Improving the Accuracy of Forecasts: Time to Ditch the Gut Feeling Approach
By Kell Sloan
At the property level, hotel revenue management practices haven't changed much since the early 1980s when yield management and price optimization became mainstream practices.
When tourism season approached, a hotel simply raised their rates. Occasionally, weekends or special events required a minimum stay, or a hotel owner would install a heart shaped vibrating bed, mirrors on the ceiling and a whirlpool tub and charge a premium for the "Honeymoon Suite." Then, sell a bottle of Baby Duck sparkling wine and a roll of tokens for the vibrating bed to make a little ancillary income.
A gut feeling indicates that there must be some market where there is unrealized pent-up demand and a healthy profit for vibrating beds. However, without intimate market knowledge, forecasting future demand relies heavily on preconceptions of this unique market and spreadsheet modeling that is sophisticated enough to recognize a variety of intrinsic factors that might influence demand and price sensitivity.
As a result, the accuracy of revenue forecasts for a heart-shaped bed market or any other market is dubious at best. Adopting the problematic forecast would lead to faulty operational planning and resource budgeting - dramatically affecting top and bottom-line performance and the revenue manager's employment prospects.
Yet, this the very situation that revenue management, sales and marketing and operational teams alike are working through in a post-COVID world. The last year and a half of unpredictable acts of government and the consequently sporadic demand have resulted in many revenue managers expressing low to no confidence in historical demand-based pricing models and occupancy trends. Despite this lack of confidence, the predominant question is "how do revenue management and operational teams develop a consistent and clear understanding of future demand?"
Future demand hinges on two fundamental concepts. Firstly, understanding who the guests staying at a property are. And secondly, understanding what forces are driving changes in local market demand and price sensitivity. Simply stated, guests and driving forces are integral components of every market ecosystem.
So, can predictive analytics or real-time data scraped from a variety of internal and external sources such as a property management systems and social media feeds be used to analyze both structured (price of a heart-shaped bed on Tuesday) and unstructured data (online review) and in turn predict future business levels?
Plenty of Data Yet Spreadsheets Rule Forecasts
Make no mistake, even 30 room motels located in rural out-of-the-way areas are awash with incredibly valuable structured and unstructured customer data and potentially profitable insights. Property management systems such as Micros Opera, whose strength is in its transaction reporting functions, have provided revenue managers with plenty of data to create revenue and occupancy forecasts.
For all that, Skift Research's The Hotel Revenue Management Landscape 2019, concludes that over 80 percent of hotels world-wide do not use anything more sophisticated than an excel spreadsheet to establish pricing or revenue strategies. Head shakingly, there are still plenty of branded and independent hoteliers whose entire revenue strategy is to price check online travel agents (OTA's) daily and undercut comparable competitive set products.
Arguing with the die-hard "undercut the competition by five dollars a night" general manager or owner about their "any paying guest is the right guest" or "gut feeling" approach to revenue management is an exercise in patience and swallowed frustration. Pointing out that hotel operators who make sensible decisions regarding pricing and distribution tend to dominate market share, operate cost efficiently, return a profit and have award-winning guest service scores can be like dropping that bottle of Baby Duck sparkling wine on the arch of the foot. Repeatedly. Profit-driven revenue managers do not like to leave money on the table. Thus, to be profitable a hotel must have some way of forecasting future demand and possess at least a degree of insight into guest's spending habits.
Why Do We Look Backward to Predict Future Demand?
Ironically, to predict the future many hotels look backward. Revenue management has relied on using historical trends such as bookings, heat maps, seasonality, holidays, feeder markets, pace by segment, pace by distribution channel, pricing as well as groups and special events to predict the future. Applying a little simple exponential smoothing followed by an excel linear regression of arrivals and room nights and the groundwork for the annual budget and forecast are complete.
Further refinement into segments such as transient, corporate, locally negotiated, seniors and/or distribution channels along with allocations for demand generators such as promotions, loyalty, group discounts are shoehorned into rows and columns and in a relative short time an annual sales forecast, revenue budget and a static 12-month inventory management plan is created, tweaked and gospelized.
