Hospitality companies have had to adjust in recent years to significant disruptions to their business. Reservation lead times shrank, while last-minute bookings grew.
A large hospitality brand saw this situation as an opportunity to rethink revenue forecast models across its myriad brands and thousands of locations. Working with ZS, a trusted partner, is helping the company reliably anticipate demand at its properties and accurately interpret data for short- and long-term financial projections. Planning teams are better able to allocate resources and provide operational guidance for downstream decision-making, including marketing campaigns and promotions.
As the economy faltered during and rebounded from the COVID-19 pandemic, the hospitality brand saw its revenue forecasting fall short. Not only was the system producing significant variances, but its dependence on traditional measures also meant it could not predict and adapt to weekly/monthly demand seasonality, factor in recent booking behavior shifts or capture external variables such as local events and weather.
ZS developed a proprietary, novel approach that uses advanced analytics and robust data sources and builds on ZS domain expertise.
We focused on three aspects:
- Use of a robust and well-integrated set of data sources. Daily, weekly and monthly variables were created to allow the system to identify seasonal periods. Supplemental markers indicate whether an event or holiday (local, state or national) is confined to a single day or likely to spread over several days. This “holiday effect” is an important phenomenon in the hospitality industry, given its potential to drive revenue gains for sales and marketing teams that develop strategic campaigns to reach targeted audiences.
The noise and the impact of outlier events such as COVID-19 and local/geographic-specific variations were removed. Additional data sources were incorporated, notably CBRE hospitality industry macro trends, STR premium data benchmarking, analytics and marketplace insights.
- Use of advanced analytical techniques. The new approach combines advanced regression analysis and machine learning models through Driverless Forecast, a ZS proprietary tool with the ability to model external features such as events, holidays and weather. Forecast metrics and algorithms were designed to incorporate historic trends and recent booking pattern changes. A ricochet approach allows the system to make daily revenue predictions and then use that data to train the forecast for the next day. This resulted in a robust modeling and validation approach for yielding highly accurate forecasts.
- Streamlined and efficient execution of forecasting analytics. Automated processes were created and deployed at scale to allow revenue teams to make frequent forecast refreshes with quick turnaround times. (The number of steps was trimmed from 13 to three.) Forecast accuracy definitions align with business requirements.
ZS followed an iterative process with the new approach, leveraging components from our Driverless Forecast solution. The project team regularly met with the client’s revenue leaders to keep them apprised of the core methodology and offer post-deployment support for questions and feedback. The built solution also can be scaled across different brands in a very quick manner.
The new revenue forecasting solution delivered right out of the gate, consistently capturing underlying booking patterns and scaling to support more than 10 brands across the hospitality brand’s portfolio. In the first three months following deployment, the company reported accuracy rates of 99%. Turnaround times for forecasts fell 80% as ZS streamlined the process by removing redundancies and minimizing the manual effort needed to produce them.