Optimistic that you can easily find the best airfare and hotel rate for your next vacation or business trip? Think again.

Travel providers now use artificial intelligence software to re-price their offerings, sometimes dozens of times a day, to maximize revenue. For business and leisure travelers, the result is a variation of the cat-and-mouse game, where travel companies are almost always the cat.

Traditionally, hotels and airlines priced their offerings depending on peak demand periods, past sales data and the number of current reservations. Individual hotel properties could make changes if, for example, their hotel was emptier than usual for an upcoming date and a lower room price would spur demand.

Now, changes in travel pricing are being made much more frequently. The practice, called “hyperdynamic pricing,” is poised for significant growth, said Angela Zutavern, a managing director at the technology consulting firm AlixPartners and the author of “The Mathematical Corporation: Where Machine Intelligence and Human Ingenuity Achieve the Impossible.”

Hyperdynamic pricing  takes into account lots of data. Along with historical and seasonal information, the new AI systems scan the web for global news events, weather predictions, trending Google searches, social media posts, local event schedules and other factors that could affect demand, Zutavern said.

Did singer and rapper Lizzo just announce a concert tour? The system may anticipate a spike in demand at hotels in cities on her tour and raise rates there. Have hurricanes been in the news? Flight prices may need to come down to entice travelers to tropical locations.


“The systems give hotels and other travel companies the ability to make more frequent changes, experiment and then see the impact of the price changes,” Zutavern said.

According to research by the Montreal-based airfare prediction app Hopper, the average price of a domestic flight changes 17 times in just two days, while international flights change a dozen times in that span. Prices on high-traffic routes like New York to London can change up to 70 times over two days.

Theresa Van Greunen, assistant vice president of corporate communications of Aqua-Aston Hospitality, based in Hawaii, said her company manages Marriott and Hilton-affiliated hotels, which each have their own proprietary dynamic pricing systems. It also manages about 40 independent hotels that use Aqua-Aston’s pricing system. All employ artificial intelligence.

“AI gives us more insights into our data, and opportunities to be more nimble,” she said, not only in room pricing, but in suggesting to hotel management how to price special offers and whom to send those offers to, based on expected financial returns.

These kinds of insights are “especially important to smaller operators trying to compete with giant global brands,” she said.

More on The A.I. Age


Some travel companies are buying technology companies to jump-start new pricing capabilities. Last fall, Oyo Hotels and Homes, an Indian hospitality chain, bought Danamica, a Danish data and dynamic pricing company.


In response to the fluctuating prices, corporate travel departments that book business trips for their employees have begun to develop strategies to catch the lowest prices. Last year, Egencia, Expedia’s booking platform for corporate travel departments, introduced a feature that would scan the internet for prices on airline tickets that were already bought through its system. If a cheaper fare for the same flight and class of service was found within 7 days of purchase, the system would automatically cancel and rebook the ticket without any action required from the traveler or the traveler’s company.

Egencia is rolling out a similar feature for hotel rooms that scans for price drops and rebooks the room at the lower rate. The company charges clients a small fee when it finds and rebooks lower rates.

Egencia is also introducing Smart Mix, which analyzes business travelers’ data to determine their preferences — taking morning versus evening flights out of their home city, for example, or staying at certain hotel chains — and combining that information with their employers’ travel policies to suggest appropriate flights and lodging.

Individual consumers can’t compete with the resources and computing power of the travel companies, but there are a few ways they can improve their odds of getting a good fare or room rate.

Gary Leff, founder of the travel blog View From the Wing, advises travelers to avoid booking hotel room rates that cannot be refunded or changed, even if the price is slightly lower than a more flexible rate. He recommends instead that travelers periodically check for lower rates on the hotel website or other sites that the hotel might match. Members of some groups like AAA can often get a rate as low as the nonrefundable rates without the restrictions, he said.

In addition, many travel agents can offer discounted hotel rooms and airfares, passing along savings they get from working with consolidators who buy blocks of airline tickets and hotel rooms at discounted fees.

Travelers can also try to learn more about the hospitality ecosystem in places they like to visit, said Fred Lalonde, chief executive of Hopper. Large cities like New York, Boston and Chicago that have a high proportion of business travelers and a large hotel selection often have relatively stable pricing three to six months in advance, Lalonde said.


Prices begin to drop sometime in the past three months before check-in, so travelers can generally find the best deals then, he said, although they may not get the exact hotel or neighborhood they prefer.

There are also more companies that track pricing, predicting if a price will rise or fall and sending alerts. In August, Google started including “insights” in its Google Flights travel booking feature, displaying, for example, airfares that were lower than average in green and airfares higher than average in red, for some routes. In the fall, it briefly tested a “low price guarantee” service that promised anyone who booked a flight on Google that it would monitor the fare and pay the difference if the fare dropped before the flight took off. The company said in an email last week that it wasn’t sure if it would run that feature again.

Hopper said it analyzed 10 years of airline pricing, monitored hotel pricing around the world and tracked customer behavior to offer travelers advice on the timing of hotel room and airfare purchases and optimal routes. It also suggests alternatives to save money. The company said it had about 45 million customers in 120 countries.

Hopper also analyzes customer interactions within its system to make suggestions. If, for example, a large number of travelers who are monitoring the price of flights from New York to Rome are also monitoring the same dates for New York to Milan, Hopper’s AI recommendation system may suggest Milan as an alternative destination for all customers monitoring the Rome prices, even if they haven’t thought of flying to Milan.

“If enough people are doing it, the system sees it as a viable alternative that customers might want,” Lalonde said.


Hotels are more likely to display different prices for the same room on different websites than airlines do, he said, so Hopper monitors the web for lodging packages, flash sales, prices available only to customers in certain locations and other rates.

“We do all the foraging,” Lalonde said. The company receives commissions for the sales it makes.

The next wave of dynamic pricing systems will use “unstructured” data in addition to the “structured” data they already use, said Wilson Pang, chief technology officer of Appen, a company in Sydney, Australia, that supplies data to travel companies and others for their AI systems.

Structured data is clearly defined, like temperature predictions and historical sales. Unstructured data usually needs to be interpreted. The location of a photo taken from a hotel balcony posted on social media with the caption “Nice view!” needs to be identified and tagged. In a hotel review, the sentiment of the user needs to be discerned.

“These kinds of information could eventually be used in the pricing model too,” Pang said.