Executive Letter

Can you imagine an airline or a travel agent telling a good customer that they can’t book their ticket? Or walking, cash in hand, into a hotel with vacancy and being turned away for no apparent reason? Smart businesses don’t alienate paying customers. But that’s exactly what happens to e-commerce customers every day. Travel merchants – due to their direct outreach, expensive advertising or carefully cultivated brand reputation – bring customers with intent to purchase to their site and then turn them away. They’re rejected due to fear of fraud, and they’re probably never coming back.

E-commerce fraud costs the travel industry $21 billion per year. Billion. With a “B.” It’s an enormous problem, and there’s a good chance you don’t realize it. $6 billion of that goes to the fraudsters. They aren’t going away. Worse, though, is the other $15 billion that the industry loses to costs associated with managing fraud. Good customers are turned away, fraud review teams are increased or reduced in response to seasonal demand and prospective customers shop elsewhere because of reputational damage.

But fraud prevention in travel is particularly challenging. Smart fraud prevention looks at as many data points as possible and makes a calculated decision. Travel merchants often can’t do that. Tight timeframes, no shipping address, international shopping, bookings for business and more all make it more difficult to determine a good order from attempted fraud. It’s tempting to choose caution, batten down the hatches and hope for the best, but that’s the wrong choice. Smart fraud management looks at both sides of that $21B coin. Chargebacks are damaging, but the only way to really address the problem is to reduce the associated costs.

Doing this well has enormous payoffs. Rather than shutting the door on a potential revenue stream, a smart approach to fraud management grows with travel merchants. It opens the door to new customers from more markets. It minimizes annoying verification methods and provides a seamless experience for your customers. It manages the costs of chargebacks but not to the point of turning away good customers.

At Riskified, we’ve been helping merchants manage e-commerce fraud since 2013. We’ve processed hundreds of millions of transactions and handled all types of threats. Given that experience, my biggest piece of advice for travel merchants is to view this as an opportunity. Your current approach to fraud management is almost certainly leaving money on the table by rejecting good customers. You spent the time and money getting them to your site; you’re already doing the hard part! By optimizing your approach you’ll capture that revenue without increasing costs.

And that’s why we partnered with Skift to develop this report. We’re excited to help travel merchants understand the true costs of online fraud in travel and how to best manage it. Smart merchants will increase their revenue, cut costs, and open their doors to more good customers. And isn’t that why we’re here in the first place?

Eyal Raab, Vice President of Business Development


activity is booming in the travel industry. Digital travel purchases have long been a critical source of revenue for nearly every sector of the industry, from airlines to online travel agencies (OTAs) and hotels. But despite digital’s long-standing importance to travel, the rapid growth in online purchases is nonetheless surprising. Consider, for example, that worldwide digital travel sales will surpass $676 billion in 2018, growing at a rate of more than 10 percent a year. And although this growth rate will slow during the coming years, global e-commerce travel sales, including airline, car rental, and lodging and transportation, are forecast to climb to $855 billion by 2021.

The shift toward digital travel purchases is just one sign of e-commerce’s growing importance to the industry. Another important signal is the increasing popularity of mobile purchases, whose contribution to overall U.S. digital travel sales jumped by about four percent last year to nearly $76 billion, with mobile transactions expected to surpass $100 billion by 2020. In fact, mobile accounted for 40 percent of all U.S. digital travel sales in 2017, with the channel forecast to surpass desktop sales by 2022. Even sectors such as aviation, where mobile ticket purchases were once almost unthinkable, are seeing increasing growth. One recent study by Atmosphere Research forecast sales of airline products via the mobile channel will grow from 1.7 percent in 2016 to more than 7 percent by 2021.

But in the midst of these optimistic e-commerce forecasts, a problem threatens to destroy the industry’s hard-won digital gains. In fact, a growing blind spot is preventing travel companies from realizing the full potential of consumers’ increasing demand for online travel purchases: e-commerce fraud.

Some travel companies may consider fraud to be a secondary concern: something they have already dealt with and have under control. But in reality the problem threatens to roll back the progress made by today’s digital-centric travel businesses. Consider an analysis of e-commerce fraud trends between 2016 and 2020 conducted by Juniper Research, which found that the number of fraud attempts based on total worldwide population increased from 1.39 to 1.49 percent between 2014 and 2015, a year-over-year increase of 6.7 percent. The issue is all-the-more pressing because travel businesses are among the largest sectors targeted for fraudulent activity. One example is the airline industry, which in a 2016 analysis had estimated fraud costs of $858 million per year, with roughly 75 percent of those costs borne by airlines. The problem is even bigger for travel intermediaries like OTAs. One 2017 analysis by eNett predicted a 24 percent increase in worldwide OTA fraud costs, from $8.8 billion in 2017 to $10.9 billion by 2020

A 2017 analysis by eNett estimated that OTAs will incur fraud costs of close to $10.9 billion by 2020, including both “direct” losses as well as indirect” costs due to higher operating expenses for manual review and lost business to competitors.

