For pricing, underwriting & risk leaders
Is your pricing model inverting?
Four signals of a blind spot in your own book, and how to find it.
A Swarmalytics field guide
You have seen what one insurer's model was hiding. The question that matters now is whether the same thing is happening in yours.
False confidence is the expensive failure: a model that looks functional while the portfolio quietly tells a different story. It rarely announces itself, but it leaves signals.
Here are the four we check. Run your own book against them.
SIGNAL 01
Your loss ratios don't spread the way your tiers say they should
Pull loss ratio by your own risk tier. If the spread between your "safe" tier and your "risky" tier is narrower than your pricing assumes, your segmentation is blurrier than your rates.
In the insurer's property book, the highest-risk tier ran a 98.1% loss ratio while the lowest-risk tier ran 33.5%. That spread is the easy case. The harder cases hide in lines where the tiers have drifted closer together than anyone has checked recently.
CHECK Loss ratio by tier, by line. Look for compression you can't explain.
SIGNAL 02
Your lines disagree about who is risky
When the same account ranks safe on one line and risky on another, one of the models is missing a signal that the other is catching.
On the liability line, the ranking had fully inverted: low-risk accounts ran a 75.4% loss ratio against 58.8% for the high-risk accounts. The property model and the liability model were describing two different portfolios.
CHECK Cross-tab risk rank across lines for the same accounts. Disagreement is a flag, not noise.
SIGNAL 03
Your best accounts keep shopping at renewal
If your lowest-loss accounts also have your highest non-renewal or quote-out rate, the market is correcting your model for you. You are likely overpricing the accounts you most want to keep, and a more precise competitor is quietly taking them.
This is adverse selection running in the direction nobody watches for: not the bad risks you underprice, but the good risks you overprice and lose.
CHECK Retention against loss ratio, by tier. Watch the top.
SIGNAL 04
Closing the gap would take a rate change you would never file at once
If bringing your worst tier to a sustainable loss ratio would require a rate increase too large to file in a single cycle, the mispricing has been compounding for years, and the model never raised its hand.
For the insurer, getting the highest-risk tier to a sustainable 50% loss ratio implied a 96% rate increase. That number is not only current mispricing. It is how long it had been invisible.
CHECK What rate change would bring your worst tier to target? If the answer is uncomfortable, that is the finding.
Why a model misses this
None of these signals require replacing your actuarial infrastructure. They usually require more resolution than standard pricing inputs carry.
We append external data to your existing policy records: property characteristics, location, land use patterns, OSHA inspection history, industry classification. On the liability line above, OSHA history alone separated claim rates two to one (14% with any inspection history versus 7% without). The models train on your own loss history. The added variables are what make the hidden pattern visible.
The patterns were already in the data. Finding them was the work.
Find it in your book.
We can run this exact analysis on your portfolio. We append our data to your records, build predictive models against your loss history, and show you where the blind spots are, on your own book, in plain numbers. No pitch deck. A real finding. It starts with a 20-minute call.
The same method finds hidden patterns in any data-heavy book: banking, healthcare, real estate, logistics. This one happened to be insurance.
Swarmalytics maintains 16,000+ variables on every US commercial and residential property in a 75-computer private data center, refined over 30 years of finding signal where others saw noise.