A Data Science Central Community
When you book 5 nights in a hotel, usually each night (for the same room) will have a different price. The price is based on a number of factors including:
I am wondering which other metrics are included in the pricing models. For instance, do they include weather forecasts, special events, price elasticity? Is the price user-specific (e.g. based on IP address) or based on how the purchase is made (online, over the phone, using an American Express card vs. a Visa card), how long in advance the purchase is made, etc.
For instance a purchase performed long in advance results in lower prices, but it is because inventory is high. So you might be fine by just using inventory available alone.
Finally, how often are these pricing models updated, and how is model performance assessed? Is the price changing in real time?
I'm interested in learning more about these algorithms. Anyone willing to provide insights?
I like this post. This brings back to my mind revenue management, hotel is a classic example of how to improve profits through RM.
Yes the rates can be assigned real-time - If a hotel has done their research, they can have many different predetermined rates. The rate determination maybe determined by asking if the customer is with a group. If customer answers yes then the appropriate rate from a "pool" of predetermined rates can be assigned real time. The rates can be determined at many different levels depending on who is booking for whom and when.
Below is some more ideas on maximizing profits through pricing.
There are many revenue options for rooms. There are primary sources (the room) and there are secondary sources (the food, beverage, activities, TV, etc). Below is a list of some of the different pricing options for a room with respect to the demographics:
Then there is the whole rate only determination. There is a "Rack Rate" or the published rate. This usually is the highest price for a hotel room. The rates decline from the rack rate. The above list determines a rack rate(different rates for demographics) and a discount is applied accordingly to the criteria below(demographics affiliated with a specific entity). Below are some of the predetermined rack rate discounts based on affiliation:
Of course all of the above requires a hotel to do a lot of market research and testing of prices with promotions/special offers. The performance metrics could be capacity at a given rate. If the capacity is 100% then more than likely the hotel needs to evaluate to be able to get higher prices. The other performance metric would be to evaluate the different affiliations, like the travel agencies bringing in enough business to justify the predetermined rate. The meaning of predetermined in this context is that there is some negotiation with the travel agency, company, and tour business...before the first booking. The individual bookings would depend largely on season, day of week, inventory, competitor pricing, location, elasticity, method of payment, and etc.
Thanks Lance for your great answer. Also, I'm wondering how "big data" and modern "data science" is being leveraged in these traditional OR pricing / inventory optimization models. Will we see prices offered in real time and customized to the user, taking into account highly granular prices from competitors?
I am not sure about the trends in the hotel industry. Here are my observations...
As an outsider observing the hotel industry, I see a lot of opportunity for big data doing competitive analytics and customer analytics and demand analytics. The smaller hotels would not be able to do much with analytics. The large hotel chains would have the data needed.
Also, it would appear through my traveling that we have real time customized price is already happening. As for taking into account the competitor prices, the competitor prices would only be a the rack rate level without discounts. From my observation, the discounts can be tied to capacity by increasing prices as capacity reaches 80% then 90% - the competitor would have no idea about the capacity at any given time.
I think there is a Big Market here, I live in Cancun, Mexico, we are talking about 400,000 Rooms, so I am sure BIG DATA and Business Inteligence could help a lot, please keep me informed about your research and your Ideas Vincent.
Here's an interesting article on this subject: http://www.cognisantassociates.co.uk/articles/Hotel_Rate_Setting.html
Great post Vincent!
I think travel portals are using big data and some sort of demand generation algorithm is in place for building these offerings. I agree that all the factors you have mentioned are used as key features for determining the price but online consumer behavior is also taken into account and thats where big data comes in. I have seen in past that a cheap motel on a weekday that costs $40/night, if you check out the prices for a week for that motel from different portals repetitively the prices start going up and can go as high as $100, and not only that hotels in same neighborhood see a lift in price as well. And pricing strategy at that point is increase the cost to double+x% so even if you are able to sell just one, you are covered for 2. I don't know how much control hotels have on their pricing except for defining the rack rate and price bands by they go up. I am no expert in this and this is purely empirical.
Large chains use yield management optimizations. They are large integer programs that decide at any time how many rooms in each price group to make available for the next time interval. Like the airline ticket pricing the routines are run often, I don't know how often-- airlines run them every few hours, I understand. They maximize revenue not profit. They take into account history of sales of each room type by day of week, events and promotions. There's a huge academic literature on 'revenue management' for many industries which you can find with google scholar. In the published papers are detailed descriptions of the algorithms and their dependencies. You can bet that the actual algorithms used, which are proprietary, are significantly more complex and use more parameters with data specific to the company and location.
Yield management is interesting because it is influence-free, in the sense that, if I am a frequent visitor to the hotel, I can't get a better deal by knowing the manager. The algorithm makes the rooms available, and they are there for anyone, like with airline tickets. The hotel does not care who it sells to. So if you are one click ahead of a high roller, you can score that suite with a view for $99. So the hotel, to keep its frequent guests happy, must now run lots of promotions for its 'loyalty' groups to keep them happy. What's the cost of that? I'm not sure that the yield management routines today accomplish a global optimization across both types of programs.
One example is Marriott's award winning "Group Price Optimizer" DSS for group customers. You can read about it here:
Please see the presentation and video, "Carlson Rezidor Hotel Group Maximizes Revenue through Improved Demand Management and Price Optimization" at
Another article that might shed a light :
I think the major part of hotel room brokers only take in acount competitor prices and inventory