Although the challenge of detecting and preventing this fraud is urgent enough, the strategy currently used by travel businesses to guard against fraudulent transactions makes the problem worse. The rigid,
overly stringent fraud safeguards currently employed by travel businesses to avoid illegal transactions are creating a growing number of “false declines,” legitimate purchases wrongly rejected due to suspected fraud. These inefficient fraud-detection systems end up creating a problem costlier than fraud itself. According to Business Insider, U.S. e-commerce merchants lost more than $8.6 billion to unnecessary false declines in 2016, a dollar figure that’s $2 billion more than the genuine fraud such efforts were designed to prevent. As if this lost revenue wasn’t enough, inefficient fraud-prevention systems slow down transaction processing times and increase operational costs for staff that must manually review uncertain purchases. On top of all this, an inefficient fraud process can create frustrated customers, squander marketing efforts to acquire new buyers, and even help drive those customers into the arms of competitors.

In response to these growing challenges, an increasing number of travel companies are taking a more sophisticated approach to fraud management and detection, shifting their strategy from one that is reactive, slow, and inflexible to one that proactively analyzes threats and adapts in real-time to new changes in the fraud landscape. And thanks to recent advances in data science, and the growing field of machine learning, more companies in the travel space can realize these benefits at scale and at a lower cost than was ever possible in the past.

The benefits of doing so are numerous. Aside from the obvious boost to the bottom line, many organizations realize a lift in customer satisfaction, experience quicker online transaction time, and reduced drain on internal company resources normally needed to monitor and fix transactions flagged as fraud. In fact, organizations that implement these next-generation fraud-prevention systems estimate that they can recover between 40 and 70 percent of all digital purchases previously declined while driving an incremental one to 1.5 percent increase in e-commerce sales.


Nearly all travel businesses that sell their products online have a system in place to prevent fraud. But too many of these systems miss the mark, creating overly rigid and imprecise fraud rejection rules that slow down transaction time, increase costs for manual review, and lead to lost revenue.


Why are the current tools used by travel organizations to manage e-commerce fraud no longer serving them? And what is it about the structure of the travel industry and its customers’ digital purchases that make companies more susceptible to problems with inefficient fraud management? Most importantly, what are the benefits of utilizing a next-generation fraud-prevention solution, and how much do merchants stand to gain as a result? In “Solving Travel’s Revenue-Limiting E-Commerce Fraud Problem,” Skift and Riskified will attempt to answer these questions, providing a roadmap for a more effective, profitable, and successful e-commerce fraud strategy for travel organizations in the years to come.

E-commerce fraud 101: what is it?

What is e-commerce fraud, and why should travel executives care about it? While the problem of fraud is widely recognized, the specific tools and techniques used to address it are not as well understood. The following is a primer on the most common terms associated with fraud prevention in e-commerce and digital travel, along with some of the inherent flaws and challenges with each approach.

  • E-commerce fraud: Online purchases made in bad faith by criminals, who use techniques like stolen or fake credit cards to complete transactions on merchant websites. Because these orders are placed online, they are, by definition, “card not present” orders (see below). These illegal purchases can be quite costly to companies selling online, as the fraud liability for card-not-present purchases is entirely borne by the merchant.
  • False decline: A legitimate transaction that was improperly flagged as fraud. False declines result in a merchant losing out on legitimate business that would have otherwise been approved but was rejected due to overly stringent fraud prevention systems. The travel industry is especially prone to false declines due to its reliance on rigid rules-based systems and the global nature of many purchases (for example, a traveler buying something outside their home country).
  • Chargeback: A purchase that a cardholder has reported to his or her bank as illegitimate. The money is refunded to the cardholder, and the merchant loses the costs associated with the goods or service and its fulfillment. On top of this, chargebacks carry additional fees for merchants that can become quite costly.
  • Fraud-prevention systems: Nearly all businesses that sell online have a system in place to prevent fraud, though some are more effective than others. Some of the most common techniques include manual review (having a human review the transaction), “issuer” review (having the customer’s bank review the purchase) or “rules-based” systems (using software rules to determine if a purchase matches fraud patterns). Although each of these fraud prevention systems can reduce bad transactions, they all have drawbacks and inevitably reject some transactions that looked suspicious but were in fact legitimate (false declines).
  • Card not present: Often called “CNP” for short, this term refers to any transaction (online, by phone) where the card is not physically present at the time of payment. While the majority of CNP purchases are legitimate, the remote nature of e-commerce activity makes CNP purchases an attractive target for online fraudsters. Airline companies are among the most affected by CNP fraud, with some estimates placing total annual losses for the sector at $1.2 billion.
  • Address verification system (AVS): A fraud-prevention technique designed to verify CNP transactions. AVS cross references the billing address provided by the customer with the billing address on file at the customer’s bank. While the system can be helpful in some instances, it is often less effective than it should be. Customers are prone to make mistakes when entering their address data, and AVS will penalize them for this mistake. Additionally, many fraudsters buy stolen credit card information with all the AVS details included, so creating a match is easier than it should be.
  • 3D Secure (3DS): A type of “issuer-based” identity-verification tool used by merchants to provide additional layers of protection for online credit card transactions. 3DS cross-references customer data provided by the merchant with data provided by the customer’s bank. While 3DS does have its benefits, many retailers believe it adds unnecessary friction to the checkout process.

