"Some people still think their own minds can do a better job than a RMS"
Published: 20 May 2008
Revenue and Pricing Strategies in Travel, TDS Europe 2008 Special
The travel industry acknowledges that the past few years have seen the introduction of new systems, both for revenue management and finance that assist hoteliers greatly in abilities to produce reliable and accurate forecasts.
But Steve Pinchuk, VP Profit Optimisation Systems, SAS Institue feels the current state of forecasting in revenue management systems has not fundamemtally improved for over a decade.
According to him, current RM systems still use single forecast models to forecast all demand types for all products for all arrival dates for all periods throughout the booking curve.
"This is intuitively weak. Different types of forecast models need to be used and the forecasts in RM systems should determine for each forecast, based on the historical data and past forecasts, what type of forecast model to use. An improvement in forecasting by 10% will add 2% more profits to the bottom line. Forecasting techniques need a radical improvement. There are dozens of different types of forecast models – why do RM systems still only use one type?" probed Pinchuk, while speaking to EyeforTravel.com's Ritesh Gupta. (A session on forecasting to optimise revenues in periods of high and low demand was conducted during Revenue and Pricing Strategies in Travel conference as part of EyeforTravel's Travel Distribution Summit Europe 2008 being held here in London).
Pinchuk also spoke about other aspects of forecasting. Excerpts:
Ritesh Gupta: An expert mentioned that even if you don't use an automated RMS with complicated algorithms, excel spreadsheets now look at segment or sub-segment level, while a couple of years ago the focus was on day by day occupancy forecast. What's your viewpoint regarding the same?
Steve Pinchuk: This means people who are building their own RM process in Excel are making their excel models more detailed. Excel could always do this it was the knowledge level of those using Excel that has improved.
Ritesh Gupta: The Revenue Management discipline is undeniably constrained in terms of segmentation and other labelling methods due to limitations in our human ability to process all of the information. What do you recommend in overcoming or how should one work to overcome such limitations?
Steve Pinchuk: Computer systems and their databases can handle this. People have been trying to mow their lawns with fingernail clippers. The problem is that some people still think their own minds can do a better job than a RMS. The issue is in volume, not ad hoc thought processes. The human brain is far more flexible and intuitive than a computer program – however the computer can perform 10 million forecasts in an hour, every hour for years. In order to analyse at the right level of detail you must let computers handle the huge volume of analysis that is needed.
Ritesh Gupta: Revenue Management systems are designed and built with two primary models: Forecasting and Optimisation. How sensitive is RM systems forecast quality about frequent price adoptions in a dynamic pricing environment?
Steve Pinchuk: RM systems used today do not model true price elasticity into their forecasts. They look at historical data for past price elasticity and assume that will apply for future forecasts. The past data is incomplete, they do not know the true value proposition and willingness to pay that led to the historical demand they are assuming will apply in the future. They also do not know these factors for the future dates they are forecasting. However, they are telling people their forecasts are price sensitive and people don't know enough to question this.
Ritesh Gupta: To ensure the forecast quality - and subsequently the restrictions and pricing decisions proposed by the RM System, hoteliers need to monitor and perform sanity checks on a very regular basis. Which in your opinion is the best way to ensure forecast quality is up to the mark?
Steve Pinchuk: Doing analysis of past forecasts and how accurate they were. A man who refuses from his past is doomed to repeat it…
Ritesh Gupta: Revenue Management requires both long-term strategic or pro-active processes and short-term tactical or reactive tasks. Do you think hoteliers are tend to be very good at the latter, but struggle to prioritise the time for long-term planning and strategies?
Steve Pinchuk: It is easier to react than to do long term analysis. Reacting is immediate and people justify that it must be done quickly so they rationalise away the need for true analysis.
Ritesh Gupta: A senior RM executive last year told me – consumer's reaction to price strategy changes can vary so much depending on the price sensitivity, demand situation and market segment that the only way to answer that question is to track consumer behaviour changes in relation to demand and include the result of this analysis in the next pricing discussion. What's your viewpoint regarding the same?
Steve Pinchuk: Someone knew what they were talking about. A markets reaction to price is the result of thousands of individual decisions or reactions. Sometimes it is not very logical and was not based on analysis or all the pertinent data. We need to try and understand behaviour and not decisions. If we can understand behaviour we can predict how people and markets will react even if they are irrational.
Comments
You must be logged in to post comments.
Pitkin said on 21 May 08:
I think I agree with Steve Pinchuk above. Maybe. Maybe not.
He says that the human mind is far more complex and intuitive than a Revenue Management System, but we must let the computers whip the large amounts of data into a recognizable shape. Yes, that's reasonable. But, it was a human mind that wrote the RMS software. Those of us who visit this site don't need to be reminded that software is notoriously buggy. Always has been and always will be. So, at some point, don't we still need the intuitive human mind to make that gut-level decision as to what to do with the recommendations and report from the RMS?