4 Things Machine Learning Schemes Miss in Online Credit Card Processing (Thus the Need for Manual Review)

4 Things Machine Learning Schemes Miss in Online Credit Card Processing (Thus the Need for Manual Review)

In today’s online credit card processing climate, there is no silver bullet to vetting legitimate orders from fraudulent ones, in whatever industry. By now merchants know well it takes a combination of human or manual reviews and machine learning – one complimenting the other – for the best results in maintaining both a low fraud level and a low instance of falsely declining orders.

Relying on one or the other is risky because neither is perfect:

  • Time constraints: Reviewing transactions manually is time consuming, of which online merchants never seem to have enough. Manual reviews are necessary, however, for when trends change, resulting in high chargeback rates or increased false positives.
  • Black and white: In machine learning, decisions are based on explicit knowledge, what the ‘machine’ is told to decide. It is an attractive fraud deterrent to merchants because it is adjustable and persuadable until a happy medium is found.

Realistically, not all online merchants can afford machine learning schemes and must rely solely on manual reviews, again, which can be tedious and time consuming (as we blogged about last month).

We know of four instances where machine learning falls short, emphasizing the need for human review.

Online shopping at the office

Be honest. We all do it. Particularly around the holidays. When a consumer is shopping from his/her office and reaches the checkout, red flags go up for several reasons:

  • The consumer is shopping from a corporate server, different from his/her home server.
  • They may be purchasing several gifts possibly with different credit cards, to be shipped to the same location.
  • Perhaps s/he is shopping on lunch hour and has a small window of time.
  • Finally, s/he opts to ship the gifts to the office – the address which differs from the billing address.

In this case, the merchant can compare the consumer’s e-mail with the address of his/her office to see if there is a match. Additionally, the merchant can use Google Maps to compare billing and shipping address locations, based on the IP address.

Shopping while on vacation

Sometimes tourists, when traveling from warmer climes to cooler (to ski for example), purchase winter clothing to be shipped to the hotel in which they’ll be staying. Understandably, most merchants would see this as merely a case of friendly fraud: a bulk purchase shipped to a hotel, likely with a stolen credit card, thus enabling the consumer to claim s/he never received the clothing.

Goods shipped to hotel and Air BnB addresses cause suspicion as fraudsters often use them as places to retrieve goods only to never be seen again. In addition to a billing/shipping address mismatch, a cross-border purchase adds to the suspicion.

The best solution for the tourist: merchants are urged to contact the customer (tourists) for more information while the tourists would do well to contact their credit card issuer to confirm the transaction.

‘But it’s cheaper in America’: Shopping globally to save money

We have a Central American colleague who, in he and his wife’s first visit to the U.S., was astounded at how less expensive brand name clothing sells for in the States. Since, his wife often tries to purchase clothing from U.S. retailers to save money. However, her pursuit has come with hurdles as her attempts to purchase are often declined – most likely the work of a retailer’s machine learning settings.

Cross-border purchases are tricky. While a confirmation phone call is warranted here from both parties (similar to the shopping tourists), customer history is also a solid indicator of whether the transaction is legitimate. Merchants are also advised to request more information when a customer is making a cross-border purchase, such as telephone number at the very least. Otherwise, implementing 3D secure on international purchases is a good practice.

College student struggles

What do college students use credit cards for? Of course, buying books, but also ordering food, buying clothes, sneakers, school supplies and other needs. But when the order is made, declines are common because the shipping address (likely a dormitory) doesn’t match the billing address (home).

A merchant is advised to do several things here:

  • Based on the IP address, case the shipping area on Google Maps.
  • Was the order made from a mobile phone? That’s a good sign it’s legitimate as students are known for mobile shopping.
  • View the customer’s history (if s/he has one) and note similar purchases.
  • Billing and shipping mismatches are common with college students attempting to purchase goods online.

Online credit card processing: No fraud strategy is perfect

While there is no silver bullet strategy to warding off fraudsters without falsely declining orders, online credit card processing experts recommend a combination of human knowledge and a machine learning or AI scheme – and finding the best balance between the two.

We’re big fans of 3D secure technology here at Instabill, which we feel is the best deterrent available, just ahead of machine learning and AI.

Instabill is partnered with acquiring banks worldwide to get your online business running and thriving. With each of our banking partners, a variety of fraud prevention tools is available along with the best payment processing solutions. By telephoning 1-800-530-2444, you’ll speak directly with a merchant account manager who will get the process under way.

Quick Approvals

Apply Now

Leave a Reply

Your email address will not be published. Required fields are marked *