Outdated fraud tools cost travel businesses big money

The costs associated with inefficient fraud management may seem small. But, for travel businesses already struggling with slim profit margins and increasingly fickle digital customers, over time these inefficiencies add up.


By now, e-commerce fraud is a well-known problem for many travel executives. But it’s not just the fraud itself that creates challenges. The reality is that inefficient management of e-commerce fraud detection can have bigger long-term consequences. When travel executives employ inefficient techniques, or rely only on existing fraud prevention tools, it leads to unnecessary lost revenue, increased operational costs, and poor customer experiences. In this section we’ll take a closer look at the unforeseen business problems associated with a reliance on these out-of-date fraud-management solutions.

Lost revenue

Chargebacks are the most-obvious source of lost revenue created by inefficient fraud management. As noted above, a chargeback is a purchase a consumer has reported to his or her bank as illegitimate.The travel companies lose the cost of the goods or service sold, and additional fees are imposed on the merchant by banks. These costs are well known by most e-commerce merchants in the travel space, and they continue to grow. According to JP Morgan Chase, chargebacks are growing annually by 20 percent, while overall online sales are growing at just seven percent.

However, an even more significant problem than chargebacks is the lost revenue caused by the improper rejection of valid transactions. As noted above, a recent study by eNett found that the costs of improper fraud management for OTAs are expected to increase by 24 percent between 2017 and 2020. Other research conducted by Riskified backs this up, with the company estimating that between 40 and 70 percent of orders that merchants typically decline are actually legitimate and could have been approved. These false declines, or mistaken purchase rejections, turn away legitimate customers for no good reason, resulting in the short term in lost revenue and in the longer term, potentially missed sales.

As those who specialize in e-commerce fraud confirm, overzealous fraud prevention methods can lead to customers who cannot complete legitimate purchases. “Merchants need to understand that the fraud cost is not only about chargebacks; it is also about customers that are not able to finish the purchase due to being rejected by the risk controls,” said Eric Olmos, revenue manager at NUK Consultants. But figuring out how to minimize such declines can be difficult. “Distinguishing between fraud, friendly fraud, and non-fraud-related chargebacks, and finding the proper balance between lost margins and chargeback cost, is key to creating and maintaining a risk-control system… Finding the right balance could boost sales five to 10 percent.”

Many travel merchants may believe their current systems for preventing chargebacks and false declines are sufficient to handle the problem. But as observers familiar with e-commerce fraud will point out, many of the most widely-used verification systems aren’t effective enough, leading to unexpected lost revenue and missed sales.

Consider the 3DS system as one example. This customer verification solution allows merchants to essentially outsource fraud prevention efforts to financial institutions and card issuers. While this may seem like an added layer of security that can help reduce fraud-prevention costs, these institutions often suffer from their own inefficiencies, slow processing times, inertia and incomplete customer data. Further, 3DS is not universally supported by banks, and it can only be applied when every link in the purchase experience — bank merchant, card issuer and billing infrastructure — adheres to the protocol. These limitations vary by card issuer, country, and retailer. On top of this, even the banks themselves don’t believe 3DS is an effective fraud deterrent. One 2018 study by the Minneapolis Federal Reserve reported that, “about three out of 10 large institutions ($1 billion or more in assets) use 3D Secure or its equivalent for online payments; however, none of them rate this method as very effective and two-thirds of those institutions rated it somewhat ineffective.”

