By Max Wang, Senior Director of Data Science at CubeSmart
Artificial Intelligence (AI), fueled by bigdata, continues to make an impact on businesses. Revenue Management (RM) has benefited from this growing and disruptive trend. In this article, RM is concerned with “selling the right product, to the right people, at the right time, for the right price, to generate the maximum revenue.” Apparently, this is an aspirational goal! However, with the help of AI technology and bigdata, organizations are taking a quantum leap towards this goal. In this article, we discuss how businesses can leverage the power of bigdata and AI to achieve the RM goal.
Factors that affect demand include price, customer’s preferences, availability of substitutable products, income, and competition. To predict the demand for your product in the coming days, months, or years, collect all data that impacts demand. For example, airlines predict the demand for their flights using the past years of sales data, including prices, booking times, destinations, demographics of customers, flight type, and seating. Wherever possible, collect similar information on your competitors. You can augment your first-party data with third-party data (e.g., income, demographics, and lifestyle). Even if your data is scarce, you can still leverage machine learning (ML) for demand prediction. There might be additional technical challenges, such as segmentation, data aggregation, censoring, seasonality, and trend detection. However, ML and AI technology are well developed to solve these problems.
Supply forecasting is essential to RM. For this discussion, we focus on a product’s availability (not general supply chain forecasting). There are a few ML techniques available for supply forecasting, such as exponential and adaptive smoothing, moving average, regression and ensemble modeling, and life cycle modeling.
There are additional factors that affect supply for some other industries. For example, the tenants of a self-storage operator may move out without giving advance notice. They need to predict the vacate rate of current customers, which is a function of various factors including but not limited to, length of stay, location, product features, time of year, day of the week, price, and tenant’s information and their relationship with the company. ML and AI technology, fueled by diverse data, powers the prediction of supply for the future.
At the core of RM is price optimization, which consists of four key components: demand, supply, price elasticity of demand, and a performance metric (objective function). There are technical challenges in estimating these components. For example, you may not know the exact demand when rooms are sold out for hoteliers as the unknown additional customers cannot be served. The objective function is specific to each industry, e.g., an airline maximizes the revenue of each flight, and hotel operators may maximize the annual revenue of each hotel night or stay. ML with AI technology is an ideal approach to solving these challenges. It handles bigdata well, it learns and improves itself, it minimizes human bias, and generates truly optimal prices.
At CubeSmart, the Revenue Management team has developed a price optimization system with personalized product recommendations. This system leverages first and third-party data, ML, and AI technology to make rate recommendations for thousands of products and hundreds of new customers daily. An integration of this system with an ML-based marketing budget allocation process has further transformed CubeSmart’s marketing, RM, and CRM decision making processes.
Promotion Management, Marketing, and Operations
What else can an organization do with RM? Let us look at the hotel industry as an example. At each location, there is a fixed number of rooms. The revenue manager has important decisions to make to determine what price to charge for each available room, whether to offer a promotion, and whether to run additional customer acquisition campaigns. ML can assist in making smarter decisions. Data Insights lead to better demand forecasting of price, by product feature, seasonality, and location. They can also help with enhanced marketing decisions. For instance, there might be no need for additional location campaigns if your organic media can attract sufficient customers for that location at that time. Alternatively, you may still run a paid program to attract more or higher value customers if your pricing advantage can overwhelm the cost incurred by the program.
ML and AI technology help you achieve the RM goal through improved Operations programs. To name a few, these include product recommendation, guided sales funnel (e.g., customized landing pages), real-time chatbot, upselling, cross-selling, overbooking (e.g., airlines), upgrading, and product bundling.
To ensure success in leveraging the power of bigdata and AI in RM, there is imperative to be both strategic and tactical support at the executive level. Strategically, there is a need for vision and collaboration. Tactically, AI-powered RM involves data collection, processing and storage, computational resources, automation, process change, A/B testing, and BI technology for monitoring and reporting.
The support starts with the collaboration of executives to define a strategic goal that aligns with the RM objective. For one company, the goal could be setting the optimal price to maximize the annual revenue and, for another, it could be to determine an optimal product mix to maximize profitability. An effective RM system requires multiple years of related data for development and deployment. The system is computationally expensive to run, maintain, and improve. This requires strong support and commitment from all CXOs.
To create an AI-powered RM, your organization starts with forming a team of data scientists and engineers. It provides them the support for continual business knowledge development and technical enhancement. It is critical to the success of any the AI-powered RM project that executives lead by example in making data-driven decisions, cultivates an analytical culture, acts as change agents, and that they are willing to take risks with a series of data-backed experiments.