# A risk review end to end: the Meridian finale — Hypothesis Testing Lab

## About these files

Meridian Assurance's telematics pilot: 60,000 motor policies over accident year 2025, their 4,663 settled claims, and 1,100 investigated claims scored by two fraud models. Generated by Charles Sambo's insurance notebook from seed 20260822.

Synthetic data: no real customer, policy or claim is in them, and the companies are fictional.

## `meridian-policies.csv`

One row per policy-year in accident year 2025: the frequency grain, with the exposure each policy earned. Rows: 60,000.

| Column | What it holds |
|---|---|
| `policy_id` | The policy, 1 to 60,000. |
| `accident_year` | 2025. |
| `region` | London, Midlands, North or Scotland. |
| `tier` | standard, or telematics (the pilot). |
| `exposure_years` | Earned vehicle-years, three decimals. |
| `earned_premium` | Earned premium in GBP, two decimals. |
| `claim_count` | Claims on the policy in the year. |

## `meridian-claims.csv`

One row per settled claim: the severity grain, which exists only for the policies that claimed. Rows: 4,663.

| Column | What it holds |
|---|---|
| `claim_id` | The claim, 1 to 4,663. |
| `policy_id` | The policy the claim belongs to. |
| `region` | The policy's region. |
| `tier` | The policy's tier. |
| `claim_amount` | The settled amount in GBP, two decimals. |

## `meridian-fraud-review.csv`

One row per claim referred to the investigations unit, where the true fraud status is known, scored by both fraud models. Rows: 1,100.

| Column | What it holds |
|---|---|
| `claim_id` | The claim, from the claims file. |
| `champion_correct` | 1 if the current (champion) model classified the claim correctly. |
| `challenger_correct` | 1 if the new (challenger) model classified it correctly. |

Every number in the chapter was computed from these files by the lab's engine, and checked against scipy 1.11.4 and statsmodels 0.14.0.

The chapter: https://hypothesis-testing-lab.azurewebsites.net/learn/risk-review-end-to-end