The lost revenue implications of a less-than-ideal solution like 3DS are real and growing. According to Riskified data, about 30 percent of OTA transactions are routed through 3D Secure, and those instances experience drop-off rates (where a customer abandons a purchase) of as high as 60 percent. That means if an airline or OTA sends 30 percent of their orders to 3D Secure and experiences a drop-off rate of 30 percent, the net effect is a direct revenue loss of 9 percent. Travel industry observers confirm that 3D Secure is a flawed solution that can lead to more missed sales. “3D Secure has long carried concerns about the impact on sales conversions,” said Christophe Kato, head of industry card services at the International Air Transport Association (IATA).

Higher internal costs

In addition to the growing costs associated with lost revenue, inefficient fraud management also leads to higher operational costs for travel businesses. In fact, the costs and extra employee time associated with fraud management can compound quickly, as businesses grow and unpredictable consumer demand causes unexpected spikes in manual review.

Manual review, or the process of verifying the legitimacy of a transaction with customer support teams, is common in e-commerce. The vast majority of merchants, roughly 80 percent, employ manual review teams to make a final determination on suspect transactions. These call centers are among the most expensive, and time consuming, solutions for fraud prevention today because they require frequent maintenance, ongoing payroll costs, and experience high rates of employee turnover.

Yet in the absence of more sophisticated fraud-prevention solutions, the reliance on manual review seems to only be increasing. According to a study by LexisNexis Risk Solutions, nearly half of all transactions flagged as potentially fraudulent are ultimately decided by human beings. Overall, merchants are spending up to 25 percent of their fraud-mitigation budgets on manual reviews. The amount of employee time spent on these inspections varies greatly, but the average number of orders a reviewer processes in a day ranges from six for small merchants to more than 100 for large merchants.

E-commerce analysts increasingly agree that relying on this process is not sustainable in the long term. “[Businesses] must move away from tools with high false-positive rates to next-generation solutions that provide much more effective fraud containment, require less maintenance and counter fast-changing fraud patterns,” wrote Andras Cser, vice president at Forrester. Because fraudsters are organized and well informed of merchants’ fraud detection rules, merchants need to seek out solutions that offer versatile and customizable architecture to cut operational costs and improve performance.

Poor customer experience

Lost revenue and increased operational costs are both problematic effects of inefficient fraud management. But travel is also an industry that is highly reliant on delivering positive customer service. That’s why another problem – harder to quantify but no less significant – is how fraud rejections impact the overall buyer experience. Too often with overly strict fraud management systems, good customers are wrongly turned away due to false declines or long purchase authorization wait times, causing them to give up or choose to complete their purchase using a competitor website. This problem is further compounded by the abundant purchase options in the online travel world, an ecosystem in which customers have little patience bad or slow experiences.

Research by Skift confirms that current fraud detection safeguards negatively impact the customer experience. According to 2018 consumer fraud survey conducted by Skift, many customers who were incorrectly declined either gave up on their purchase, called for help, or went to a competitor website. Among respondents who said they previously had a purchase declined, more than 43 percent of the respondents said their next action was to “give up,” while more than 38 percent said they contacted customer service, resulting in increased operational costs. Meanwhile, more than 21 percent went to a competitor website, resulting in a lost sale. The long-term implications of losing a customer due to a poor experience aren’t limited to a single purchase alone: when considered through the lens of customer lifetime value, a lost sale can be even more painful if the competitor successfully converts the buyer into a loyal customer.

Inefficient fraud detection can result in customers who give up on purchases, tax customer service resources, or create situations where the buyer defects to competitor websites to complete their purchase.

Each of these cause-and-effect situations, whether that involves a customer giving up on a purchase, contacting customer service, or going to a competitor, is a problem on its own; no one likes unhappy customers or lost sales. But over time, each instance contributes to the overall brand experience, collectively hurting customer sentiment and leading to negative impacts on profitability and loyalty. When sales are routinely lost because an outdated solution falsely rejected the purchases, it can quickly grow into a much larger problem with long-term business consequences.

Why fighting fraud is so hard for the travel industry

Key dynamics in how digital shoppers buy travel products online complicate travel executives’ ability to identify and prevent fraud using existing methods. The industry’s high degree of price transparency, time-sensitive product inventory and international nature can often thwart efforts to fight fraud using existing methods.

All industries face challenges and hidden costs related to inefficient fraud detection and management. But travel organizations face a unique set of challenges. Perhaps the biggest challenge is the industry’s price-driven, hyper-competitive, e-commerce ecosystem, where an unending push for low prices creates tight profit margins. Thanks to the category’s numerous OTAs and metasearch websites, digital travel e-commerce has a high degree of price transparency, meaning consumers are likely to shop around for the best possible deal. In such an environment, there’s very little to stop a customer who was falsely declined on one e-commerce site from taking their business to a competitor.

