challenge

The holding's analytical infrastructure looked like this: 4 analysts served over 300 users across 60+ Power BI workspaces, connecting data from more than 40 source systems. Around 500 dashboards had accumulated in the system – and 40% of them were one-off ad-hoc requests that had turned into permanent digital deadweight. The math was relentless: 90% of the analytical team's time was spent on operational routine, and only 10% went to research and projects. The natural result – 30% of new analysts did not complete their probationary period. Business teams suffered too: any question regarding data turned into a waiting game of days instead of minutes.

Why now?

Attempts had been made to solve the problem, and the first try was an approach using Data Marts in Power BI: the data team builds "data kiosks," and business users build the visualizations themselves. The idea was right, but the execution failed. The Data Mart was built on raw data without a proper semantic model, and most business users simply could not figure out the technical nuances. The failure of the first attempt sharpened the understanding: the problem lay not in the choice of tool, but in the approach to data architecture and the very culture of working with it.

Fears and how we addressed them

The main fear was completely justified: initial efforts to build self-service analytics had already been made and had failed. The core question was: would business users actually use the new system, or would it suffer the same fate as the previous one? The answer was a bet on a fundamentally different level of simplicity: BigQuery as the single layer for all the Holding's data, and Looker with its LookML semantic layer ensuring a unified, clean data model. A Drag-and-Drop builder and an AI assistant utilizing natural language eliminated the technical barrier completely.

solution

We built a comprehensive analytical ecosystem consisting of three complementary components. First – baseline BI reporting for a seamless transition from the previous BI system. Second – a self-service platform based on Looker with a single source of truth: through a Drag-and-Drop report builder, every manager can independently construct visualizations and dashboards without involving analysts. Third – an intelligent AI assistant integrated with the platform: an employee simply asks a question in natural language within a chat, and the system instantly analyzes the data and builds the required chart or table.

result

• Saving 70% of time: The analyst team was finally freed from routine, shifting focus to deep forecasting and strategic insights.

• Instant Time-to-Insight: business teams no longer wait for days.

• A true Data-driven culture: data ceased to be a privilege of the technical department, becoming a daily working tool for every manager.