How Vipra Software transformed an insurance group's reporting from a 6-hour daily ordeal to live 15-minute C-suite dashboards, deploying a modern Snowflake + dbt + Looker analytics stack across 9 business lines.
A diversified insurance group operating across 9 business lines — life, general, commercial, health, travel, marine, property, liability, and reinsurance — had a reporting architecture that was consuming more analyst time than it was saving. Every morning, a team of 8 analysts spent 4–6 hours extracting data from 9 policy administration systems, reconciling premium and claims figures across divisions, and manually assembling the Group Performance Pack that CXOs reviewed in their 9am operational meeting.
The manual assembly process introduced regular errors. Figures shared in the executive meeting would be revised during the day as analysts discovered reconciliation discrepancies, creating a credibility problem for the finance function. CXO requests for ad-hoc analysis — "show me combined ratio trends by product line for the past 18 months" — could take days to fulfil, as the underlying data needed to support such queries lived in formats that required bespoke extraction and manipulation for every new question.
The analyst team was also at capacity. Eight analysts spending 30% of their working week on mechanical data assembly had no capacity for the value-added analysis the business needed. Attrition risk was elevated — experienced insurance analysts were not joining the organisation to build Excel pivot tables from system extracts.
Vipra Software selected the Snowflake + dbt + Looker stack as the best-fit combination for an insurance analytics use case: Snowflake's separation of storage and compute for cost-optimised multi-workload analytics, dbt's SQL-native transformation framework enabling the finance team to own and validate transformation logic, and Looker's LookML semantic layer providing governed self-service without ungoverned sprawl.
The data architecture follows a clean medallion pattern implemented in Snowflake: Raw (source replicas), Staging (typed and cleaned), and Mart (business-ready insurance domain model). The dbt transformation layer enforces single definitions for all KPIs — the combined ratio calculated for the life business line uses the identical formula as the commercial line, eliminating the definitional inconsistencies that had caused reconciliation failures in the legacy process.
Looker's LookML semantic layer adds a governed abstraction above the Snowflake mart tables, translating the physical data model into business-domain concepts that insurance professionals recognise. Underwriters explore data using insurance terminology (written premium, loss development, IBNR reserves) without needing to understand the underlying data model. All LookML measures reference the certified dbt metric definitions — changes to business definitions propagate from dbt through LookML to every dashboard simultaneously, eliminating definition drift.
The daily Group Performance Pack is now available at 7:15am, fully automated, with no analyst involvement — compared to the previous 10am–12pm delivery window requiring 6 hours of manual work. The 8 analysts who previously assembled it have been redeployed to commercial pricing analysis and claims pattern detection — work the business had been unable to resource for 18 months.
Executive confidence in the numbers increased materially within the first month. The absence of intra-day revisions — a regular feature of the manual process — removed a persistent credibility friction that had been cited in multiple board-level discussions. CXO ad-hoc analysis requests now receive same-day responses as analysts use Looker's self-service explorer rather than bespoke extraction and manipulation.
The Snowflake cost model delivered an additional benefit: the previous on-premises data warehouse had been running at 100% utilisation continuously. Snowflake's auto-suspend reduces compute costs to zero outside peak query windows, delivering a 55% cost reduction compared to the previous always-on infrastructure even after adding the dbt and Looker licensing costs.