Fraud prevention in our way

Our solution enables huge variety of businesses, including e-commerce companies, classifieds, airlines, hotels and many others to conduct business online. The inevitable feature of doing business online, which frequently frustrates merchants and their clients is fraud. Another disappointing part here is that internet businesses are liable for fraudulent purchases according to the card association rules – while usually their toolkit for fighting fraud is quite limited. Using our expertise as a payment solution provider, we are there for you to cover your back, while you as a merchant can focus on product and customer experiences and not on fraud, so we’ve developed a suite of modern tools for fraud detection andprevention.

During the last couple of decades, very many fraudsters have moved online, with internet becoming a goldmine for criminals, whose level of sophistication is constantly evolving. Significant amounts of money and constantly increasing volumes of information are changing hands over the internet, and skilled cybercriminals have all the necessary tools to gain illicit access to them. Fraud is a complex feature of selling online, requiring us to constantly reevaluate and adjust our strategies.

The purpose of this guide is to provide more details on the features of anti-fraud solution that we have designed and how we use it with specific attention to such features as machine learning, device fingerprinting, online persona detection and data-driven analytics we use to adapt to the specifics of our merchants and their individual fraud exposure.

Payment fraud

The most typical example of fraud online is a fraudster obtaining someone else’s credit card details and using it for making purchases – obviously unauthorized. For example, the criminal buys a high value item—say a jewelry—from an internet business for $500 and then resell it offline or using digital marketplace with a significant price downturn - for $100, still making profit. The legitimate owner of the card will find the unauthorized transaction and initiate a chargeback with his or her bank – long and costly process of charging money back to his account. Usually all the costs of chargeback are ultimately born by a merchant.

Those costs are represented not only with the value of the good/service sold but also any additional fees levied for the dispute, especially significant for merchant with a low average order value – chargeback costs may multiply losses. When a certain business becomes targeted by cybercriminals, those costs will become a significant financial burden.

Online commerce has been steadily growing over the past years. Although selling online provides merchants with significant gains and consumers with the added comfort of making purchases in a real-time fashion, the losses from fraud are also increasing rapidly and sometimes even faster than volume of online sales.

There is another side of fraud impacting the financial position of the merchants. While situation with false negatives described above – fraud thatwas not identified and mitigated before transaction occurred, there is another side – false positives. Those are transactions of legitimate users that are suspended by various fraud prevention tools —are also costly to merchant in terms of revenue lost.

The reality of trading online that you always need to choose balance between those two – the fewer false positives you have the more false negatives will occur and vice versa. It is important to note that businesses are not always fully capable to control this equation since there are regulations of card associations which place limits on allowed number of chargebacks and in case of breaching those regulations sanctions and costs will follow. Consequences of non-compliance may be severe in some cases not only in monetary terms of fines and penalties, card associations might also disqualify merchant and such a situation will effectively result into even more severe losses up to bankruptcy.

Solid solution

Bearing in mind the experiences of merchants and our expertise as a payment solution provider, we have developed our own multilayer anti-fraud solution, geo-targeted to African specifics and adaptive to features of our merchants from different industries.The algorithms of our solution track and evaluate every transaction for fraud probability evaluation based on number of criteria and allows range of action to be taken. While obvious fraud cases are blocked by default, we still allow our customers certain degree of customization – their experience and knowledge of the business are those key details of building individualized fraud prevention system that we cannot exclude.

Our antifraud is inbuilt directly into the payment flow and works out of the box – no set-up fees or other associated costs will be incurred. Once you’re integrated with us, our antifraud solution is also ready to launch. And due to frequent changes in fraud patterns the business will also need to adjust and invest in anti-fraud significantly, while we can do it for you – our system is constantly updated, we have access to the blacklists and suspicious cardholders from all over the world based on our cooperation with merchants from different industries and with different specifics of fraud that we have already mitigated.

Solid’s solution uses predictive models based on both historical and real-time data received from our customers world-wide, allowing us to show you a more detailed picture of what typical fraud signals might look like and thus to help you to reach correct decision on whether to block each specific transaction or not. Such an approach to fraud detection provides merchants with connectivity and deep correlation between merchant specific data. Another significant advantage of our solution is the self-learning ability of the system - we can always develop new rules and triggers around changing circumstances.

Self-learning system or in other terms, machine learning has been a recent trend in fraud prevention and we have fully utilized it rather than using a traditional rule-based system. The algorithms which underpin our solution make predictions in real-time about the integrity of a transaction based on information from legitimate and fraudulent transactions. Machine learning is usually discussed in context of fraud prevention, but it can also play a role as a transaction approval tool – and we offer such an opportunity to our merchants. Using evaluation of patterns in good transactions, we can make definite distinctions between normal and fraudulent transactions and separate them into two distinct groups, that will either flow differently or the latter ones will be blocked.

