Forecasting with Confidence: The principles, process and unparalleled power of revenue science

Whether you sell hotel rooms or parking spaces, confidence in your forecast, and ultimately your pricing decisions, begins with a solid understanding of available demand in your market. This starts by applying a scientifically sound approach to assessing your unconstrained demand – demand that isn’t constrained by the capacity or restrictions of your business and could be sold if your property had an unlimited number of assets available. For example, picture a 100-room hotel or 100-space parking facility with demand for approximately 100 rooms or spaces on a particular date. Charge a fair rate, and you should be fine. Now think of the same property, but with a demand for 1,000 units. How you optimise your business between these two circumstances may require very different actions. Ultimately, in both situations, you may reach full occupancy – but the true demand is very different. In the latter case you can afford to yield more aggressively, charge higher rates, and still fill up your property. Any historical data you reference after the date in question has passed will be influenced by many factors, such as capacity and pricing decisions. It’s important to understand what the demand would have been, had it not been influenced by these constraining factors. This allows us to know the true demand, so we can really optimise profits by understanding the demand in the market – instead of looking at constrained demand. I am sometimes asked why we don’t use lost business data to determine unconstrained demand. Perhaps a few decades ago, you could’ve effectively kept track of regrets and denials, but it’s difficult to track this type of data cleanly and accurately in today’s complex online environment. This is especially so when you don’t know why a person isn’t willing to pay a particular rate. A regret no longer occurs simply because of price but because of a combination of many factors like location, reputation and online content. Using lost business data is flawed, adds a lot of bias and ultimately may not produce an accurate estimate of demand. The most advanced revenue management systems use scientific methods to understand true demand – this approach has consistently been proven to have the highest reliability due to the integrity of the data utilised. Demand has a lot of uncertainty, so it’s best to think of it as a signal which is a forecastable part to which ‘noise’ has been added, where ‘noise’ represents all the seemingly random or arbitrary fluctuations in demand. Noise cannot be forecasted, and it’s not possible to cleanly separate a signal from the noise using traditional forecasting methods. The job of forecasting is to isolate the signal and use it to produce a forecast. But what remains, after you have taken the signal out, is random noise. Therefore, you must understand that for whatever forecast you produce, you will be uncertain about it and must account for the random noise. To forecast accurately, your revenue management system must: Avoid inaccurate or ‘noisy’ data that dilutes the reliability of a demand forecast. Develop machine-learning, decision-based systems that adapt quickly to changes in your business or the market to continuously improve forecast performance. Harness the power of hundreds of analytics models, all of which must be finely tuned for specific business scenarios. Predict demand by incorporating historic and future data, competitor pricing, and forward-looking market demand intelligence. Consider demand and how it varies due to pace, season, day of week, year-over-year trends, shift, length of stay and asset type. Understand the unique relationships between properties, market segments, and their booking patterns while accounting for uncertainty or volatility in the market. Integrate pricing and market demand intelligence data directly into demand forecasts and strategic decisions to optimise revenue performance. At IDeaS, when we talk about forecasting we’re not just talking about forecasting an outcome but also about forecasting the reliability of that outcome. We use the forecast of demand, and the uncertainty, to cope with the risk associated with the uncertain nature of demand in the marketplace in our decision-making process, instead of ignoring it. Dr Ravi Mehrotra, President, Founder & Chief Scientist IDeaS Revenue Solutions