Managing 70 airports spread across 14 countries, welcoming 320 million passengers a year and generating €4.5 billion in revenue: Vinci Airports’ operational challenge is commensurate with its global footprint.
The group has embarked on a large-scale data and AI transformation, leveraging Google Cloud Platform.
Dependence on Mastering Passenger Flows
The Vinci Airports model rests on two pillars: aeronautical charges, representing about 50% of revenue, and commercial revenues generated by passenger spending (duty-free, dining, parking), which account for 30–40%.
This structure requires precise anticipation of passenger flows. From this predictive capability cascade the sizing of teams, calibration of the commercial offer, financial projections, and investment strategy.
The challenge lay in the extreme heterogeneity of the network. Each airport has its own IT systems, local constraints, and operational specifics. Data were scattered, non-standardized, and siloed.
A Data Factory to Federate the Network Data
To meet this challenge, Vinci Airports launched in 2023 the construction of a global Data Factory hosted on Google Cloud Platform.
This infrastructure centralizes data collection from heterogeneous systems across 14 countries, automates data quality checks, and harmonizes data structures to enable reliable cross-platform comparisons.
The architecture relies on the proven ingredients of Google Cloud: BigQuery for storage and analysis of massive datasets, Vertex AI for training and deploying predictive models, Cloud Run and Streamlit to provide accessible business interfaces, Cloud Storage for centralized model management, and Cloud Build for continuous integration.
This architectural choice allows airports to keep their local tools while exposing their data in a common repository, where it is cleaned, structured and usable to feed dashboards, analyses and AI models.
Three Use Cases at the Heart of the Transformation
The first and primary use case focuses on predicting passenger traffic. Multi-scale predictive models were designed to meet the needs of each management level: annual projections for the executive team, weekly or daily views for operational teams, and localized airport-by-airport analyses. These models cross traffic history, exogenous variables and faint signals to simulate trajectories and optimize trade-offs.
The second axis concerns operational efficiency. By analyzing scanned boarding passes, it is now possible to anticipate arrivals at security checkpoints. Cross-referenced with the capacities to process queues, these data enable real-time staffing adjustments with the aim of keeping wait times under ten minutes.
The third use case touches on commercial optimization. By combining traffic data and purchasing behaviors, Vinci Airports identifies consumption profiles by passenger typologies and destinations. A British traveler in transit does not spend in the same way as a French passenger on a domestic flight: these patterns allow retailers to adjust assortments to the actual profile of their customers.
Data Quality at the Heart of the Project
Artefact intervened from the early stages of the project with a collaborative approach that involved operations teams from day one.
This approach is described by Benoît Forest, Director of Operations and Data at Vinci Airports, as a sine qua non: “The onboarding timing of the operational teams is very important. It must begin on day one. This ensures that data scientists understand business concerns and incorporate highly operational needs into the solution design.”
Data quality represented another major challenge, with automated mechanisms to detect missing files, verify data structure, and perform integrity tests, with automatic alerts in case of anomalies.
These safeguards, operated entirely on Google Cloud Platform, prevent silent drifts that erode business trust in the models.
“I think the success of this project lies in understanding the strategic business challenge, defining the scope, deploying the solution, training, and today the tool is used daily by business teams,” sums up Benoît Forest.
A Step Towards Generative AI
Vinci Airports then moves into a second phase focused on systematic prediction and generative AI. Three use cases are under study: a GPT-powered assistant integrated into dashboards, a conversational data querying mode, and an automatic extraction system for documentary content (procedures, audits, reports).
The aim is to go beyond the limits of current Power BI dashboards to enable every employee to query the entire dataset directly through autonomous AI agents capable of providing answers to complex questions without requiring additional development.
“AI enables us here to move from local intuition to shared knowledge, without replacing teams, but by giving them the means to save time and focus on decision-making,” concludes Benoît Forest.