We have integrated machine learning into automated fraud screening layer and merchants can now use it for identifying high-risk transactions, profiles and accounts, detecting and analyzing changes in user behavior and combatting account takeover – quite a few uses apart from traditional transaction screening. Although it’s at the forefront of our innovation today, the methodologies that underpin machine learning are certainly not exclusive to fraud detection. In fact, frequently referenced examples of how algorithms influence our daily lives include movies and box sets recommended to us by the likes of Netflix, and Spotify’s music suggestions made based on our listening habits and taste.

Online persona fingerprinting

Each user that enters merchant’s website has number of features that by default may be used in analyzing the nature of transaction and those include not only obvious points such as card details, geographical locations, velocity but also the previous uses of the card, untypical transaction amount combined with velocity (like 5 payments per minute for unusually high amounts), browsing history, cookies, use of anonymizers and many others that can hint us on potentially fraudulent payments.

Behavioral models and online history of potential buyer on merchant’s website are giving us all types of actionable insights about visitors. Having access to a range of data including email addresses, phone numbers, unique device fingerprints together with behavioral data, such as where and how often users are clicking and how much time they are spending on different pages. Our solution can quickly and efficiently process all of this information and find patterns, helping to have much clearer picture of users’ intents.

Application of such technics allows us not only to reduce fraud, but also to go a step further and design a variable payment flows to reduce a friction for good users. For example, traditional offline financial institution may ask every applicant, whether trustworthy or not, to fill out the same form with a predetermined number of questions, an online merchant, leveraging device fingerprinting and behavioral analytics can confirm trusted users and send them through a frictionless payment flow while requiring extra verification from new or suspicious users. Optimized payment flow will make a use of merchant’s website much more enjoyable for trustworthy users.

Human input into learning the machine

Together with fully automated anti-fraud with inbuilt machine learning algorithms, Solid also allows merchants to add customized rules based on the business specific – for example to block or manually review payments from certain countries/devices/anonymizers and so on.

Setting such a rule may be viewed as a manual help to the automated algorithms – while it takes some time to collect data and learn, the ready-for-use information derived from merchant’s knowledge of own business will help to adapt faster. After you set up a new rule with our antifraud solution you will also get a full set of analytics showing the number of matching transactions that were actually disputed, refunded, marked as fraudulent to evaluate the performance of the trigger set.

When it comes to your attention that machine learning skips certain fraud triggers, that are obvious to you and can easily isolate fraudulent transactions – it would be a great addition boosting the performance of customized anti-fraud for your business and this is exactly what we mean by individualization available for every merchant.

We are also providing you with a toolkit for manual review for specifically flagged transactions – sometimes merchants might choose to invest some additional human work just to be on the safe side in their fraud prevention logics, and we have foreseen that as well – functionality for more traditional rule-based approach is also included in merchant’s dashboard.

Data at the forefront

Access to a wealth of data is the crucial resource in building of fraud preventing measures, however it still can be a challenge to turn that data into actionable and self-learning system. Tuning and customization of anti-fraud toolkit in line with constant shift in trends and customer behaviors requires non-stop analysis of available data. Here at Solid we believe this process is most effective when it is driven by the dedicated fraud experts – those who appreciate the unique nature of fraud challenges attached to each and every merchant and who can offer you the most applicable, efficient rules and strategies in order to reach out your clients with a frictionless experience and help you to grow.

Fraud rules, predictive models and application of machine learning should be built on an exquisite understanding of the information available and strategy is a subject to continuous adaption to various emerging trends, objectives of a merchant and market developments (global and local). Our interpretation of data with the analytical expertise available can also deliver value to other parts of merchant’s business and not only in terms of preventing fraud. Geographic expansion, introduction of new products and payment types have their own risks attached, often requiring adjustments to fraud prevention strategy. It is this level of tailored, expert support which can make our fraud prevention toolkit to work for you at maximum capacity.

Fraud rules, predictive models and application of machine learning should be built on an exquisite understanding of the information available and strategy is a subject to continuous adaption to various emerging trends, objectives of a merchant and market developments (global and local). Our interpretation of data with the analytical expertise available can also deliver value to other parts of merchant’s business and not only in terms of preventing fraud. Geographic expansion, introduction of new products and payment types have their own risks attached, often requiring adjustments to fraud prevention strategy. It is this level of tailored, expert support which can make our fraud prevention toolkit to work for you at maximum capacity.

By using proactive screening and analytics tools combined with reasonably developed system of alerts, knowledgeable analysts can review transactions, monitor the efficiency of rules set, forecast and identify emerging trends in real time, spotting and stopping fraudulent activity before it reaches its target.

Last, but not least

The reading of this guide should have helped you to understand the basics of fraud prevention algorithms that we can offer. If you have any questions or feedback regarding our fraud prevention toolkit or are interested in working with our anti-fraud solution, please feel free to contact us.