There’s also decreasing brand loyalty among travel consumers, which compounds the false decline problem. According to a survey of businesses and agencies by Econsultancy, 46 percent of respondents said it is very easy for customers to switch to competitors in the travel industry. Losing these increasingly fickle customers to a competitor due to a declined purchase can occur quickly. Yet that seemingly brief interaction can have significant long-term consequences, cutting into the lifetime value of customers. The willingness to switch between travel brands makes unnecessary fraud declines all the more painful.

This in turn adds to the industry’s high customer acquisition costs to bring in new digital buyers. A number of recent studies confirm that customer acquisition can be a large expense for many travel brands. One 2015 examination by Market Realist found that close to 50% of the annual expenses of high-profile travel firms like TripAdvisor went to “selling and marketing,” while another analysis of the hotel sector found that between 15 and 35 percent of guest-paid revenue was spent on acquisition efforts.

On top of this, the typical buying habits of travel customers don’t always make it past existing fraud-detection rules, placing added pressure on existing fraud-detection solutions to keep up while allowing unnecessary chargebacks and false declines. “Online travel exposure to chargebacks is unique for several reasons,” said Olmos. “The pressure on immediately delivering the service, the trend to collect as little information from the customer as possible to make funnels frictionless, and the fact that no physical good is delivered make the risk assessment even more difficult. Most of the classic data points that an online travel company could use to evaluate risk are easy to fake. That, combined with the huge volume of sales, makes it easier for the fraudster to pass under the radar.”

In the section below we’ll examine some of the other unique complications facing the travel industry related to fraud management, including the timing of travel purchases, the geographic distribution of buyers, and travel shoppers’ growing comfort with mobile purchases.

Purchase timing

The real-time nature of most travel pricing, which requires companies like airlines and OTAs to approve and confirm digital orders rapidly or risk losing the sale, is one big challenge for today’s travel companies looking to get a handle on fraud. Real-time purchases lead to tight time windows that are often too narrow for any hiccups to occur during the path to purchase.

One reason for the growing importance of real-time purchases is the increasing popularity of low-cost airlines, which have helped to popularize shorter booking windows with consumers. While this change is good news for buyers, it also puts more strain on merchant systems and manual support teams that must be constantly available to deal with problems. The need for fast approvals or denials has compounded under these time constraints.

The global nature of the travel environment also means that today’s e-commerce-focused travel companies are open for business seven days a week, 24 hours a day, with merchants accepting orders from all corners of the globe. Even the most well-staffed and trained businesses will find it extremely difficult and inefficient to rely on manual fraud review 24 hours a day.

A further timing-related challenge for merchants is the complicated purchase habits of today’s digital travel shoppers. Consider the typical behavior of “backpackers,” a consumer segment frequently-associated with travelers between the ages of 18-25 years old. The age of the travelers and the short time window between their purchase and their flight are often seen as potential “red flags” for a travel organization looking for fraud. Yet as data from companies like Riskified reveal, these spur-of-the-moment purchases are often safer than you might expect. An analysis of merchant transactions processed through the Riskified system found that just 2.9% or less of these supposedly high-risk transactions made four days or less before a flight were actually fraudulent.

The fast-paced, time-sensitive nature of travel purchases mean that many organizations struggle to accurately assess the fraud threat. One example is the challenge presented by the purchase habits of last-minute younger travelers. While the group’s purchases may seem risky at first glance, the use of more sophisticated fraud detection analysis often proves this not to be the case.

Using an overly aggressive, rigid, or simplistic fraud-prevention system to process these fast-paced transactions not only disregards typical traveler behavior, it also risks alienating a key segment of young consumers who may avoid buying from a travel company for years to come due to a bad purchase experience.

Geographic considerations

Another example of the difficulties inherent in travel-related fraud detection has to do with the geographic location of buyers. This often causes problems when customers are asked to input their physical address, a data source that doesn’t always make sense. “In travel and transportation, the address information makes sense only regarding invoice purposes as there is nothing to be shipped,” said Yago Casanovas, Payments and Fraud at Air Europa. “Most customers do not need an invoice so even if the data has to be input, many customers do not do it correctly and they can give erroneous information like the country if it is displayed in a combo box. We see many invoices issued to Afghanistan, being the first in alphabetical order. Fraud screening tools do not like countries like that so they reject.” But many customers choose Afghanistan out of convenience, and these customers are actually from geographies where fraudulent purchases are much less likely. As a result, these otherwise good orders were lost unnecessarily.

