Your pricing model
has a blind spot.
Swarmalytics finds the patterns that your current data stack misses: at the property level, the portfolio level, and across every risk tier in your book.
The same patterns that cost carriers millions in mispriced accounts every renewal cycle.
What we found in a Top 20 P&C insurer's book.
Case: Commercial Property Portfolio Analysis
A Top 20 P&C insurer came to us with what they believed was a routine data question. What the models returned was more uncomfortable than what they asked for. Their pricing was wrong in both directions — simultaneously under-pricing their highest-risk accounts and over-pricing their best customers. Their current models and analytic partners weren't surfacing either problem.
The data to see it had existed for years.
The counterintuitive finding: the liability line was running the opposite direction — low-risk accounts at higher loss ratios than high-risk ones. The book was mispriced on both lines at the same time.
The diagnostic showed exactly where, and by how much.
98.1%
Loss ratio on highest-risk ExCAT accounts
96%
Rate increase needed to reach a sustainable 50% LR
2.9x
Gap between highest and lowest risk tier loss ratios
Three ways Swarmalytics
works inside your portfolio.
Pricing Segmentation
We run 16,000+ variables per account against your existing book to surface exactly where your pricing model is working against your portfolio — in both directions. Under-priced risk and over-priced good customers show up in the same analysis.
Catastrophe Risk
We predict structure-level damage before the first raindrop using satellite data, topography, stream proximity, and elevation. Standard models aggregate by geography. Ours resolves to the individual property which is the difference between a house that floods and the one 400 yards away that doesn't.
Portfolio Intelligence
Fraud signals, claims prediction, compliance scoring. The same swarm intelligence architecture applied to the behavioral and financial patterns inside your customer base. At Bank of America, we applied this to a $1B annual credit card write-off problem. Any improvement on that number pays for itself many times over.
Why the resolution is different.
Most risk analytics tools are working with the same data stack. The carriers using them get the same picture. Swarmalytics runs on a different depth of signal: 16,000+ variables per property, 140 million US homes in the database, and a swarm intelligence architecture that finds patterns across all of them simultaneously.
True swarm intelligence runs many models in parallel, each sharing information, each converging on the same problem from a different angle. Gartner puts this architecture 6-plus years ahead of mainstream adoption. We've been running it in production for years.
The result is resolution that standard approaches don't reach. Patterns that exist in your book right now, invisible to the tools you're currently using.
30 years. The world's largest institutions.
American Express
VP of Analytics at Epsilon (acquired by AmEx). Grew the analytics group from 2 people to 50. Worked with 50–100 Fortune 500 clients. Commissioned one of the first commercialized predictive systems — the direct predecessor to what Swarmalytics runs today.
Bank of America
AI targeting applied to a $1 billion annual credit card write-off problem. Fraud pattern identification: people steal in round numbers. Legitimate charges look like $16,462.15. Fraud looks like $5,000 flat. The trailing zeros were predictive signal.
AARP
Scored 60–70 million senior records for membership likelihood. Built communication optimization tracking every contact to find optimal cadence. Saved $8M per year. That model is still running.
Aetna
Combined Aetna's data with clinical records to build a medication compliance prediction model. Scored every patient for likelihood to comply. Diabetic non-compliance leads to hospital stays costing tens of thousands. Metformin costs almost nothing. Aetna trusted our models more than their in-house team.
What's in your book?
A portfolio analysis starts with your existing data. We show you what's there that your current models aren't surfacing. No obligation before the findings.