Just like the engine is to a car – powering it and getting you where you need to be – your revenue management forecast powers your entire system and its outputs. Forecast accuracy translates into dollars and is crucial for your business success. Choosing an RMS with science-based forecasting capability helps you more effectively yield while preventing dilution during periods of high demand. It also prevents overpricing and guides you in developing targeted promotions during low demand. Hoteliers in turn can more effectively evaluate accuracy. And the more accurate your forecast, the better your revenue management system (RMS) recommendations, and therefore the greater your revenues will be.
Are you measuring forecast accuracy in ways that makes the most sense for your business?
Measuring Forecast Accuracy
Hotels cannot accurately calculate forecast error by simply averaging the daily margin of error. For instance, if a hotel is forecasted to fill 100 rooms for two days, filling 90 rooms on one day (-10 margin of error) and 110 rooms the next day ( +10 margin of error), the average net error becomes zero because the misses up and down net against each other. This makes it appear that you’ve forecasted accurately when in fact you have not. You can resolve the issue by working with absolute values instead.
Mean Absolute Error (MAE)
As the name suggests, mean absolute error (MAE) is simply the average of the errors after dropping the positive and negative signs. It’s calculated by averaging the absolute values of the difference between the forecasted values and the actual values.
While MAE works fine for a single hotel, a problem arises when benchmarking multiple hotels within a portfolio. MAE gives the illusion that your smallest hotels are more accurate, because the error numbers are lower. For example, a 1,000-room hotel compared to a 100-room hotel will have a bigger error, simply because you’re working with larger numbers.
Mean Absolute Percentage Error (MAPE)
To deal with this problem, we can find the mean absolute error percentage (MAPE).
MAPE allows you to compare forecasts across multiple properties because by looking at it on a percentage basis, you’re normalizing for the differences in the sizes of the hotels. A 10-room miss in a 100-room hotel will have the same implications from a forecast accuracy perspective as a 100-room miss in a 1,000-room hotel – both amount to a 10 percent error.
Key Factors to Consider
In addition to the measurements you use, there are some key factors to consider when calculating forecast accuracy.
Level of Granularity
With a monthly or quarterly roll-up of forecasted and actual revenue, you may have a lot of misses that ultimately net against each other. Hitting your monthly forecast spot-on might feel like a great achievement but it could mask some missed opportunities stemming from forecast misses at the day level. Perhaps you underforecasted a key Wednesday night, resulting in an early sell-out, a missed rate opportunity, and disruption to your weekly stay pattern. This miss may have been offset by an overforecast on a non-constrained day, where you may have missed the chance to promote more demand through an offer or package rate. Where the rubber meets the road, tactical revenue management decisions are made on a daily level, so it’s important to measure your forecast accuracy at a daily level of granularity as well to ensure you are achieving the full potential of your hotel.
Number of Days Out
Given that most good revenue management forecasts update daily, you also want to consider lead time. Forecasts by nature are typically less accurate further ahead of time. While you’d expect to see a larger error when measuring accuracy outside of your hotel’s booking horizon, this metric will likely not be very meaningful given that bookings are not yet materializing and that your hotel is making relatively few revenue management decisions. Conversely, measuring accuracy at a point where most of your bookings have already materialized is sure to give you a nice looking result but typically is too late in your horizon to be making decisions that will materially affect your performance.
Another factor to consider when gauging forecast accuracy is the volatility of your business. It’s similar to forecasting for the weather. It’s going to be much more difficult to forecast weather for a city that experiences big temperature swings versus a city that’s warm and sunny year-round. Therefore, it is important to manage your expectations regarding your level of accuracy accordingly, and perhaps use metrics that help benchmark against other forecasting methods to neutralize the effect of volatility on your accuracy metric.
Constrained vs. Unconstrained Demand
Revenue management decisions should be based on unconstrained demand forecasts –measuring your total opportunity – versus constrained demand. Anything that affects capacity can cause you to miss on a constrained demand forecast. Anticipating and managing capacity is certainly an important aspect of revenue management but should be looked at separately from demand. And using constrained demand to make decisions also negatively impacts your optimization results.
For example, say you have all 100 rooms booked at your 100-room hotel, and your constrained forecast showed 100 percent occupancy. You may think, “I made a high forecast and I’m spot on.” However, your unconstrained forecast would have revealed you actually had demand for 160 rooms. In this situation you may have under priced your hotel and missed out on lucrative revenue opportunities.
For optimal results, hoteliers need a forecast that goes beyond looking at historical data to consider factors such as lead time and granularity – measuring daily versus monthly. And you also need the flexibility to choose the right metrics in order to control for things such as hotel size and the volatility of your business. In the end, attaining a high level of forecast accuracy, and using unconstrained demand, are vital for your business success because they directly impact your revenue potential.