Another recent example from Air Europa shows the strain a false decline created by geography can place on travel businesses, regardless of the ultimate outcome. The airline recently reported a problem when a potential customer in Madrid attempted to purchase a one-way flight from Caracas, Venezuela to Las Palmas in Spain’s Canary Islands. Because of the unique flight path, the airline initially flagged the transaction as fraud. “The customer rang our call center a bit angry,” Casanovas said. Unfortunately, the fraud-prevention service that Air Europa was using didn’t account for the large community of Canary Island expats now living in Venezuela. The airline’s customer service agents became the last line of defense and they were able to approve the transaction with confidence because of their knowledge of the market. Although the issue was resolved to the satisfaction of the customer, the airline absorbed higher costs during that process.

The rise of mobile

Yet another unique challenge for the travel industry is the growing popularity and utility of mobile devices for booking.  As noted earlier in this report, more travelers are booking travel products like hotels or rental cars on the go and at the last minute, accounting for a greater share of the travel industry’s overall digital purchase activity.

Yet despite the continued growth of mobile purchases, many e-commerce organizations continue to rely on the misguided perception that mobile purchases are more likely to be fraudulent. The widely-held belief is that mobile devices are often easier to break into than a traditional desktop computer, especially when they don’t have a lockscreen code. There’s also the added complication that mobile devices aren’t connected to a static IP address, making identity verification more difficult. On top of all this, many organizations worry that mobile devices can be more easily used to fraudulently complete purchases. It’s one reason why a recent study by LexisNexis found that the cost per dollar impact of fraud in the mobile channel in 2016 had grown to $2.33, a 12% increase over the year prior. But is this concern about mobile devices truly warranted? Or is another example of how overzealous enforcement and rigid fraud systems can contribute to lost revenue?

Skift’s own research confirms that mobile devices are generating a significant share of consumers’ purchase decline problems. Among the people who had an online travel purchase declined in Skift’s consumer survey, smartphones tied with desktops as the most frequently used device, with more than 37 percent of respondents confirming that they were using a smartphone when their purchase was declined. This statistic is all the more notable when considering that mobile purchases currently account for 40 percent of digital travel purchases, suggesting the mobile purchases tend to be declined at a higher rate than those made via desktop PCs.  

According to Skift Research, travel purchases made on mobile devices were declined for fraud at the same rate as purchases made on desktops, this despite the fact mobile purchases represent less than half of travel merchants’ e-commerce activity. The discrepancy suggests a more sophisticated fraud approach could help to reduce false declines on mobile.

But despite the growing use of mobile, Riskified research suggests that some mobile transactions are actually substantially less risky than purchases made by desktop.
Consider a review of the fraud rates on last minute online flight ticket orders between the two devices, which found that fraud rates for mobile were significantly lower for mobile devices.

The challenge presented by the industry’s higher rate of mobile false declines threatens to become all the more pressing as mobile accounts for a greater share of overall digital travel purchases. As noted earlier in this report, travel purchases made on mobile devices are expected to surpass those made on desktop devices in 2022, making it all the more urgent that travel providers move quickly to streamline their mobile approval process.

The results above suggest that travel companies need to take a closer look at how and why they reject mobile transactions. But perhaps even more importantly, travel organizations’ need to take a different approach, utilizing fraud-management systems that are flexible enough to adapt to the behavior of the mcommerce environment and to identify and interpret the unique purchase details that come from such on-the-go orders. In other words, detecting, reviewing and processing these sales requires a fraud-management system that can offer higher efficiency and better real-time adaptation.

Fighting back: machine learning and the future of the industry

Digital travel platforms will always grapple with fraud. But there can be a dramatic reduction in the number of false declines when travel companies implement better fraud-management solutions. In fact, today’s best-in-class e-commerce fraud-management tools offer travel organizations a more flexible, efficient, scalable, and holistic strategy. Taking this approach is the only effective way to make sense of the increasingly complex range of channels, transactions, geographies and business models that make up today’s digital travel ecosystem.

As we’ve explained, there are plenty of existing fraud solutions on the market. But while each has their uses, many are unnecessarily inefficient and cost businesses even more than the fraud they were designed prevent. In this challenging environment, machine learning has emerged as a promising solution. This modern approach is not only more flexible, but also contextual, scalable, and able to learn from past fraud patterns, constantly adapting a company’s fraud safeguards based on the organization’s evolving customer base, industry sector, and historic purchase habits.

