# unifi.ai

> Competitive intelligence from public property-casualty insurance rate filings: what
> competitors asked for, what regulators approved, and what their loss ratios did next.

## About

unifi.ai turns the public rate-filing record into competitive intelligence for
property-casualty pricing actuaries. It joins three things that are published
separately: the rate change a carrier filed, the change the regulator actually
approved, and the carrier's realized direct loss ratio in the years that followed.

That last join is the differentiator. Filing data is sold by several vendors; the link
from a filing to the writer's subsequent realized loss ratio is not, because it
requires joining the filing record to a separate free source (the NAIC Market Share
Reports) on the NAIC group code.

Operated by 5G Vector Inc., Atlanta, Georgia.

## What it does

- **Rate actions vs. results**: approved rate changes per year beside the loss ratio
  that followed — one carrier alone, against a peer group the user names, and against
  every writer NAIC publishes.
- **Requested vs. approved**: how much of the ask the regulator granted. California
  only, because California is the only state that publishes both figures.
- **Rating variables across carriers**: one disclosed variable compared across
  writers, including deductible relativities rebased onto a common base so curves
  published against different base deductibles become comparable.
- **Filing reader**: structures a filing the customer supplies from any state.

## Coverage today

- **States**: California (live). Texas is the next source; it is file-and-use.
- **Lines**: personal auto, commercial auto.
- **Realized loss ratios**: NAIC Market Share Reports, 2019 onward.

## Pricing

- Free — one state, current year, no export
- Analyst — $200/month, one state, full history, exports
- Team — $800/month, five seats, all covered states, API access
- Carrier — custom, priced on scope

## Data sources and how they are obtained

All data comes from public sources accessed through their intended workflows:

- **Texas open data** (`data.texas.gov`) — Socrata SODA endpoints published under a
  statutory open-data mandate. No clickwrap. Published crawl delays are honoured.
- **California Department of Insurance** — public notice and Approval-and-Closed
  spreadsheets from `insurance.ca.gov`. California uniquely publishes both the
  requested and the approved rate change per filing.
- **NAIC Market Share Reports** — free published reports, for realized direct loss
  ratios.
- **Customer-uploaded filings** — any state; the customer retrieved the document
  through the intended workflow and unifi.ai structures it.

unifi.ai does **not** scrape SERFF Filing Access. Reaching it requires accepting an
NAIC user agreement that prohibits automating the download of data, so those 49 states
and DC are out of scope; licensed vendors are the lawful route to that breadth.

## Stated limits

These are limits of the public record, and unifi.ai states them rather than filling
the gaps:

- **Direct loss ratios only.** The free source carries no expense data, so a combined
  ratio cannot be derived and is never implied.
- **NAIC publishes only the top ten writers per state and line.** Most filings
  therefore do not join a realized loss ratio, and "all writers" means all *published*
  writers, not the whole market.
- **Disclosure is partial.** Selected trends appear in roughly 20% of filings and
  development factors in roughly 14%.
- **A blank is never a zero.** A filing that disclosed no percentage is reported as
  undisclosed, not as a 0% change. About 40% of Texas auto filings disclose no
  percentage at all.
- **File-and-use states publish no approved change.** Texas dispositions read
  Reviewed, Withdrawn or Rejected — never Approved — so no unifi.ai figure describes a
  Texas change as approved.
- **Every figure carries its source and an as-of date.**

## What it is not

unifi.ai is not a comparative rater. It does not quote premium for a given risk, build
market baskets from quote data, replicate credit-score models, or run disruption
analysis on a customer's own book.

## Methodology

The extraction pipeline, the loss-ratio join and the peer-group mathematics were built
and validated for a Casualty Actuarial Society 2026 Ratemaking call paper. Measured
accuracy against California's published approved change: 79.1% exact and 2.2
percentage points mean absolute error per distinct filing. Product figures are kept
reconcilable with that published work.

## Target customer

Pricing and product actuaries, state-filing and compliance teams, and actuarial
consultancies in US property-casualty insurance.

## Links

- Homepage: https://unifi.ai
- Pricing: https://unifi.ai/pricing
- Blog: https://unifi.ai/blog
- Sign in: https://unifi.ai/login
- Create an account: https://unifi.ai/signup

## Contact

- Email: hello@unifi.ai
- Website: https://unifi.ai