TECH: French Champagne and Rock and Roll

 

Ideas at Dexibit come from many places – most often, driven by visitor attractions we work with, in their pursuit for innovative ways to predict and analyze visitor behavior. Ideas for new forecast models hold a particularly special place in our mission to power our cultural future. So far, we’ve built forecasts that project daily visitation, hourly attendance, advance pass attrition and simulate activities such as exhibitions. Recently, we’ve been working hard on forecasting an important metric in for profit enterprise and equally so for commercial sustainability in non profit or government venues: revenue. Enter our new revenue outlook model, which started life as an idea inspired by two great things in life: French champagne and rock and roll.

At the Rock and Roll Hall of Fame in Cleveland Ohio, visitors can experience a legendary music experience like no other and marvel at the world’s greatest rock stars. The Rock’s team were refined in financial planning for revenue, investing time in producing detailed analysis to support accuracy and converting these into daily targets by line of business. The complexity of the business model makes this a tricky problem to solve – some revenue is linked to ticketing, from ticketed admissions of various products, types and campaigns; others lines of business have a different dynamic that results from the need to convert visitors into commercial outcomes such as cafe or retail. The revenue forecasts needed to be not only accurate but granular and ideally much faster to generate than by manual analysis.    

Enter: champagne. A dataset well known in the data science world when looking at financial projections is the historic sales data of a French vineyard, champagne brand Perrin Freres, which dates back to the early sixties. Given the crossovers between entertainment and education, it turned out there were many commonalities between elements of our data sets for visitor attractions and that of Perrin Freres – even if a typical visitor attraction is a far more complex business model than wine. This classical data became the beginning of our exploration into how we could get accurate results – despite the unique challenges of the fact visitor attractions host myriad market offerings. Some lines of business like ticket products, types and campaigns respond to factors influencing the visitor’s decision in getting to the venue, other products and services are specific to the influences on visitor’s behavior within their experience. 

Trialing various techniques over time, we combined the traditional time series champagne data approach with various machine learning methods to build our final revenue outlook model. The techniques evolved from the champagne to look purely at top line revenue set a consistent pattern that our more complex algorithms could learn from easily. The depth of these machine learning methods then let us capture and analyze influences on visitor behavior, whether these be external such as the weather seasonality or internal such as the venue’s events and how these affected revenue. The two became a hybrid approach combining a traditional statistician’s classical time series forecasting and a machine learning toolkit. The result gave us the best of both worlds to forecast revenue specifically for visitor attractions.

Back at the Rock and Roll Hall of Fame, we set the new model to work, predicting revenue down to the day and up to a year or more ahead and differentiating between various lines of business. As we began beta testing on venues further afield across the visitor attractions sector, we found that at least 18 months or more of recent historical data is required in order to train the model for optimal results capturing year on year growth trends. We also discovered new applications for future features, such as the need to predict cash receipt versus revenue recognition, especially in venues with lengthy advance pass booking windows.

The revenue outlook forecast model is now available in Dexibit – check your forecasting module for more, or talk to your success manager for details. And as you may have guessed, we celebrated the launch with a bottle of the finest.