Insights · Rating errors · 5 min read
What a $4M rating refund teaches rating teams
A 2026 California consent order shows what a rating error costs once it reaches policyholders: refunds, interest, a penalty, and customers who had already left. Five lessons for teams that implement rate changes.
In December 2025, a carrier group met the California Department of Insurance to self-disclose that its own processes had created a rating error, and that policyholders had been charged incorrect and unapproved rates. The consent order that followed in 2026 is public. We are not naming the carrier, because the lesson is not about them. It is about how rating errors behave.
What the order records
| Policies that needed a premium return | 399 |
| Premium refunds | $4,374,694 |
| Interest on those refunds, at 10% | $1,766,905 |
| Policyholders still insured when the order was signed | 85 of 399 |
| Penalty | $50,000 |
The carrier had to make its best efforts to find the 314 former policyholders so the refund checks could reach them. It also agreed to additional steps in its rate filing practices.
Lesson 1: The interest is not a rounding error
The interest on the refunds came to about 40% of the refunds themselves. A rating error does not cost what it overcharged. It costs that plus time, and the time is how long the error goes undetected.
Lesson 2: Most of the customers were already gone
Fewer than a quarter of the affected policyholders were still customers by the time of the order. A rating error that overcharges does not just create a liability. It quietly pushes customers out, long before anyone knows why.
Lesson 3: Self-disclosure is the good outcome
This carrier found the problem itself and went to the regulator. That is the best case, and it still cost refunds, interest and a penalty. The alternative is a regulator or a market conduct exam finding it first.
Lesson 4: The fix after the fact is a process, not a check
When a rating error surfaces, the natural remedy is more review: more people re-reading filings and spreadsheets. Review helps, but people re-reading tables is exactly what missed the error the first time. The durable fix is a check that does not depend on someone choosing the right rows to look at.
Lesson 5: The check belongs before go-live
Every one of these costs grows with time in force. A check that runs when the rate change is implemented, and before it takes effect, turns a refund into a defect fixed in QA. The questions that check has to answer are simple to state:
- Does the system rate what we filed, for the right effective date?
- Did only the intended rows move, with our deviations still on top?
- When something is wrong, which row, and which step?
Questions
How common are rating errors that lead to refunds? Regulators publish consent orders and market conduct findings regularly, and rating and filing compliance is a recurring theme. Most errors never reach an order, because they are found as undercharges nobody measures.
Who should own the pre-release rating check? The VP of Pricing owns what is filed. The Head of QA runs the check. Both need the same evidence: every changed row, checked against the filing, before the change goes live.