This year was different.
This was the first time we’ve heard the conversation shift from “the promise of machine learning” to a serious conversation about what is the best use of this technology in Hospitality Revenue Management.
It was refreshing to hear people talking about real-world application and digging deeper (buttressed with an increased understanding of exactly what machine learning is) into which Hospitality problems are seemingly tailor-made for a machine learning approach.
Shortly after we presented in the lightning round selected result from the 2019 Revenue Management Trends report (download here) there was an especially great conversation with two senior Revenue Management leaders: one from gaming, the other from a luxury hotel chain.
We were discussing previous attempts made at group forecasting, and the disappointing results. It seems that failing to accurately predict group business is a near-universal experience for hotel Revenue Managers. And that’s not surprising. Group forecasting is just a lot harder than transient. It’s “chunky” and inconsistent. Plus, different types of groups (and even groups within a segment) act differently.
Simply put, group demand is just harder to predict.
As the two Revenue Management leaders discussed past failings in trying to predict group business, our conversation turned to new approaches that could help overcome the group forecasting challenge.
We reflected on the complexity surrounding groups and the significant increase in factors that impact demand for the segment—and the possibility of applying machine learning to group forecasting. This technology, we agreed, could be the answer for predicting highly variable group business.
The gaming Revenue Management leader started describing all the factors that could influence how a specific group builds over time. He talked about how he has observed higher booking against the block for family reunions from the South, whereas family reunions from the Northeast may only book up a fraction. He qualified this observation, however, by saying, “At least, that’s what I think, but machine learning should be able to help me determine that!”
He was exactly right. The more variables that could impact the behavior of specific customers and segments, the more difficult it is to accurately predict demand using time series forecasting—but that’s not the case with machine learning.
A machine learning approach to forecasting can process all those factors, determine which are the most important in terms of driving demand, and leverage those in its prediction.
In fact, Revenue Analytics has deployed machine learning on our platform specifically against the problem of group forecasting, and we’re seeing that application result in greater accuracy that drives incrementally better pricing and inventory decisions—whether they’re made by a system or a human.
Machine learning provides a much-needed leap in the right direction for improved group forecasts.
It was a great conversation, and it was indicative of the subtle shift in the Hospitality industry’s dance with machine learning. It’s no longer a promise, but something that is here, and with very real application against a perennial problem for hoteliers.
Most importantly of all, it works. Machine learning goes a long way toward answering this age-old hospitality problem of accurately forecasting groups.
Frankly, we were thrilled to no longer be talking about the promise of machine learning at ROC. Instead, we were talking about how the Revenue Management leaders in our industry are finally able to utilize it.
Have questions? Let’s grab some time to chat.