Project overview

Designing a scalable data warehouse on Azure

A major Spa and Personal Care franchise in North America partnered with VRIZE to modernize its fragmented data systems by building a cloud-based enterprise data warehouse on Microsoft Azure. The solution unified diverse data sources, automated reporting with Power BI, and enabled faster, data-driven decision-making across business functions. The new platform improved data quality, reduced reporting time, and enhanced customer insights — laying the foundation for advanced analytics and AI-driven personalization.

Client overview

Hand & Stone Massage and Facial Spa is a renowned wellness franchise with hundreds of locations across North America. Known for offering luxury spa services at affordable prices, the brand caters to a large and diverse customer base. With the expansion of digital channels and multiple customer touchpoints, Hand & Stone recognized the urgent need to upgrade its data infrastructure to keep pace with business demands and rising consumer expectations.

Business challenges

Despite having access to large volumes of data from various channels such as customer bookings, campaign performance, sales metrics, and loyalty programs, Hand & Stone was limited by a fragmented data ecosystem. Information was scattered across databases, APIs, and flat file systems, including CSVs and JSONs.

Reporting processes were manual, time-consuming, and lacked the agility required for real-time decision-making. Marketing teams struggled with inconsistent insights, incomplete views of cross-channel performance, and difficulty in segmenting and targeting customers effectively. There was an urgent need for a centralized, reliable, and scalable platform to unify data and empower the business with actionable intelligence.

Our methodology

VRIZE architected and delivered a modern enterprise data warehouse on Microsoft Azure, designed to handle both structured and semi-structured data with enterprise-grade performance, security, and scalability. During the implementation process, a strategic decision was made to continue using Azure Synapse, where the client was already operating, rather than building a new platform on Azure SQL Database. VRIZE further optimized the existing Synapse environment to maximize performance and scalability.

Phase 1 enhancements

• Converted full loads to incremental loads wherever possible, reducing overall batch processing time by more than 25%.
• Built pipelines to handle late-arriving fact changes, eliminating discrepancies against source system.

Phase 2 enhancements

• Optimizing existing pipelines by rewriting stored procedures.
• Converting data flows into stored procedures for better performance.
• Improving data distribution across nodes by selecting the right indexes.

Key architectural components

1. Data ingestion layer

VRIZE leveraged Azure Data Factory to orchestrate the ingestion of data from multiple sources, including RDBMS systems, third-party APIs, and flat files in formats such as CSVs and JSONs, into Azure Data Lake Storage. To ensure scalability and ease of maintenance, our teams developed pipelines and reusable components, enabling efficient handling of diverse data formats and simplifying future enhancements.

2. Data staging and transformation zones

In the next step, we implemented a multi-zone data architecture comprising raw, staging, and curated layers. In the Staging layer, data was systematically cleansed, validated, and enriched using a combination of Azure Data Flows and Databricks notebooks, ensuring that raw data was securely stored and readily accessible for further processing.

3. Data warehouse layer

The data warehouse was modeled using the Ralph Kimball methodology, leveraging a combination of Star and Snowflake schemas optimized for query responsiveness and analytical flexibility. From the Persistent Layer, the transformed data was loaded into the Data Warehouse Layer, and depending on the use case, the warehouse gave the flexibility to either employ a Star Schema or Snowflake Schema, enabling efficient querying and reporting. This architecture further helped organize data into fact tables, which stored measurable business metrics, as well as dimension tables, providing relevant contextual details for analysis.

4. Reporting and analytics layer

We integrated Power BI with the data warehouse to enable the development of automated, self-service dashboards tailored to the needs of the marketing team. To promote consistency and reusability, we also built centralized datasets and semantic models, empowering marketing teams to access reliable insights across campaign performance, customer behavior, and conversion trends with ease and more confidence.

5. Documentation & Governance

All architecture blueprints and escalation workflows were documented and maintained in a centralized knowledge repository, ensuring transparency and ease of collaboration across teams. Additionally, we also implemented DevOps pipelines to streamline deployments and enable seamless rollbacks, supporting a stable and agile development environment.

The VRIZE edge: a cut above

VRIZE delivered a tailored, efficient, and future-proof data solution by focusing on key strategic pillars:

• Accelerator-based delivery: VRIZE used its internal library of Azure pipeline templates and schema mapping frameworks to reduce implementation time and increase consistency.
• Marketing-centric data models: Unlike traditional data warehouses, the models were customized and purpose-built for marketing teams, making it easier to run campaign lift analysis, track attribution, and segment customers based on lifetime value.
• Future-ready design: The solution was designed to be flexible and future-ready, making it easy to add new data sources, integrate AI/ML models, or connect to downstream systems, all without the need to rebuild or rework the existing setup.
• Embedded Co-innovation: VRIZE worked closely with the marketing and IT teams, conducting weekly demos and design reviews, ensuring alignment throughout the engagement.