BAT Data Always at Hand: Transforming Reporting from Instability to Control

Tags:

Business Intelligence (BI)Data Engineering

About the project

New Data Structure for BAT – Clearing the Smoke for Transparent Numbers

British American Tobacco (BAT) is a global giant operating in 180 countries. The company, uniting over 50,000 people, is known for its brands: Dunhill, Lucky Strike, Pall Mall, and Vuse. Today, BAT is undergoing a massive transformation from traditional tobacco industry to electronic cigarettes and tobacco heating systems. The leader actively invests, develops technologies, and supports sustainable development.

Client Challenges and Needs:

The client faced difficulties with reporting. The system was unstable:

metrics were calculated with delays or inaccurately
performance was suboptimal
reports periodically became inaccessible

This created risks for operational processes and complicated decision-making.

The client needed a reliable tool that would provide stable, continuous access to accurate data in a business-friendly format.

How we solved the client's problem

Analytical reporting at BAT no longer brings chaos, as information is stored and processed in a multi-level system.

Dispatching mechanisms ensure reliable control and instant notifications in case of failures.
Our approach: We began with a detailed analysis of existing Power BI reports. We implemented initial corrections and partially stabilized the system’s operation. Key issues were identified in data sources and their transportation process to Power BI. Our specialists developed several solution options with data warehouse implementation using various tools.

Work results:

Reporting became accessible, comprehensible, and reliable, while data ceased to be a problem – now it works for the business, not against it.
Optimized Power BI: conducted an audit of existing reports, corrected calculation logic, improved performance.
Built Data Warehouse architecture: implemented a multi-level model with Bronze and Silver layers for efficient data storage and processing.
Set up ETL processes: automated data collection, cleaning, and transfer, minimizing risks of manual errors.
Implemented dispatching: created a mechanism for data quality control and loading stability with automatic notifications in case of failures.
Formed an infological model: clearly described all entities, their relationships, and analytics usage logic.