If only inventory management was all there is to revenue management, life would be much simpler. Transactional in nature, inventory management is the static process of managing rates to meet budgeted goals. In other words, establishing pricing to sell a certain number of rooms that fit into predetermined segments or "buckets" in the annual budget.
Revenue management, like any form of business intelligence, is a dynamic process. The anticipation of and fulfillment of demand changes constantly. Concurrently, strategies and pricing used to sell highly perishable rooms change to reflect changes in demand, changes in supply, and price elasticity.
Change in Available Supply is an Overlooked Metric
Curiously, change in available supply is an often-overlooked metric – particularly considering current labor shortages where post-COVID many hotels are struggling to attract and retain staff for essential duties such as housekeeping and maintenance. A hotel may have plenty of supply in their physical inventory but can't service rooms to meet demand. From a predictive analysis perspective, hotel labor shortages offer short term deliberate opportunities to push and pull price elasticity at the expense of the local competitor set. Who doesn't love a market dominating STR report week after week?
In these situations, inventory management does little to optimize revenues from the demand that the hotel can fill. While the hotel may adjust forecast on a monthly basis, using rolling 12-month forecasts, this does little to address the fact that sales and room revenue forecasts have been geared to meet predetermined targets and rates are adjusted to optimize this goal. The approach is reactive, rather than proactive, because the hotel is not able to respond quickly and decisively to supply and demand opportunities and challenges.
As a result, hotel sales managers often offer groups underpriced rooms with no options for ancillary revenues to make up revenue offsets, or the hotel offers deep discounts to Opaque sites and front desk staff miss perfect fill opportunities because the hotel sells its available inventory early or leaves occupancy on the table.
While revenue management systems such as Marriott's One Yield, IHG's Concerto, ChoiceEdge by Choice Hotels or Best Western's Best Rev use price-elasticity models to recommend optimal daily rates for statistically derived market segments, these systems are only as effective as the end-user. The principals of Garbage In, Garage Out apply to the five years of prepopulated historical data and segmentation that many of these systems use to calculate optimal pricing.
Frankly, COVID skewed data, limited real-time local information, conflicting trend predictions from 3rd party sources and acts of government require revenue managers to exercise a great deal of contextual and technical judgement in managing day-to-day pricing and inventory management.
Constraints such as labor shortages, supply chain delays, acts of government have made forecasting demand patterns complicated and unpredictable. Further, complicating accurate demand forecasts and pricing models is a lack of understanding who our guests now are. The front desk and hotel staff may see the guests daily, the GM may interact with guests during property walks and sales managers may be busy filling out group block bookings and conducting corporate cold calls.
None of this provides clarity as to whom our guests really are. A robust predictive analytics approach can help shed light on who the guests are, predict occupancy trends, improve top and bottom-line business results and help inform whether investing heart-shaped beds offers a competitive advantage and return on investment.
Revenue Management and Marketing Agree – Some Guests Are Worth More than Others
One of the many things that marketing, and revenue management share is the belief that some guests are worth more than others. The one-off guest that digitally checks in and out of their hotel room and never make use of the hotel's other facilities such as a swimming pool or exercise room has a certain value. A repeat guest who books multiple golf and dining packages when they check in with their extended family has an entirely different value.
Identifying guests with a higher Lifetime Value (LTV) to the property is the key to profitability. In order to maximize profit and bolster guest service scores, individual hotels, management companies and brands need to find creative ways to increase guest spending.
Whether increasing the guest spend is a result of satisfying basic expectations such as providing clean and comfortable rooms or by offering access to exclusive loyalty programs, every opportunity to encourage repeat business needs to be capitalized on. Creating demand has traditionally been the forte of the marketing department. And as marketers have access to potentially insightful customer data that revenue managers do not generally access, the opportunity to break silos and work together to gather, analyze and find insight in the structured, semi structured and unstructured data that resides in a variety of systems including the Property Management System (PMS); point of sale (POS), the customer relationship management system (CRM), channel management tools, competitive shopping software services, the hotel's own website and a variety of social media channels and developing a clear understanding of guest and market dynamics is the very heart of a predictive analytics process.
