Global ecommerce volumes are expected to rise from €3.24 trillion today to €5.3 trillion in 2026, making up 27% of all retail sales. In conjunction, merchants are faced with a rising fraud problem, costing them over €80 B across key verticals in 2021.
With the increase in ecommerce usage, comes a plethora of data with 2.5 quintillion bytes of it being generated by customers daily. Despite this abundance, merchants continue to depend on traditional data sets that allow them to see only blocks of facts. Revealing connections within these large data sets can further pr ove to be a time consuming and costly exercise.
This calls for merchants to move away from such siloed data sets and leverage the power of data enrichment and linking analysis to better assess customer risk and spot and fight against emerging fraud types.
Value added by Data Enrichment
Data enrichment merges raw data points such as Bank Identification Number (BIN), IP address, email address, device data with a larger database that could be internally sourced, provided by a specialized third party or a mix of both. Each datapoint, when enriched, produces unique insights which are then fed into advanced Machine Learning models to assist in accurate risk scoring and fraud decisioning. Some of the insights gained include:
This allows merchants to gain key information required to form a richer picture of the customer without inducing additional friction in the customer journey. This is especially useful for those expanding into new markets or experimenting with newer sales and delivery channels.
Understanding how Linking Analysis works
In essence, linking analysis aids in connecting dots within a merchant’s data by evaluating certain attributes such as email, phone number, IP address, shipping and billing addresses, to link transactions together and form entities.
Each entity corresponding to a customer then includes aggregated information such as total amount spent, number of transactions, delivery address or methods of payment used in the past, which can be leveraged by risk scoring engines in real time. As opposed to traditional databases that store data in rows and columns, data here can be visualized and stored into graphs, making it more intuitive and easier to understand.
Let’s take an example and see it in practice:
- A new transaction, namely T7, worth 200 EUR is recorded. This transaction uses the same IP address as another transaction, T1.
- T1 is further connected to four other transactions i.e. T2, T3 and T4 based on the same name. T1-T4 are cumulatively worth 600 EUR.
- Two additional transactions, namely T5 (50 EUR) & T6 (70 EUR), trace back to T4 as they use the same shipping address. T5 & T6 are further connected to each other by using the same email address.
Effect of Linking Analysis: The linking logic allows for all the above transactions i.e. T1-T7 to be linked together and be viewed as a single entity, corresponding to a single customer. Transactions analyzed here not just correspond to a single merchant, but those engaged in by a customer across multiple websites serviced by the same payment service provider (PSP).
This helps to get complete details on the total number and value of all transactions in which the customer engages and gets a 360 degree view of their activity to accurately assess risk. Had the transactions been viewed individually or clubbed differently, it would paint an entirely different picture of them, which could lead to false positives and in turn harm the merchant’s top line revenue.
Benefits of using Linking analysis and data enrichment
Getting a complete and enhanced view on new and repeated customers, allows merchants to decrease overall false positives and reduce any uptick in fraud levels. Whitebox AI further aids in giving complete explainability over decisions achieved via linking analysis and data enrichment, further assisting in accurate rule writing and manually adjusting risk thresholds. This helps increase approval rates without increasing friction that may lead to cart abandonments.
In cases of suspected fraud, linking analysis and data enrichment can reveal hidden connections between fraudsters and form a complete profile of their transactional history. In turn, merchants can spend less time on manual screening and analysis. Sophisticated tools leveraging Machine Learning further allow for large chunks of data from multiple sources to be analyzed at a fast pace and in real time. This makes fraud decisioning extremely scalable and efficient for merchants.
Merchants adopting fraud prevention tools which leverage these technologies benefit from faster integration cycles and need lesser data touch-points in the customer’s purchasing journey. For example, data points extracted at checkout, namely name, sign up details, card details etc, would be enough to detect whether a transaction is at risk of being fraudulent, without the need for additional data to be collected across any previous steps in the customer journey.
Emerging fraud types uncovered by linking analysis and data enrichment
Synthetic Identity fraud
Synthetic identity fraud involves fictitious identities created by fraudsters by combining stolen and fake user and payment information such as name, social security numbers, shipping address, and other payment information. These fraudsters then mimic the behavior of a genuine user before ‘busting out’ and entering into a series of high value purchases.
Linking analysis can be extremely useful in listing all transactions associated with that IP address, email or shipping address. Any previously rejected transactions associated with the credentials would also be revealed to highlight increased risk levels.
Account Takeover (ATO) Fraud
ATO fraud when a fraudster gains control of an account that belongs to a genuine customer and engages in unauthorized purchases. The merchant is then left to face increased chargebacks from the original account holder and a loss of reputation.
In such cases richer insights on geolocation and IP gained via data enrichment aid in spotting anomalies or mismatches such as a stark contrast in current location and shipping address. Linking analysis can further identify other transactions linked to the name that may stem from a different IP or geolocation, hence pointing towards a mismatch and increased likelihood of fraud.
Fraudsters may set up multiple accounts to take advantage of one-time offers, signup bonuses and referrals.
Using linking analysis, these users can be identified based on their shared details such as card information, the same IP address or device, etc.
This article was contributed by Fraugster