settled on a system where we use contextual when certain, and posted when contextual reports ambiguity about which road a driver is on (i.e. access road vs freeway). This hybrid fallback approach maximized the accuracy improvement while maintaining model robustness in edge cases — critical for an insurance pricing system where errors compound.
Execution & Prioritization
Earning executive sponsorship through the roadmap process
Through the normal quarterly roadmapping process, I earned executive sponsorship for the product increment, focusing on leveraging existing data engineering pipelines, data broker relationships, and strong reach of a change to our core safe driving model. By framing the increment as a margin improvement with known infrastructure dependencies — rather than a new system — I minimized perceived risk and accelerated approval.
Impact & Metrics
$8M in additional annual margin
About $8,000,000 per year in additional margin across our small customer base of ~75,000. The improvement came from more accurate risk pricing — better alignment between our driving safety scores and actual driving conditions meant fewer pricing errors, less adverse selection, and stronger unit economics across the entire book of business.