Skip to content
Arcwright

Cambridge Mutual

Claims rebuilt around the handler, not the mainframe

Cambridge Mutual is a 140-year-old mutual insurer with 1.4 million household policies and a claims process built around a green-screen policy admin system. We left the mainframe in place and rebuilt the work on top of it, cutting the average claim from eleven days to four.

Client
Cambridge Mutual
Sector
Insurance
Run
12 months · Jan–Dec 2024
Filed
2025-01-28
Capability
Platform modernisation · Systems integration · Security and compliance
Stack
C# · .NET 9 · React 19 · SQL Server 2022 · Azure · Azure Service Bus · Playwright · Terraform

Impact

  • 4 days

    Average household claim cycle, down from 11 days

  • 63%

    Fewer manual re-keying steps per claim

  • 1.4M

    Policy records migrated with no reconciliation breaks

Challenge

Insurance
12 months · Jan–Dec 2024

A handler settling a burst-pipe claim moved between a 3270 terminal, two shared spreadsheets and an email inbox, re-keying the same policy number up to nine times. Thirty-eight per cent of claims were touched twice because information arrived after a decision had already been made. Consumer Duty reporting required evidence of consistent outcomes across the book, and the only way to produce it was to export the mainframe extract and count by hand. Replacing the policy admin system outright had been costed twice and shelved twice; the board would not approve a third attempt.

Solution

So we did not replace it. A facade service wrapped the mainframe screens behind a typed API, and new claims capability was built beside it in .NET rather than inside it — the strangler-fig pattern, applied one claim type at a time starting with escape of water because it was the highest volume and the least contentious. Handlers got a single React workspace that assembles policy, cover, history and correspondence on one screen and writes back through the facade, so the mainframe stayed the book of record throughout. Every decision now carries a structured reason code, which is what turned Consumer Duty reporting from a manual count into a query. We migrated 1.4 million policy records across four weekends with balance and cover reconciliation on each, and closed with 1,900 automated tests that the client’s own team runs.

Plates

  • Claims handler workspace showing a single household claim with policy details, cover limits, prior claims history and correspondence assembled in three columns, and a decision panel with structured reason codes.
    Plate 01Nine re-keyed fields became none.
  • Diagram of the strangler-fig facade: the mainframe policy admin system at the base, a typed API facade above it, and four migrated claim types drawn as completed segments alongside three still routed to the legacy screens.
    Plate 02Migrated claim type by claim type; the mainframe stayed authoritative.
  • Consumer Duty outcomes report showing settlement times and acceptance rates broken down by claim type and customer vulnerability flag, with a month-on-month comparison chart.
    Plate 03Reason codes made the regulatory report a query.

Testimony

In week two they told us that about a third of the programme we had scoped was not worth building. That conversation saved roughly £400,000 and I have never had it with a consultancy before.

Fiona Doyle · Chief Information Officer, Cambridge Mutual5 out of 5

Adjacent records