Why analytically forecasting for uncertainty impacts your profit potential
- There’s more to forecast performance than accuracy
- Forecasting for uncertainty is a critical component of revenue technology
- Accounting for noise in the analytical process minimizes risk & increases profits
Companies with data-driven business cultures are more productive and profitable than their competition. Hotel organizations, specifically, use data-driven approaches to increase profits amidst market complexity and volatility.
One component supporting a data-driven revenue strategy is an analytically-derived forecast. And with the increased profits it brings to the bottom line, hotels have a natural proclivity to validate their demand forecast through its accuracy.
However, there’s more to forecast performance than accuracy.
Take the game of heads or tails, for example. If you had to guess the probability of heads over tails, your answer would be easy: Heads over tails is expected half the time.
And 50%, well, that’s technically a perfect forecast.
But, as you can already imagine, a perfect forecast isn’t usually achieved with a limited number of actual tosses.
Play the game 10 times in a row—tallying up every time it lands on heads or tails—and you’ll more often than not end up with a percentage different from the original forecast.
Why is that?
Because a coin is a coin and a flip is a flip. Coins don’t have a memory; the coin doesn’t know or care what happened in the previous tosses.
Apply this to hotels: Compare 10 Tuesdays in a row by evaluating their 10-day forecast against their actuals.
You’ll find that no two Tuesdays are exactly alike, and each are also subject to the independent odds of their individual data elements.
When it comes to forecasting hotel demand, we can never truthfully say we know exactly what’s going to happen, rather what is most likely to happen.
So, then, if we look to demand forecasts to understand our most likely outcomes, what inputs help produce them?
While most hoteliers are familiar with forecasting elements such as historical data, recent trends and pacing, there are other considerations not as commonly discussed.
One of these is the role “uncertainty” plays in the analytical forecasting process.
Uncertainty refers to the amount of unpredictability or volatility that exists for any given day, in any given market, for any given market segment—in relation to the number of days to arrival.
It represents “the unknown,” and is best described as noise in the forecasting process. It’s also a large reason why your perfect forecast isn’t usually your actual forecast.
Unfortunately, demand uncertainty hasn’t received much attention as of late (and, in some instances, is ignored entirely by many revenue technologies). However, it plays a very critical role in your hotel’s forecast performance.
One of the reasons it’s impossible to generate a perfect demand forecast is because of this noise—seemingly random or arbitrary fluctuations in demand.
Noise is not forecastable.
The limitation of technologies using traditional demand forecasting is they use deterministic models that view all data as exact…including uncertainty.
They take exact values as input, and they output exact values. Because of this, their forecasting calculation is unaware of the uncertain nature of the demand.
With a traditional forecast, it’s impossible to cleanly separate signal (data with a predictive value) from noise (data devoid of predictability). Any deviation in demand—however normal it may be—is considered an error since the forecast is an exact number.
This noise also shows up as variability, and traditional forecasting models account for this in exactly the same way as the variability in the signal.
So how does IDeaS handle this process differently?
Unlike traditional methods, IDeaS separates the signal and the noise. Rather than attempt to forecast the noise—something futile in its efforts—IDeaS demand modeling instead improves the forecast by progressively isolating the signal from the noise.
In IDeaS demand modeling, everything is “stochastic.” Stochastic modeling systems train their sights on more accurate forecasts by modeling probabilities and factoring in random behavior.
This means a stochastic outcome can have any value within a range, and each value has a certain probability of occurring.
For example, rather than saying the predicted outcome of tossing a pair of dice is 7, the stochastic profile of the predicted outcome would be a range of outcomes from 2 to 12—with a probability of 1/36 for 2, 1/18 for 3…all the way up to 1/36 for 12.
This is important because if uncertainty is not managed correctly, a hotel’s demand forecast, its subsequent pricing decisions and overall profits are left exposed to higher volumes of risk.
Solutions that can’t properly understand or quantify uncertainty—and analytically consider them the same as they do the easily-distinguishable demand patterns—ultimately produce sub-optimal forecasts and decisions because of it.
By managing and accounting for uncertainty within the analytical process, these risks are minimized and hotels have better revenue opportunities, better restriction controls and higher hotel profits.
After all, the odds of achieving optimal forecast performance without accounting for uncertainty? No dice.