Let’s see first how impactful the problem of false positives and fraud is. Could you please describe the state of false positives, how much the rate has increased, and what ecommerce industries are most affected?
According to the Merchant Risk Council’s global fraud survey, the average merchant declines 2.6% of all orders due to suspected fraud. The higher the price, the more likely a transaction is to be declined. In 2019, Aite Group estimated that merchants lose up to 75 times more revenue to false declines than they do to legitimate fraud. This research was done before the pandemic, while over the last year the global online sales grew in size by more than 24%.
Many merchants tried to address this problem by investing in fraud tools but discovered a different problem: increasing false positives. For consumers, false declines are simultaneously insulting and a nuisance. One-third of new shoppers whose transactions were falsely declined will abandon the retailer and switch to a competitor. In fact, Aite Group has estimated that merchants could lose as much as USD 430 billion to false in 2021.
If there is little transaction data available, what is the most effective way to avoid false positives with unlabelled data?
The accuracy of fraud detection tools is dependent upon the quality of the data. The more limited the data for identifying fraud, the looser the thresholds for fraud must be to limit false positives. Fortunately, when it comes to electronic payments, it is possible to base fraud detection on complete, unaltered, end-to-end network data regardless of where a merchant, issuer, acquirer or processor sits in transaction.
When working with limited unstructured transaction data, unsupervised machine learning can be the most effective approach to detection false positives. With the data available to it, unsupervised machine learning compares new incoming data with past data to detect anomalies between common and uncommon behaviour.
Anomaly detection accuracy also depends on the population from which the model is generated and applied. When generated from a broad population (transactions from an entire channel, demographic, or region), the resulting model will be generic and legitimate transactions may be falsely declined.
A more accurate approach is to generate and maintain a unique machine learning model for each customer, card, and device. This removes the noise inherent in generic models, and results in fewer false declines.
How can AI tell the difference between a potential fraudster and a legitimate customer in real time?
By capturing payment transaction data off the network and processing it in memory, it is possible to decode, process, and analyse data using AI in real-time – before it even reached the authentication host. This enables fraudulent transactions to be blocked before they complete.
How can merchants reduce fraud attempts for transactions made via guest checkout, as the scenario is a bit more challenging than in the customer checkout case?
The use of guest checkout may limit the ability of the merchant to authenticate the identity of a customer before permitting a transaction depending upon how much friction the merchant is willing to add to the process.
Once a transaction from a guest account is attempted, the ability to detect fraud comes back to the quality and completeness of the payment data available to the fraud AI in real time. By analysing every field in every transaction anomalous or suspicious behaviour that would otherwise go unnoticed will be detected. For example, with the ability to capture and decode X-Forwarded-For information in real time, INETCO can determine the original IP address of the customer to establish if they are in a suspicious location or embargoed jurisdiction – even if the customer deliberately attempts to use intermediate servers to make the transaction.
What strategies does INETCO bring to the table to help out businesses in beefing up their fraud prevention measures while keeping a seamless digital experience on the consumers’ end?
For detection and blocking payment fraud, our company provides INETCO BullzAI – a single platform that combines a real-time fraud detection platform with a fraud transaction application firewall. The platform captures and decodes in real time all payment network traffic from Level 2 to Level 7 for the entire end-to-end transaction journey.
When a suspicious or fraudulent transaction is detected, INETCO BullzAI not only generates an alert, and opens a case in our fraud case management module and notifies analysts, but it can also decline individual transactions before they complete – independent of yet complementary to declines made by an authentication host.
As it monitors the entire end-to-end payment network, INETCO BullzAI can also detect and block or throttle network attacks such as DDoS, MitM, bot, and velocity-based attacks at the message level without any negative impact on legitimate transactions.
Our whitepaper ‘Preventing Fraud Earlier, Faster, and With Precision’, produced in collaboration with Aite-Novarica Group, further explores the traditional approaches of legacy fraud detection solutions and their inherent weaknesses to combat the rise and complexity in online fraud. So we recommend it to ecommerce merchants, processors, and financial institutions looking to reduce false positives and stop fraud.
What additional fraud challenges do you see as causing hurdles for merchants as we move into 2022?
When the world moved online for grocery shopping and other necessities, driven by COVID-19 restrictions, many merchants had to navigate a new territory, quickly setting up or improving their online stores. The challenge for merchants continues to be how to deal with surging customer demands, increased sales volumes, while still preventing losses from fraud and false positives.
The pandemic gave a huge boost to account takeover (ATO). Fraudsters have been doing ATO for years but were able to put a new spin on it with the rise in services such as Buy Now, Pay Later (BNPL). Also, recently, fraudsters started relying more and more on the creation of synthetic identities.
To stay profitable in 2022 while saving brand reputation means prioritising payment security and being prepared for new unseen fraud patterns.
This interview is part of The Fraud Prevention in Ecommerce Report 2021/2022, the ultimate source of knowledge that delves into the evolutionary trail of the payments fraud ecosystem, revealing the most effective security methods for businesses to win the battle against bad actors.
About Stephen Lazenby
About Inetco
INETCO Systems is a Vancouver, Canada-based global fintech company that helps financial institutions, payment providers, and retailers in over 35 countries detect and block payment fraud with granular precision, reducing false positives and fraud losses. INETCO solutions help clients improve payments security and customer engagement, while reducing operational costs.
Every day we send out a free e-mail with the most important headlines of the last 24 hours.
Subscribe now