Machine learning for fighting fraud

The value of modern approaches to fraud prevention compared to outdated methods used by travel companies is significant, and nowhere is this more obvious than in the realm of machine learning. This branch of artificial intelligence is built on the idea that machines can analyze huge data sets like e-commerce purchases, learning to identify and adapt to common purchase patterns with minimal human interference. Best of all, machine learning and its respective results get smarter as more information comes into the fold. When more purchase transactions are safely approved or denied, machine learning can incorporate those additional pieces of data into the decision-making approach, better predicting patterns and improving the accuracy of future decisions. By applying this machine learning approach to e-commerce fraud detection, businesses can boost transaction approval rates and reduce chargebacks with a new level of confidence.

Why is machine learning so useful to today’s cutting-edge digital businesses? Consider the example of Netflix and Spotify, two consumer companies that use the technique to help make their huge media libraries more personalized for a previously unimaginable audience of hundreds of millions of customers around the world. How does Netflix know so well what you want to watch next? How can Spotify build surprisingly accurate music playlists tailored to individual tastes? Netflix and Spotify aren’t the first entertainment services to provide their customers with recommendations, but they can attribute much of their success to the power of machine learning and its ability to identify personalized tastes at massive scale.  What’s innovative about their approach is that these two companies are able to use machine learning to identify consumption habits for millions of users in real-time, while also having the flexibility to evolve their recommendations based on feedback and as users’ tastes inevitably change.

It’s that same foundation — the power of machine learning — that makes today’s fraud-detection solutions scalable and adaptable. Instead of telling the system how to determine recommendations, travel businesses can feed historic data to algorithms powered by machine learning and empower the system to use patterns to learn how to achieve better and more efficient outcomes. After gathering massive amounts of this data on legitimate and fraudulent transactions in travel, machine learning can be applied to find which data points (or combinations of data points) are most important and how they should be weighed. Thousands of data points can be considered and weighed to determine the legitimacy of every travel booking.

These solutions also minimize the need for traditional fraud-prevention approaches like manual review, rules-driven software, or those based on broad limits provided by card issuers or financial institutions. Rules-driven approaches often result in situations where the most basic of factors can render a transaction fraudulent when it should otherwise have been approved. Consider the example of an airline that decides to establish a rule against purchases from a specific country that is more susceptible to fraud: if every transaction that originates out of the country is automatically denied, a huge number of legitimate customers might be turned away. In contrast, a well-designed machine learning models can comprehend the full context of a booking. Instead of weighing each data point separately like a rules-based system, these cutting-edge systems identifying various elements that are statistically positive or negative to determine the final decision.

Empowering machine-driven algorithms to oversee fraud detection has three crucial benefits: scalability, speed and constant recalibration. Scalability is key for any rapidly growing company. Speed is particularly important in many travel purchases, and computers can make smarter, more contextually aware, decisions in milliseconds. And machine learning can apply “deep learning,” a more sophisticated form of problem solving inspired by the human brain, to progressively improve and gain proficiencies by iterating and learning from past mistakes.


Identifying common patterns

Human brains excel at detecting patterns in large groups of information. Today, sophisticated machine learning is helping companies in a variety of industries apply this same pattern recognition ability to a variety of business problems, doing so at scale for huge data sets where humans are not as efficient.

Machine learning is also more effective at understanding some of the most-common causes for e-commerce false declines in the travel industry. Many of the common situations that cause false declines, such as mismatches in customers’ shipping and billing details, or geographic discrepancies between credit card bank identification number (BIN), address, and IP address, are common and shouldn’t be used to automatically decline orders. Travel firms often have trouble addressing these types of data mismatches using existing techniques. Thankfully, machine learning can improve on this approach by connecting the dots between specific data points to more accurately predict if a purchase is fraudulent.

Machine-learning algorithms will consider a wide range of unique data points, including a passenger’s date of birth, the legs or flight routes of a trip, ticket type (round trip or one way), the number of passengers on an itinerary, age of travelers, and the nature of last-minute purchases to prevent fraud without increasing costs or adding complexity to a travel business’ operations. Addressing market-specific purchasing behavior in different locations is also paramount because it’s a common pattern that requires the flexibility, breadth and understanding that only machine learning can provide.

How might this work in practice? Consider one example from a Scandinavian travel company. Prior to partnering with Riskified, this travel merchant was being overly conservative in its approval rates for orders that originated outside of the region. Almost every order originating from Scandinavia was being processed without issue, but approval rates were much lower for other geographies. It’s not that the Scandinavian buyers were inherently safer. The problem was that this merchant’s system didn’t know how to evaluate their purchases. This meant that the company was leaving significant amounts of money on the table because purchases from outside of Scandinavia violated their rules. While some of those transactions carried higher risk, many did not and should have been approved. Riskified’s solution verified the legitimacy of those transactions and turned rejections into approvals.


