How does Simulation work?

In Simulation, Dexibit provides a forecast of what would have happened in normal or usual times, from which you can simulate scenarios. This provides a hybrid approach of machine learning and manual assumptions.

Firstly, the machine learning forecast works by using statistics to find patterns in your venue’s own data history (preferably, referencing at least 12 months of daily data – though more is preferable), to train a model specifically designed to predict visitation. Then, you can manually adjust this base visitation into a scenario, using playable controls. These controls use a combination of scaling factors, alternate predictions and factor insights.

Because these simulations are based on an underlying forecast using machine learning models, the simulation will provide a quality indicator for the base prediction accuracy.


How accurate is a simulation?

Simulations are built from a machine learning based forecast of what would have been, during normal or usual times. The underlying forecast accuracy is displayed as a bronze (70 – 80%), bronze (70 – 80%) or gold (90%+) accuracy, measured on a weekly basis. These should be considered as a contingency on the final result, as a +/- range.

Premium plans have access to the full machine learning model, which features all sorts of factors into the forecast and results in higher accuracies. All other plans use a basic model incorporating time series, public holidays and school terms for the venue’s specific region. Providing additional years of data may increase underlying forecast accuracy.

From this underlying forecast, manual controls are then used to produce a simulation for scenario planning based on the user’s business judgment.

During unpredictable or unprecedented times, consider scenarios a hypothesis rather than a prediction for a wide range of possibilities.


What assumptions can I use for a simulation?

The following assumptions can be used to create an assumption:

  • Closures - Initial reopening, rolling subsequent closures

  • Visitor origin - International tourism (by region), domestic visitor origin (local, drive in or domestic fly in)

  • Capacity

  • Opening days and hours

  • Demand - Pent up, slow growth and subsequent reopenings

  • Custom segments



Sourcing data for Simulation

Training data for the underlying forecast comes from your venue’s visitation data, based on the data sources and business rules you have configured under venue management. Historic visitation data is visible in the simulation as an orange line, which can be toggled on or off by clicking on ‘Actual visitation’ in the legend. The baseline forecast is visible in the simulation as a dotted shadow line, which can be toggled on or off by clicking on ‘Forecast’ in the legend.

Context for the model comes from an almanac of calendar events happening in and around the venue.

Did this answer your question?