For data-heavy businesses making million-dollar decisions
What is your model not seeing?
A Top 20 insurer asked us to validate its pricing model. We found their riskiest accounts underpriced, their safest accounts overpriced, and the liability line inverted. The issue was already in their data. Their analytics stack did not surface it. We did.
Same data. Different visibility.
The finding that the existing stack missed.
98.1%
loss ratio on their highest-risk accounts. Underpriced. The data called for a 96% rate increase.
33.5%
loss ratio on their lowest-risk accounts. Overpriced, and quietly pushing good business away.
inverted
the liability line. The segment that looked safest was carrying the risk.
Their model looked functional.
The portfolio told a different story.
This one happened to be insurance. The method is industry-agnostic. The same blind-spot detection works anywhere the data runs deep enough to hide a pattern: banking, healthcare, real estate, logistics, risk of any kind.
Why we see what your stack doesn't
Most analytics describes what already happened. We make testable predictions, and then we test them.
16,000+ variables per entity, refined over 30 years of finding signal where others saw noise, for institutions like American Express, UBS, and Bank of America. The depth of signal is the difference. Generative AI creates. Predictive AI decides. This is the deciding kind.
30 years of findings
The same depth, across industries.
$8M / yr
AARP. Membership scoring savings, still running 5+ years later.
$1B / yr
Bank of America. Fraud & default prediction against the write-off base.
700k / day
Predictive targeting that replaced a cold-call center.
150 cases
Regional hospital. Undiagnosed congestive heart failure surfaced.
The 2 Week Pilot
Proof on your own data.
Send us a defined slice of your data; for an insurer, a cut of the commercial book. Two weeks later you sit down with Doug for the readout: where your risk concentrates, which signals your current stack can't see, and what both are costing you.
Paid as a pilot, credited in full if we go further. If we find nothing worth acting on, we'll tell you that too.
We reply within one business day to scope it. NDA first; your team spends hours on the data pull, not weeks.