QANW (Quality Assured National Warranties)

The organisation

QANW (Quality Assured National Warranties) is one of the UK's leading providers of insurance-backed guarantees and deposit protection insurance, established in 1994 and headquartered in the UK with an office in Glasgow. QANW supports tradespeople and consumers across the domestic and commercial home improvement and renewable energy sectors, including schemes aligned to bodies such as RECC and MCS, giving homeowners confidence that work is protected if a contractor ceases trading.

The challenge

QANW was going live on a new vendor-built core platform hosted on AWS that holds current state only, offers no database access, and exposes data solely through its own user interface. From go-live day, the business would generate operational history it had no way to keep, with every week of delay representing data that could never be recovered. Compounding this, Excel spreadsheets were not just the reporting tool but also the integration channel to external partners and the de facto business glossary, meaning any solution had to replace several roles at once without disrupting a fixed go-live date.

The solution

We designed a cross-cloud bridge that reads the core platform's data lake through its own serverless SQL endpoint over a private connection, landing data in Azure as an immutable bronze layer before curating it upward into silver and gold, without the vendor system ever being touched, load-tested, or changed. A self-hosted integration runtime inside QANW's own Azure subscription bridges the two clouds, with Azure Data Factory driving incremental, watermark-based pulls into an ADLS Gen2 Delta Lake. Because bronze is directly queryable through serverless SQL, existing reporting could point at it from day one, before any modelling existed, acting as a safety net while the curated model was built. Silver consolidates current-state entities and gold delivers a curated star schema with business measures computed centrally, the semantic layer the estate had never had. The curated model was shaped around the questions an insurance and claims business actually has to answer, and the entities behind them: policy and guarantee registration, premium and scheme performance, claims notified, reserved, settled and declined, claim lifecycle timestamps, contractor and installer performance, and scheme-level exposure by product, region and accreditation body. Because bronze is immutable and append-only, every one of those measures can be recalculated as at any past date, which is what turns loss ratio, claims frequency, severity and development triangles from a spreadsheet estimate into an auditable number. Slowly changing dimensions preserve the state a policy or contractor was in at the moment a claim was notified, rather than only the state they are in today, and reserve movement is captured as a history rather than overwritten, so the business can show a regulator or a reinsurer how a position developed. Data protection was designed in rather than retrofitted. Claimant and homeowner personal data is isolated in the silver layer with column-level controls, gold serves analytics from de-identified keys wherever a report does not need identity, and partner sharing runs through a metered, audited boundary instead of emailed workbooks. Rather than pitch this as a proposal, we built the design as an interactive walkthrough application, an animated current-state diagram colour-coded by risk, a phase-by-phase target state, and two costed delivery options, so architecture and scope decisions were made against working software.

How we ran it

1. 01

   Discover

   Mapped the current-state-only vendor platform and identified that Excel doubled as the reporting tool, integration layer and business glossary all at once.

2. 02

   Design

   Designed a cross-cloud read boundary over the vendor's own serverless SQL layer, reachable from a self-hosted integration runtime inside the client's Azure subscription.

3. 03

   Model

   Shaped the curated layer around insurance and claims questions: policy and guarantee registration, claims notified, reserved, settled and declined, lifecycle timestamps, contractor performance and scheme exposure, with slowly changing dimensions so a claim can be read against the state of the world when it was notified.

4. 04

   Develop

   Built the medallion lakehouse, with Azure Data Factory driving incremental watermark pulls into an immutable bronze Delta Lake ready for interim reporting from go-live day.

5. 05

   Protect

   Isolated claimant and homeowner personal data in silver with column-level controls, served analytics from de-identified keys where identity is not needed, and replaced emailed workbooks with a metered, audited partner sharing boundary.

6. 06

   Deliver

   Presented the design as an interactive prototype application, letting QANW choose between a four-week focused build and an eight-week full day-one reporting estate.

The result

History capture began from go-live day, in QANW's own Azure tenant, closing the single most time-critical risk in the programme without any change request to the vendor or added load on the live system. Reporting logic scattered across Excel formulas is being replaced by a genuine semantic layer, so figures stop drifting between reports, and claims and scheme measures are computed once, centrally, rather than re-derived in each workbook. The business gains a defensible view of claims frequency, severity, settlement time and scheme exposure that can be restated as at any date, which is the basis for reserving discussions, reinsurer conversations and pricing reviews. A costed, sequenced roadmap covering forty-two day-one reports gives QANW a clear route to self-service Power BI and a governed path to metered, audited data sharing with external partners, replacing spreadsheet exchange with a controlled channel.