Segmentation is one of the most critical tasks in Pricing and Revenue Management, yet getting it right is the most daunting task. Throughout my career at Rainmaker, I have had the opportunity to work with extremely talented hotel and casino-hotel revenue managers who, via extensive experience at their properties and a strong analytical skillset, are able to identify segments with impressive accuracy. For instance, they are able to deduce that guests booking between two weeks and 60 days prior to the arrival day have a distinct behavior and sensitivity than other guests. Or that groups arriving on Mondays and staying two nights should be priced differently than groups arriving on Thursdays for the same length of stay. More often than not, data analyses performed by the Rainmaker Science team corroborate the revenue managers' insights.
However, a comprehensive segmentation approach requires classifying all the possible behaviors of individual guests or groups, not just the ones that stand out. If guests booking in the two-week to 60-day window exhibit a particular behavior, what about the ones booking beyond 60 days? Should they be their own segment or is there an opportunity to segment them further (e.g., 60 days to 90 days and beyond 90 days)? Does this behavior hold true for all the booking sources or only for a subset? And what about groups arriving on Monday but staying only one night? What about the ones staying three nights or more? If you consider all the possible lengths of stay, arrival dates, booking times and channels, there are thousands of possible combinations. Even the most sophisticated revenue managers cannot precisely segment more than a subsection of all their bookings because the task is simply too complex and time-consuming for a human being.
Furthermore, these segments may be correct now, but may not be six months down the road in a fast-changing field where new sales channels constantly open and hotels and casinos continuously seek new markets.
Segmentation requires a prolonged and ongoing effort. Which begs the question: When should it be reassessed or revised? Ideally, as often as possible, but revenue managers simply cannot realistically be expected to have the time to continuously keep an eye on their segmentation when they have so many other responsibilities and duties.
Many new revenue management gizmos on the market would like to lure you into believing that the laborious responsibility of segmenting guest or group behaviors should be undertaken by the revenue managers themselves. They market them as the essence of flexibility and offer the false promise of giving users the enviable capacity to custom-segment the data any way they see fit. And yet what these gizmos really do is merely dump one of the most intricate and critical tasks of revenue management onto the lap of their users. Manually segmenting the data is not only mind-numbing, but it also can cross the eyes of even the most knowledgeable, patient and disciplined among us.
The resulting segmentation may be good in best-case scenarios, but should the revenue management team be responsible if it is not? A sub-standard segmentation will always result in sub-standard performances of the revenue management system.
In a perfect world, the segmentation should be completely data-driven, automatic and dynamic in order to free the revenue manager from all the tedious grunt work and guesswork. In other words, a modern revenue management system should reduce the workload of the revenue management team not add to it.
For those who regularly track the work of academia and, by extension, strive to bring to their customers the best that technology has to offer, today's state-of-the-art tools present amazing algorithms that can automatically perform these tedious tasks much faster and more accurately than any human being. And doing it in an automated fashion means that the segmentation can be revised as often as necessary, making the process as fluid as the fast-changing environment in which hotels and casinos compete. Why waste all this human capital on tasks such as segmentation when technology is perfectly suited for this critical role?
The most effective and reliable revenue management system is one that constantly and relentlessly crunches numbers to ensure that the segmentation is always optimal. As time goes by and more historical data are available, the segmentation becomes even more precise as the revenue management system finds new patterns and refines existing ones. And all this is done based on data, flawlessly and dynamically, without human intervention or guesswork. For example, state-of-the-art algorithms continually sift through thousands of group quotes in order to extract patterns and segment behaviors based on attributes such as arrival day, arrival day of week, peak days, length of stay, lead time, conversion rate of the booking channels, market types, non-room revenue, rates of rooms compliance, and more. This is the way segmentation should be conducted. It is segmentation so precise and so surgical that we at Rainmaker refer to it as micro-segmentation.
To conclude this post, I would like to quote an article from The Economist* on the 150th anniversary of the publication of Maxwell's equations of electromagnetism:
"He showed that nature ought not to be taken at face value, and that she can be cajoled into revealing her hidden charms so long as the entreaties are whispered in mathematical verses."
We believe the same lessons apply to revenue management and more generally to our pursuit of better understanding how pricing drives demand and profit.