Machine Learning

The future of fighting fraud in the travel industry

An introduction by Sift Science and PhocusWire

A 20-something frequent globehopper arrives in Italy, pulls out her phone and searches for a place to stay.

Visiting her favorite booking site, she instantly sees suggestions for places within walking distance, tailored to her budget.

Even though she’s browsing from a new country and booking at the last minute – classic signals of fraud – she’s able to secure a deal for later that night and book it with a single click.

This isn’t a vision of the future of travel. The technology to enable easy, personalized, and safe bookings is already here.

As more and more travelers research, arrange, and buy all aspects of their trips using a digital device, the onus is on travel companies to innovatively deliver these standout experiences.

Trust versus Risk

If you’re a travel company that sells online, you’ve undoubtedly encountered the flip side of enabling convenient and engaging features like last-minute bookings and streamlined checkout. Fraud and abuse threaten to run rampant.

That’s why it’s imperative for businesses to know which users they can trust, and which they can’t. If you mistakenly allow a bad actor to buy something, create a fake account, or compromise a good user’s account, the repercussions to your business can be devastating.

But if you’re able to instantly recognize your trusted customers, you can deliver the best experience possible – ultimately deriving more revenue and strengthening long-term customer loyalty.

The bottom line: the travel industry is evolving, and market-leading companies must invest in digital trust to stay ahead, attract new customers, and retain existing customers.

And what is the technology that powers digital trust? Machine learning.

Fighting fraud in travel: Top challenges

Let’s look at some of the obstacles faced by modern travel companies that are dealing with fraud and abuse.

Rising competition, small margins

The overall volume is increasing across different travel products, but many companies are facing a small margin – particularly in the airline and OTA sectors.

High ticket value, with a low margin, creates a higher business risk. And “inventory” – which could have gone to a legitimate traveler – can’t be recovered once it’s lost to fraud.

Moving into new markets

The incredible rise of online travel has introduced great opportunities for businesses to expand into new markets. However, every new market has a different risk profile. Travel companies need to adapt their tools, as well as potentially hire new people with specialized knowledge of these markets.

Increasing product offerings

Travel companies are under pressure to grow and expand. Just look at Expedia’s 2015 acquisition of Homeaway, which many say is now helping Expedia close in on Airbnb in the home-sharing market. You may start out offering a single product, like hotels.

But then you expand into vacation rentals, insurance, activities, airport transportation, etc. Each product will also have a unique risk profile, which needs to be accounted for in any fraud prevention models.

Collecting information while keeping it simple

For many travel products – like airline tickets – you need to collect a lot of information. However, thanks to Amazon and other digital pioneers, customers are growing more accustomed to a streamlined buying experience. If they face too much friction, they’ll turn to one of your competitors instead. How do you keep the experience simple and intuitive, without increasing risk?

Monitoring mobile bookings

Some 40% of digital travel sales are expected to be completed on mobile in 2017, according to stats from eMarketer. And Google research found that 31% of leisure travelers and 53% of business travelers have booked travel on a smartphone.

Companies have to be prepared for both legitimate and fraudulent mobile bookings headed their way. When it comes to fighting fraud, mobile data is fairly unique – you can get extra data points like mobile carrier, device type, and the pressure someone uses to tap that help pinpoint fraud.

Last-minute travel is growing

The window of time between booking and traveling is shrinking for all travelers. One-third of millennials make travel plans at the last minute, and 72% of mobile hotel bookings are made within one day of a stay. This change in traveler behavior means real-time decisions are critical to your ability to stay competitive.

In many cases, there’s just no time to manually review risky orders and logins. Meanwhile, fraudsters take advantage of last-minute bookings to evade detection. Our research found that day-of hotel reservations are 4.3 times more likely to be fraudulent.

Flexible bookings / changing travel plans

Customer experience is high on travel companies’ list of priorities – and that means providing flexibility and convenience. However, fraudsters often take advantage of the ability to change their booking at the last minute.

For example, they may buy a ticket that appears low-risk ahead of time, then change it at the last minute. Not all travel companies rescreen for fraud when a booking changes.

