From Question to
Model in
24 Hours
Upload your outcomes. Pinpoint builds and validates a custom model, production-ready in a few clicks.
Overview
Pinpoint's Discovery Models let carriers build a custom predictive model for any defined two-outcome event — churn or retention, fraud or clean, claim or no claim. Built for teams who want to test a predictive idea against their own data without standing up a new project.
Move From Idea to Production, Quickly
Test new predictive ideas, validate what works, and deploy models into real workflows.
Define the outcome
Any event you can label as did or did not. Churn or retention. Adopted or declined.
Upload labeled records
A correctly pulled dataset from your own history. Minimal internal effort.
Pinpoint trains and validates
Feature prep, training, testing, and validation, automated end to end.
Results in 24 hoursDeploy
Production-ready through configurable APIs, into the workflow you already use.
A few clicks-
Build Models Around the Decisions You Care About
How It Works
A binary classifier predicts one of two outcomes — cancelled or retained, converted or lost, fraud or clean. It trains on your own labeled history, so it learns your book's patterns rather than an industry average.
What You Get
- The ability to model any yes/no outcome
- Fast validation of new ideas using real data
- Clear signals you can use across your workflows
Build Models Around the Decisions You Care About
How It Works
A binary classifier predicts one of two outcomes — cancelled or retained, converted or lost, fraud or clean. It trains on your own labeled history, so it learns your book's patterns rather than an industry average.
What You Get
- The ability to model any yes/no outcome
- Fast validation of new ideas using real data
- Clear signals you can use across your workflows
Pinpoint's explainable visualizations reveal what drives human behavior, and also the interesting reasons why conventional data have missed the signal on this important issue. We see this as an important step forward in predicting behavior and ultimately reducing frictional costs and premiums.
Key Discovery Benefits
Test Ideas Quickly
Validate a new predictive idea in 24 hours, before committing to it.
Build What Your Business Needs
Model any two-outcome event: churn, premium leakage, program adoption. No fixed catalog.
Deploy Without Friction
Commercial clients receive annual allocations to build and run models, so experimenting doesn't need a new commercial conversation.
Case Studies
Results in Production
Personal auto · 714,577 policies
Redirecting acquisition spend with pre-quote Loss Predictions
Personal auto · 714,577 policies
Severity predictions applied across a 715k-policy book
Personal auto · 130,000 claimants
Attorney representation flagged at first notice of loss
Homeowners · ~300,000 policies
Frequency predictions at the top of the funnel
Homeowners · ~150,000 policies
Non-weather water risk priced before the application
Personal auto · 80,000 policies
Holding loss ratio while the book contracted 7%
Homeowners · 200,000 policies
Book roll reshaped by pre-bind Loss Predictions
Frequently Asked
Questions About Discovery
What are Pinpoint’s Discovery Models?
How do Binary Classifiers work?
What business challenges can Discovery Models help solve?
Do carriers need their own data science team to use these models?
How can carriers access Discovery Models?
Fast-Track
Profitable Growth
Data Science That Delivers Business Impact
Purpose-built AI for insurers, uniting actuarial expertise and data science to drive smarter, fairer, and more profitable decisions.