Case Study: Clearcover accelerates model velocity and deploys with confidence using Arize
Key results
The challenge
Clearcover's ML team initially planned to monitor production models with a business intelligence tool, but setup took more than two weeks per model and dashboards were not real-time, causing roughly 24-hour delays in detecting performance degradation. Data scientists spent an estimated 9-10 business weeks per year manually monitoring models, and BI tools were not built for ML needs such as feature-drift detection.
The solution
Clearcover selected Arize in early 2021 after a competitive proof of concept and implemented its ML observability platform for automated threshold-based monitors, concept and feature drift detection, data-integrity checks, and performance tracing. Monitoring is wired in through Arize's Python SDK so it activates automatically on model deployment, with alerts routed immediately to Slack and email.
“We recently deployed a model that went from inception to production in 46 days - hardly a small endeavor given the model is relied on to score over 50,000 insurance applications daily. Arize is a big part of that success because we can spend our time building and deploying models instead of worrying.”
APAlex PostLead Machine Learning Engineer, Clearcover
The results, in context
Clearcover reported deploying a model from inception to production in 46 days, a model relied on to score over 50,000 insurance applications daily. Automating monitoring cut per-model setup from 2-4 weeks to instantaneous, freeing more than 400 hours per year across the ML team and enabling roughly 10% more models deployed into production annually. The company reported a payback period under nine months and over 150% ROI in the first year.