Multiple types of fraud

Not only do travel companies deal with payment fraud in the form of stolen credit cards, but they also face fake accounts, content abuse (if they incorporate reviews and other user-generated content), and account takeover (loyalty fraud).

The teams responsible for managing these different types of abuse may be located in different departments, using disparate tools, which makes it challenging to take a unified approach to managing risk.

Why legacy fraud-prevention solutions come up short

Payment fraud and loyalty program fraud have been rampant in the travel industry, yet many companies are still relying on an outdated and reactive fraud prevention approach.

With higher-value purchases at stake, some companies have been hesitant to move on from legacy rules systems to layer on a machine learning approach.

The result? Higher rejection rates, an unnecessary and expensive reliance on two-factor authentication, and more customer insults. Fraud prevention directly impacts their top line.

Why are companies so hesitant to switch?

An organization may have already invested significant time and money into developing a system of weighted rules.

It could feel intimidating to consider giving that up for an unfamiliar approach that can feel like a “black box.”

And they may not want to face a sunk cost.

Here’s a quick table listing the pros and cons of each approach:

Machine learning: why it’s ideal for protecting and growing your travel business

While rules may still be used to prevent fraud today, machine learning gaining prominence as a more nimble approach advanced by forward-thinking travel companies.

The quantum leap in computing and big data power, as well as the increase in API-based machine learning solutions, mean that machine learning can now be leveraged by any company looking for a scalable way to grow without increasing risk.

Industry leaders are leveraging machine learning to increase their top line and expand into new markets while keeping fraud at bay.

Ability to positively alter the user experience

Machine learning systems are not just good for identifying and stopping bad behavior.

They can also be used to identify good users, so you can dynamically adjust their experience.

Real-time decisions

To stay ahead of fraudsters, travel companies need to gain actionable intelligence from all possible data inputs instantaneously, so you can act as quickly as possible.

Machine learning analyzes huge volumes and varieties of data to deliver real-time decisions, without sacrificing accuracy.


Machine learning models are capable of continuous learning, adapting to sudden changes in fraudsters’ strategies.

With a steady stream of quality data and feedback, a machine learning system’s predictions will get more and more accurate.


Data from LexisNexis reveals that merchants spend up to one-quarter of their fraud prevention budget on manual review.

Machine learning enables teams to automate tasks that don’t necessitate human review.

This provides an efficient, streamlined process that saves you money and valuable time.

Network effects

Machine learning systems that draw upon a global network of data allow that data to be shared – and increase your chance of identifying emerging threats before they even reach your site.

With Sift Science, our Live Machine Learning reevaluates your risk every time a user takes action on any site or app across our network.

High-volume data ingestion

Machine learning can take in a huge amount of data. The important thing to know is that so much of this is passive data based on user behavior.

How does a legitimate customer’s behavior differ from a fraudster’s?

For example, a traveler browsing a site may do multiple searches, spending time comparing a bunch of different options. They might return to the site many times, or forward a suggested itinerary to a friend.

The typical transaction time will be hours long. A fraudster is likely to spend less time, and complete many fewer of each of these actions.

At Sift Science, we have over 16,000 signals we look at.

Here are just a few examples:

  • Account age
  • Time until event
  • Seat selection
  • Order size
  • Destination
  • Buyer location
  • Device type / ID
  • Fare class

Some details about Sift Science and its Digital Trust Platform for travel brands

Machine learning is being widely used to prevent fraud and increase revenue, but not all machine learning systems are created equal.

The fastest-growing trust platform in the travel industry, Sift Science prevents fraud and increases revenue for top travel companies around the world.

We are the only company to use Live Machine Learning, automatically utilizing learnings from our global network of customers, as well as your own data, to provide a clear view of good and bad users.

This lets you reduce checkout friction and the need for costly authentication systems.

Sift Science is ideal for growing companies and enterprises, offering a holistic solution that can solve multiple fraud problems through a single integration.

Download the full ebook to learn more about top signals of travel fraud, issues around loyalty programs, and more!

For more information, visit or contact

More on Technology

The Latest

From Our Partners