:focal())
With more than two million visitors per year, our customer is one of Europe’s most awarded leisure parks, it is also a highly complex ecosystem where restaurants, shops, ticketing systems, and events all operate simultaneously, generating large volumes of data every single day.
The Challenge
This diversity creates a beautiful visitor experience, but it also creates a real operational challenge. At certain moments of the day, some restaurants become overcrowded while others remain half empty, even though they have available capacity. Consequently, businesses face longer wait times, increased operational pressure on staff, and lost revenue potential.
To address this challenge, the organization partnered with Beyond Data Group to design and implement a modern data and analytics platform powered by Microsoft Fabric and Power BI.
The Solution
We built a fully integrated predictive analytics solution within the Microsoft ecosystem, combining reliable data, advanced forecasting models, and real-time reporting through Power BI, all developed inside Microsoft Fabric.
Our goal was to:
Predict, one week in advance, the expected affluence for each restaurant and horeca location.
Enable the park to influence visitor choices through their app and on-site signals.
Improve visitor experience while optimizing operational performance.
Phase 1: Creating a clear and reliable data foundation
Before predicting anything, we first needed to understand the operational reality in detail.
All relevant data sources were centralized into a Microsoft Fabric Lakehouse, including:
Visitor attendance
Horeca sales and transactions
Historical affluence patterns
Time of day and seasonality
Day type (weekday, weekend, holiday)
Weather conditions
Previously, data existed in fragmented systems and isolated SQL queries. By centralizing and harmonizing everything into one structured and governed environment, we ensured that all departments worked from a single, reliable source of truth.
This step was essential, since without clean and structured data, AI cannot generate meaningful predictions.
Phase 2: Predicting affluence
Once the data foundation was ready, we developed machine learning models capable of forecasting the number of visitors expected in each horeca location for the upcoming week (illustrative graph bellow).
The model learns from historical patterns and continuously analyses how visitor behaviour changes depending on:
Time of day
Seasonality
Weather conditions
Special events
Overall attendance
:focal())
An important nuance is that not all restaurants have the same capacity. One restaurant might be at 100% capacity while another is only at 50%, even if the absolute number of visitors is similar.
To make this understandable and actionable, we integrated capacity levels directly into the model and visualizations, so predictions reflect real operational pressure, not just visitor counts.
The model was trained, tested, and optimized carefully to avoid overfitting and ensure stable performance in real-world conditions.
Most importantly, everything runs locally inside a secure environment. No external AI services are used, ensuring full data confidentiality and governance.
Phase 3: Real-time visualization and decision support
Predictions alone are not enough if they are not accessible and understandable.
That is why we connected the predictive models directly to interactive Power BI dashboards, making it immediately visible where crowding is expected and where capacity remains available.
This way operational teams can anticipate staffing needs, stock adjustments, and service organization.
At the same time, visitors can receive recommendations via the park’s mobile app or on-site communication, subtly influencing their choices and distributing traffic more evenly, enhancing their experience by reducing friction and waiting time.
:focal())
The Impact
By transforming raw operational data into predictive intelligence, our customer now benefits from:
Better crowd distribution across horeca locations
Improved visitor satisfaction due to reduced waiting times
More efficient staff allocation
Optimized stock and logistics planning
Real-time visibility on operational performance
This project demonstrates that AI does not need to be abstract or overly technical. When built on a strong data foundation and integrated directly into operational processes, predictive analytics becomes a powerful and practical tool for improving both experience and profitability.