Machine learning and your bottom line

Machine learning does appear to simplify and streamline the inefficient processes long associated with fraud prevention. But many travel executives rightly wonder about costs and benefits of implementing such systems on their bottom line. As it turns out, the benefits of an effective fraud-prevention strategy are numerous. Lower operational costs, an easier path to scaling the business, and unlocking greater revenue (and profits) are among the most immediate returns, but other gains can be achieved as well.

As some e-commerce observers increasingly recognize, secure transactions and fast, simple, transactions can exist side by side. “There is the age old conundrum that security and fraud prevention do not sit well with a fast and smooth process,” said IATA’s Kato. “It is true that fraud checks and post-transaction reviews hamper a fluid process,” but the level of sophistication enabled by new tools, devices, and techniques have created stronger and more efficient possibilities to overcome these challenges.

What kind of benefits do travel organizations realize from deploying machine learning in their fraud prevention systems? Consider the example of a real midsize OTA, an organization that generates somewhere in the neighborhood of $200 to $300 million a year in revenue and previously rejected between $40 to 80 million of their transaction volume as fraudulent. With the help of a new machine learning-driven fraud system, the company experienced an immediate recovery of $10-$15 million in valid purchases that were previously rejected as fraud. The company also believes a further boost in recovered transaction revenue of approximately $35 million, previously lost to false declines, is possible.

Another large European-based OTA with more than $2 billion in annual sales used a new machine-learning-based fraud system to boost its transaction approval rate by three percent, bringing its total approval rate to 97 percent. The OTA was also able to achieve higher-than-expected approval rates on its most profitable products, such as travel packages and hotel reservations, while eliminating many of the historic chargebacks associated with airfare. Most companies that adopt such systems also experience significant operational savings, reducing the need for manual review of potentially fraudulent transactions and recognizing that their current process for fraud prevention simply isn’t scalable in the long term.

Yet another example comes from a smaller online travel agency, which was able to successfully scale its e-commerce operations after switching from manual and rules-based processes to a new fraud-detection system driven by machine learning. The OTA was looking to expand abroad to serve new customer segments while also growing its product inventory to offer flights from low-cost airlines. But the drive to scale also created growing pains and higher costs. Due to the company’s manual-review process, it was difficult to meet the purchase-timing requirements to approve their new low-cost flight transactions. Meanwhile, the introduction of new fraud filters led to an increase in false declines and a 30 percent reduction in sales. It was only after switching to an automated, machine-learning fraud system that these problems were alleviated, with the company witnessing an immediate reduction in chargebacks, better approval rates on high-risk customers, and a speedy approval timeframe for fraud reviews. All of this has helped the company experience 30 percent year-over-year growth since the switch.



has never been more vital to the success of the travel industry. But as much as this ever-expanding flow of digital commerce has been good news for travel organizations, it has also had unintended side effects, including a growing problem with e-commerce fraud. Fraud is simply a fact of life when it comes to digital commerce, and many travel organizations have put in place a variety of safeguards to help guard against the threat.

But too often, these systems cause more problems than than they solve, increasing the number of falsely declined purchases, creating new costs for manual review of declined transactions, and leading to dissatisfied customers that may never come back. Unfortunately, in the highly competitive travel industry, most companies do not have the luxury of losing these sales. Margins are tight, and customer acquisition costs are high, meaning that in order to survive and thrive, all digital-focused travel businesses need to do more to get a handle on their fraud-prevention efforts.

To accomplish this, tomorrow’s most successful travel e-commerce companies need to rethink the way they approach the long-ignored discipline of fraud management. Instead of relying solely on simplistic and costly solutions like rules-based filtering, manual review, or issuer-based fraud safeguards, today’s best-in-class travel organizations are increasingly utilizing fraud techniques based on cutting edge tools like machine learning.

Not only are these systems easier to scale to millions of customers, they also adapt and grow over time, keeping pace with ever-changing travel purchase habits and new types of illicit transactions invented by criminals to get around existing safeguards. The benefits of adopting this new fraud detection approach are numerous, including increased peace of mind for travel executives and customers, increased cost savings and faster transaction times.

Fraud may be a fact of life for today’s travel businesses, but poorly-handled fraud prevention strategies shouldn’t make the problem worse. It’s time for the travel industry, a sector long considered to be the pioneer of the digital-first business model, to bring their fraud detection and prevention efforts into the 21st century.

Download PDF Version