The Problem with So Much Data
As most existing hotels acquired different systems at different times from different vendors, transactional data is often difficult to extract, let alone load and consolidate into a format such as Structured Query Language that can provide meaningful insights. There is a no shortage of raw data to harvest – far from it – the challenge for revenue managers, marketers, sales, finance and a variety of other departments is to turn application, manual and 3rd party sources of data into cohesive guest profiles and insights. Data is insightful, but only if it can be analyzed and acted upon.
If a request for an extra pillow by one guest is a data point and the review where the guest mentions they didn't receive the extra pillow is a second data point, it's easy to visualize the volume of data that can be collected incrementally for each transaction a hotel has with a guest through the entire reservation through post-check out journey. Understanding the value that an individual guest or corporation brings to the hotel can be a challenge to quantify.
One of the strengths of a predictive analysis process is that automation can track individual guests across the entire portfolio of a management company's and brand's properties. This level of would be a giant step toward achieving guest-centric revenue management and consistent profitability.
Total Hotel Revenue Management and Predictive Analytics
Guest-centric revenue management or total hotel revenue management uses the guest's lifetime value to derive an individualized price for each guest based on the value that each guest brings to the brand. Many airlines, casino hotels, online retail platforms and brick and mortar retailers use versions of total revenue management to set individualized pricing that is designed specifically to entice the guest to book - based a deep understanding of a guest's price-sensitivity and product preferences.
Predictive analytics offers revenue managers, marketers, sales and front-line teams alike an opportunity to individually price and promote ancillary services. By knowing who will be visiting your hotel, the updated guest profile including interests, history and potential spend allows even limited-service hotels to promote and price room upsells and ancillary revenues such as golf tee times at local courses.
Of course, price-sensitivity and product preferences are not drivers of demand. In the absence of substitutes, preferences only indirectly influence guest buying behaviours. A guest may prefer a king room with ocean view but will happily accept a heart shaped bed and whirlpool tub if they need a room. Especially if the price is within their budget constraints. From a revenue management perspective, total hotel revenue management hinges on developing a relatively robust historical profile of a limited number of individually "known" guests. While there are definite benefits from a marketing and operational standpoint to nurture Long Term Value through promotions and loyalty programs, individualized demand generators don't fill consistently and profitably fill rooms.
Real-time Data Sources Provide Insight into Future and Current Demand
Neither individualized guest profiles nor an increase in the volume of data will automatically lead to better revenue management decisions but using new and real-time data sources to provide a fuller picture of future demand will lead to more revenue generating opportunities. The strength of predictive analytics is in its ability to collect readily available data on customer behaviour, buying patterns, and lost sales. Large data sets can be processed quickly through free data mining and analytic tools such as R and Python. Integrating readily available 3rd party data can provide actionable insights on local market demand and price elasticity constraints.
The most common visualizations such as text mining, heat maps, dendrograms and scatter plots are straightforward and most excel users who are comfortable with basic programming concepts such as variables, data types, functions, conditionals and loops will thrive in a predictive analytics environment. Problems such as defining the most profitable combination transient and group business on specific dates with the aim of maximizing total expected revenue through a mix of physical and transactional fences while managing hotel supply constraints become manageable and actionable.
There are a lot of techniques to forecast future demand and price sensitivity and hotels are guilty of using everything from naïve techniques of adding a few percentage points to the demand for the next year to using a gut feeling approach. Successful revenue managers have relied on excel spreadsheets and historical demand to develop revenue forecasts and inventory management plans, but post-COVID forecast accuracy, budgeting and operational planning have been dealt numerous curveballs including sporadic demand and acts of government.
The use of real time structured, semi structured and unstructured data collected and analyzed through a process of predictive analytics may allow revenue managers to increase accuracy of future demand forecast and price sensitivity. Additional benefits such as total hotel revenue management or profiling guests purchasing patterns will promote ancillary revenues.
In the meantime, the use of real time data from a variety of accessible sources will allow revenue managers to forecast both current and future demand and price sensitivity with greater accuracy.
