Many hotel groups offer loyalty schemes, whereby members can gather points when they make repeated stays in a hotel or hotel group and redeem these points for free nights (or off-set a percentage of the cost). Hotels traditionally offer a fixed price for rooms when loyalty points are redeemed. With COVID-19 having an unprecedented impact on the industry, how can hotels provide incremental value to retain loyal members and drive demand, while maintaining a strict control on costs?
The mechanics of hotel loyalty programmes
Hotel loyalty programmes are primarily evaluated on their redemption options (that is, how many ways a member can redeem their loyalty points) and the point value members realise for those redemptions. In other words, what will my points get me?
With a fixed redemption price per hotel for the entire year (i.e. it always costs 20k points to book a room), the value of a loyalty point essentially fluctuates wildly between peak and off-peak seasons, since room rates are significantly higher during peak season and lower during off-peak.
Variation in point value results in an inconsistent member experience, and a difficulty in driving member value during low demand periods. With overall hotel demand falling to historical lows due to COVID-19, this problem becomes even more important to solve. Dynamically varying point value to keep in tune with hotel rates and overall demand can ensure better member value and a more consistent experience.
Reward night pricing has historically been a manual process, with flat prices set for the hotel for the entire year and prices being reviewed annually, primarily to control costs. With the hotel industry having been on a strong growth trajectory in the last decade until COVID-19, hotel occupancy and average room rates have also grown. This has been reflected in the growth of loyalty programmes, the popularity of reward night redemptions, and the cost of redemptions.
Hotel loyalty programmes have constantly raised the flat redemption prices to better control increasing demand and costs. This has resulted in inconsistency in member value between peak and off-peak seasons. With advancements in machine learning and pricing automation, loyalty programmes can develop sophisticated algorithms to automate pricing based on demand and cost forecasts, thereby improving member experience by providing consistent value.
Challenge: What are the best methods to optimise reward night pricing?
Our comprehensive analysis identified that combining customer transactional data and hotel performance data creates a stronger demand forecast model than traditional auto-regressive forecast models based solely on performance data. This in turn leads to better price optimisation. The main objective of the pricing engine was set to be the maximisation of member value and consistency while being cost neutral.
With this optimisation objective, 3 factors were identified to be key drivers –
Demand for reward nights is primarily based on hotel location and season. Ensuring affordable options for high-demand properties and seasons is critical for a good customer experience.
Redemption point value is dependent on the cash rate set by individual hotels. Point prices need to be in sync as cash rates fluctuate, in order to maintain consistent member value. Most hotel groups operate under a franchise model where hotels are required to set inventory aside for reward nights and, when those rooms are used for reward night redemptions, the franchisee is reimbursed by the hotel group. This reimbursement amount is variable and is typically dependent on the occupancy and average room rate for the day, making the cost unknown at the time of pricing a reward night.
Reimbursement rates have a high correlation with room rates but are not perfectly correlated. This creates unique opportunities where high point values can be provided to members at a low cost, as well as risks: costs could be too high for certain periods where reward nights are in high demand.
Approach: optimising prices with machine learning
We built various Machine Learning models to predict reward night demand and reimbursement costs by looking at historical customer transactional data, as well as performance data like bookings and cancellations.
Redemption point values were estimated for different price options based on prevailing room rates at the hotel. Final prices were selected using an optimisation framework that maximises member value while maintaining cost neutrality.
The use of customer transactional data vastly improved prediction accuracy, which in turn helped set more optimal prices. Additional business rules were applied to ensure consistent member experience across brands and markets globally. The optimisation framework and business rules also had configurable parameters that enabled tweaks to objectives based on the macro-economic environment. This enabled the pricing engine to be flexible enough to accommodate, for example, artificially suppressed demand due to COVID-19, and to automatically adjust pricing when demand returns.
Outcome: 22% increase in member value at no additional cost
The pricing engine was deployed globally, resulting in a 22% increase in member value for reward night redemptions, while maintaining reimbursement cost at prior levels.
This impact was primarily achieved by driving additional nights, through lowering reward night prices when demand was predicted to be low, and where reimbursement costs were also predicted to be low.
The number of members redeeming reward night programmes increased by 20% and the range of member point value decreased by 35%, leading to more consistent value for the hotel group and the loyalty programme members.