Unified Cloud Data Platform: a Single Source of Truth for Retail

Replace a fragmented legacy data warehouse (SAP BW) and manual reporting with one cloud data platform that serves as the single source of truth for commercial and logistics analytics of a nationwide supermarket chain. Goals: put the data of Commerce and Logistics in one storage with one dictionary of KPIs, so that business units stop arguing about numbers; shorten the path from source system to report; let business analysts build reports themselves instead of waiting for SAP developers; automate report preparation; cut the time to deliver new analytics; and lay a foundation for the next domains (pricing, procurement, e-commerce, finance, HR) and for ML tools.

• Data quality: one KPI dictionary with an agreed LFL store definition; daily reconciliation with SAP BW during migration.
• New analytics in ~1 week instead of ~1 month.
• 800 users onboarded (July 2026).
• Self-service instead of SAP-only development: the Operations domain was delivered by business analysts; ad hoc analytics in SQL and Python.
• Data ready daily by ~9:00. 110 datasets, >10 TB, 25 report views, 42 metrics, 50+ dashboards.
• Estimated saving of 1-3 hours per analyst per working day: 6,000-19,000 hours a year across 25 analysts (3-9 FTE, about a quarter of an analyst's day). The estimate is conservative: it excludes the 800 users and data that did not exist before. 

Microsoft Fabric (F64 capacity): OneLake (hot/cold layers), Data Factory pipelines, Lakehouse/Warehouse, semantic models and data marts, Power BI. Source systems: SAP S/4HANA, SAP EWM (warehouse management), 1C, POS systems; reporting of the Korzinka Go application was moved from Tableau to Fabric + Power BI. Legacy SAP BW warehouse was decommissioned on 1 October 2026.

The platform was built while the legacy SAP BW kept running, with sources in SAP S/4HANA, SAP EWM, 1C and POS and more than 10 TB of data. Challenges: reconciling KPIs between BW and the new platform without discrepancies before decommissioning; capacity needs turned out higher than planned and resource pools had to be tuned during the project; standardizing in-house data modeling, workspaces and access for 800 users; building system analysis and data modeling competence in-house; and embedding new reports in the daily work of category managers so that adoption keeps pace with development.

Initial situation

Before the project, 25 analysts in five domains prepared reports manually: exports from systems, consolidation in Excel and in local Power BI models, which cost extra working hours every day. Reports lived on the legacy SAP BW warehouse, where only SAP developers could build new ones. Data was scattered and refreshed slowly; KPI definitions differed between units - for example, there were several methods to classify a store as like-for-like (LFL), so figures did not match and gave a false picture of efficiency. New analytics took about a month to deliver. With growing data volumes and a chain of 150+ stores in 12 regions, this prevented timely decisions.

Implemented solution

A unified cloud data platform on Microsoft Fabric for the Commerce and Logistics domains (including distribution-centre operations and stock visibility): 110 datasets, more than 10 TB, 25 report views with 42 metrics and 50+ dashboards. Integrations with SAP S/4HANA, SAP EWM, 1C and POS were built; Korzinka Go reporting moved from Tableau to Fabric + Power BI. Documentation was unified and report preparation automated.

Implementation specifics

Delivery went in stages: design and architecture, integrations and datasets, BI reports, then change management. An in-house team of about 12 (architect, data engineers, BI lead, analysts) built standard domains, the integrator took complex cases. An analytics school with a grading system supports talent. Adoption was driven through live meetings and webinars plus separate training for top management, and tracked weekly (usage, SLA, report performance). Access is managed centrally by role and department. To ensure a seamless transition to the new data platform while both systems operated in parallel, key Commerce metrics were reconciled daily against SAP BW data. This daily comparison validated the accuracy of reporting migration to the new platform. SAP BW reports were phased out and the warehouse was fully decommissioned on 1 October 2026. The hardest part was embedding new reports into the daily work of category managers.

Changes after implementation

Analysts no longer spend manual hours on report preparation: the estimated saving is 1-3 hours per analyst per working day, i.e. 6,000-19,000 hours a year across 25 analysts (3-9 FTE, about a quarter of an analyst's day). Data is available by ~9:00 AM; new analytics are delivered in ~1 week instead of ~1 month. The Operations domain was closed entirely through self-service, with ad hoc analytics in SQL/Python. Data quality became managed: one KPI dictionary with an agr

The platform covers the whole retail chain: 150+ stores in 12 regions of Uzbekistan, and distribution-centre operations.