Need to Screen Your Customers for Fraud? Use Name Matching

Merchants Need to Combat Fraud and Bad Actors Aggressively

A minority of customers will attempt to cheat merchants in many different ways:

• Customers will violate provisions in their membership or passholder contracts to buy and resell discounted items.
• Fraudsters will use stolen credit cards to buy goods and then resell them directly or through unauthorized dealers.
• Customers will illegally sell partially used multi-day passes.
• Shoplifters will return items they have stolen for a refund.
• Customers will use a stolen credit card at e-commerce sites.
• Customers will contact a bank about a previous purchase from a merchant and attempt to reverse the credit card charge.
• Customers will refuse to pay and remain delinquent.

These are just a sample of the ways customers can cheat merchants. There is an infinity of other ways of doing it.

In order to counter these fraudulent activities, merchants typically maintain a blacklist consisting of data elements from previously established fraud cases, such as person name, address, phone number, and email address. This allows them to turn away repeat offenders by refusing their business. In the service industry (e.g., entertainment, hospitality), companies want to keep bad actors away before providing the service. Retailers want to screen customers in real time at the point of sale. In both of these cases all data provided by a customer is checked in real time against the blacklist. If a match is made, the service is refused or the purchase is declined.

Fraudsters Will Frequently Provide Different Biographical Information to Evade Detection

In e-commerce, would-be fraudsters will frequently enter data elements that differ from what they have entered previously.

• They will vary their names:
  • They will use nicknames: Joseph vs. Joe.
  • They will misspell their names: Smith vs. Smyth.
  • They will use their middle name instead of their first: John Michaelson vs. Fitzgerald Michaelson.
  • In the case of foreign names, a customer may vary the name in a way that is unique to the relevant language; for example, an Arabic surname may vary between alRasheed, arRasheed, or Rasheed.
• Addresses may be altered:
  • 1 Beaver Park Road vs. One Beaver Pk Rd.
• More complex cases occur where more than one person is involved in fraudulent activity, such as with gangs:
  • A person approaches the merchant whose name has not been previously blacklisted, but who shares a residential address with someone who is.

Name Matching Is a Technology That Can Help

What is needed is technology that can match on each data field and return a score of likelihood of a valid match:

• It needs to be highly accurate, minimizing both false positives (low precision) and false negatives (low recall) or allowing for a user preference to err on one or the other.
• For each data field, it needs to apply matching that is suitable to the characteristics of that field (e.g., person name, address) and needs to handle variations such as those mentioned above and many others.
• It also needs to be able to return a combined score for all fields.

How State-of-the-Art Name Matching Works

Unlike traditional approaches such as Soundex, metaphone,  edit distance, etc., state-of-the-art Name Matching uses AI and Machine Learning to train its matching models using a very large sample of real-world name variant data. It has different matching models for each entity type such as person, address, etc., thus achieving the highest possible accuracy for each type.

Name Matching is tunable. In the case of catching fraudsters, a high match on any one of the data elements – name, address, etc. – might be enough to reject the transaction. Of course, each company needs to be able to set its own business rules. A match involving a common person name (think “John Jones”) may not be enough for an automatic rejection. On the other hand, in the case of potential gang activity, a match on just the physical address would be a strong indicator of nefarious activity. Advanced Name Matching provides the tunability necessary to support this.

To provide more capability, advanced Name Matching also allows the setting of numerical weights on each field to specify precisely how important they are for overall matching. In addition, it provides a way to combine the field results to produce a combined score for an entire record. Finally, Name Matching must scale well to handle matching of large volumes of names.

Summary

In sum, Name Matching allows quick and accurate screening that will help merchants match effectively and in real time at the time of transaction to avoid costly fraud.