Accelerated billing data processing time from 10 hours to 3 hours
Reduced new solution delivery time by 40% with code-first approach
Unlocked €100K+ annually by optimizing cloud infrastructure
Finance becomes one of Delivery Hero's biggest data mesh consumers
SAFIA AI agent automates monthly P&L reviews for 15 financial analysts
Delivery Hero scales AI-driven finance by modernizing SAP BW with Google BigQuery.
Delivery Hero scales AI-driven finance by modernizing SAP BW with Google BigQuery.
When finance couldn't keep up with the business
Delivery Hero connects tens of millions of customers with restaurants, grocery stores, and local shops across nearly 65 countries on four continents. At a company processing billions of orders across 11 global brands, financial data is mission-critical—and for years, getting to it reliably was anything but fast.
The enterprise runs on this data, but unlocking it from the core ERP systems to feed the broader, cloud-based organization presented a massive engineering challenge. Acquisition-driven growth meant each new brand brought its own technology landscape. Sharing mission-critical finance data across the organization required labor-intensive CSV pipelines that were becoming costly to manage at scale.
The catalyst for the migration was data velocity, and the testing ground was billing. While Delivery Hero’s SAP systems handled trillions of data points flawlessly, replicating that across the wider organization took up to 10 hours.
Finance possessed an incredibly rich dataset, but the architectural boundaries of our ERP made sharing it with a cloud-native organization complex and costly. We didn't need to find our data; we needed to democratize it.
Mario Sejdia
Director of Digital Transformation, Delivery Hero
Moving to BigQuery wasn't an infrastructure rescue mission; it was a strategic necessity to achieve the near-real-time data sharing the rest of the business demanded. A proof-of-concept changed the conversation. The team migrated billing, ran both architectures in parallel for three months, and saw a clear path forward. The decision was made: move everything.
Building the Finance Data Hub
The migration, led by Uttpesh Vyas, principal architect of the migration and head of data engineering, was executed entirely in-house by a team of seven people over 12 to 24 months—while simultaneously maintaining existing SAP-based reporting for the business. "We were managing both systems in parallel," Uttpesh recalls. "It was pretty much double the work."
To support our enormous transaction volumes, we needed a platform built for limitless scale. Moving to Google Cloud delivered that scalability and marked a significant paradigm shift in data sovereignty, allowing us to move toward a modular, code-first architecture in which business logic is a portable, transparent asset.
Uttpesh Vyas
Manager Finance Data Engineering, Delivery Hero
BigQuery was the natural fit for its serverless architecture. Because it scales elastically without manual intervention, the Data and Analytics team could focus on building data products/assets rather than managing infrastructure. "With BigQuery, we don't have to worry about scaling at all," Uttpesh explains. "It seamlessly handles our massive data growth without interrupting business operations."
The migration also represented a profound learning journey. Transitioning from legacy systems required a completely new, code-first mindset. The team successfully upskilled in modern data engineering—embracing standard SQL using dbt, Git-based version control, and Python—and explored new horizons with Looker and Tableau, transforming how to deploy robust solutions.
To accelerate the migration of 337 SAP tables, the team drew on the open-source Google Cloud Cortex Framework—not off-the-shelf, but customized to fit Delivery Hero's specific data model. Cortex provided pre-built SAP logic and reusable data product foundations. For transformation and quality assurance, the team adopted dbt, which enabled them to clean, test, and document data models in standard SQL and track every change through Git.
For systems outside the SAP transaction layer—like Workday, Salesforce, OneStream, and SAP's cloud products Concur and Ariba—the team built Python-based Apache Airflow operators to connect directly to BigQuery, finally replacing the manual CSV exports that had been the fragile bridge between finance and the rest of the business.
The result is Delivery Hero's Finance Data Hub: a dedicated node within the company's broader data mesh, structured across three layers.
- Foundation Data Products supply reference data—customer primary records, exchange rates, and country codes
- Business Data Products transform that foundation into use-case-specific outputs: trial balances, spend analysis, consolidated financials
- Analytical Data Products sit at the top, feeding Looker and Tableau with ready-to-consume insights
Finance analysts can now self-serve in Looker, querying accounts, building dashboards, and running ad hoc analysis in seconds rather than hours.
A finance team that now looks forward
The results came quickly. Billing data that once took ten hours to process now runs in three hours; other domains have improved even further. Solution delivery, from business request to production, dropped from an average of five months to one quarter. By optimizing on-premises infrastructure, the team unlocked more than €100,000 in annual cost savings, which were reinvested.
But the more durable transformation is cultural. Today, Finance is one of the largest consumers of Delivery Hero's entire Google Cloud infrastructure, and one of its most active internal advocates. When the platform team introduced a new Google tool six months ago, finance was the first to try it. "We are continuously looking for ways to innovate," Mario notes. "The team is finding opportunities to optimize, scale, and launch new ideas faster than ever."
The platform is now the foundation for AI-driven financial intelligence. In Q1 2026, the team deployed SAFIA (Smart Assistant for Financial Insights Analytics), a Gemini-powered agent built on the Google Agent Development Kit.
It unlocked a completely new way of working. We moved from static reporting to actually building with AI, leveraging the platform to develop Finance agents.
Mario Sejdia
Director of Digital Transformation, Delivery Hero
Each month, SAFIA automates the performance review process for over a dozen financial analysts: generating P&L summaries, brand-level commentary, and market comparisons in natural language from the BigQuery data layer. The next phase will integrate non-financial signals—competitive dynamics, weather events, macroeconomic trends—to add the "why" behind the numbers.
It is a shift that Mario puts simply: “We moved from static reporting to leveraging the platform to build AI agents with our data. This is the new era of finance.”
Delivery Hero connects millions of customers with local shops and restaurants in ~ 65 countries.
Industry: Retail
Location: Germany
Products: BigQuery, BigQuery Connector for SAP, Looker, Cortex Framework, Agent Development Kit, Gemini Enterprise
