Every rate revision, checked against what you filed, before it goes live.
Circulars, deviations and new editions arrive as tables. VeriQuant reads them as data, generates the cases that read the changed cells, reports which cells each case read, and grades your system’s premium against them to the cent.
Today a revision is checked with spreadsheet macros and a few dozen hand-picked quotes. The defects that cost money are the ones nobody picked: a row changed outside the revision, a row dropped, last year’s edition still selectable.
Rate books live as CSV or Excel files beside the rating rules, countrywide with state overrides. Your columns map onto our table format through a declared mapping, and every row is accounted for.
A before and after pair becomes a change set, carrying the effective date of the edition it produces.
Bureau circulars load as change sets with no re-keying. A direct importer for PolicyCenter rate-book exports: Roadmap
| Set-up | Revision | Back-test | |
|---|---|---|---|
| Reference | Your current rate books, accepted as the baseline | The filed source as received: the circular, the carrier deviation, the effective date | An older circular, applied to the edition in force before it |
| What it proves | The simulator reproduces your system on the cells the matched cases read. It does not prove the tables match the filing. | Your system matches the filing on the rows the revision changes. Untouched rows are reported as baseline, not checked against the filing. | Rows that disagree surface as candidates. A person classifies each one. |
The verdict reads “reproduces your system on N of M cases”, with every case that does not match listed.
For every state, line and date window, VeriQuant names the edition that wins and whether it is the one you filed. Competing, unfiled and orphaned editions are findings. Active is not the same as in force.
Each table is graded cell by cell against the change sets: as declared, missing, wrong value, or unexpected. A premium mismatch names its first diverging rating step and the cells it read.
An agent on the model you already use, over MCP, drafts change sets from the circular and your deviation filing, and registers every assumption it made. Drafts stay marked as AI work until a person approves them. The reference is always your filing. Your policy system is always the side under test.
And the business you have already written. Re-rate the policies you have already issued against the edition that should have applied, to find leakage in written business, with the row and the step for every policy that differs.
See the three flows live in a briefing. The demo runs on synthetic tables, a fictional carrier and fictional circulars: every number is invented, and no licensed ISO content is used.
Would your tests notice a broken rate? The public fleetquote report ↗ plants defects in a rating rulebook and shows which ones the tests catch. It is a rating example, not a rate-revision demo.
On PolicyCenter? Guidewire rating testing →
Read the revision as data: the circular or filing becomes a change set with the effective date of the edition it produces. VeriQuant generates a case for every row the revision changes, computes each expected premium from the filed source, and grades your system’s premium against it to the cent.
It proves your system matches the filing on the rows the revision changes. Untouched rows are reported as baseline, not checked against the filing.
Yes. ISO ERC files import directly as change sets. Other rate books map from CSV or Excel through a declared column mapping. A direct importer for PolicyCenter rate-book exports is on the roadmap.
ISO and ERC are marks of Insurance Services Office / Verisk. VeriQuant is not affiliated.
The next step
A rate revision. A bureau update. A new state. An edge case nobody trusts. In 45 minutes, a senior member of our team walks your VP of Pricing, Head of QA and CIO through how the simulator is built for that case, and how your team runs the comparison in your own QA environment, exactly as you test today.
45
minutes
or email info@karatelabs.io
Need a rating check we don’t list? We build it during your evaluation.