The biggest problems with automatically declining transactions
Posted: Sun Dec 22, 2024 5:03 am
One of the biggest challenges in using traditional fraud filters and fraud rates is not preventing fraud itself, but ensuring that legitimate transactions are not being declined.
It’s clear that merchants need a better transaction review process. When fraud rate tools or fraud filters are too sensitive, the result is a high number of false positive transactions, which leads to card declines for legitimate purchases.
In fact, automated review processes are not the best way to combat fraudsters who reject so many legitimate orders. When the review process is done automatically, sellers typically flag up to 27% of all transactions as high risk, the majority of which are automatically rejected by fraud prevention solutions on the market today.
So what happens when legitimate orders are rejected email list providers in germany fraudulent ones are approved? Here, we explore two of the main consequences.
1. You are missing out on profits
Fact: False declines cost more than fraud.
There can be up to 40 false positives for every true fraud attempt. This means that up to 95% of transactions flagged as high risk could be legitimate. These false positives result in declined cards, a huge loss of sales, frozen accounts, and an overall poor experience for your customers.
An average of 15% to 30% of consumers fail identity verification tests based on personally identifiable information (PII) and life history questions, while up to 60% of criminals and fraudsters do. Furthermore, Gartner’s state government clients report that more citizens have had their identity compromised than those who have not.
What’s worse, depending on the margins generated by a given business model, it could take a dozen or more legitimate transactions to cover the costs of one fraudulent order that was wrongly approved.
Retailers lose more money to false declines—$118 billion per year—than they do to credit card fraud, which totals about $9 billion. What’s more, 32 percent of customers who experience a false decline never shop with the same merchant again.
Contrary to what many merchants think, a “tougher is better” approach doesn’t actually deter criminals. In fact, it’s likely to just irritate your legitimate customers.
2. Your anti-fraud protection model is not efficient
Advanced analytics increase the concentration of fraud relative to legitimate transactions at high rates and decrease the concentration of fraud relative to legitimate transactions at lower rates. However, it is important to remember that legitimate transactions will be among those with a high fraud rate.
For example, if you have a customer who typically buys expensive items quickly, purchases at odd hours, and uses multiple shipping addresses, most merchants would assume this is a fraudulent transaction and flag it as having a high fraud rate. But it is possible that this customer is wealthy, a night owl, and impulsively buys gifts for family or friends – making this purchase completely legitimate.
A fraud rate is only as good as the information used to compute the rate. If the data analyzed is incorrect or does not represent the real world, it is not useful. This concept is simple, but often forgotten by merchants.
For example, in a risk analysis position, people often think of data as being useful for extrapolating the probability of future events. But this is only true if we have data showing that the events we are concerned about occur with the same frequency and with the same patterns as in the real world. You might conclude that this is not true for businesses that reject all high-risk transactions, assuming that they are all fraudulent.
Achieving an Effective Transaction Review Process
The difficulty in achieving an effective and efficient review process is not unique to small and medium-sized businesses. Enterprise-level companies face similar obstacles.
Most companies face the same dilemma in the review process: whether to over-process or under-process transactions. In fact, most high-risk orders are not reviewed, and companies alone do not have enough or specific information to approve the transaction.
When building any predictive index, the general rule is the more information, the better. However, while it is important to analyze all data sources when building and managing a fraud index, corrupt data can lead to erroneous answers in the final index.
Corrupt data can come from many sources—inconsistencies, incomplete or duplicate data, and more—and result in a messy analytical model.
How to build fraud rates correctly
The data used to build a fraud index should represent your future customers or orders that will be placed.
Let’s use this fraud index as an example:
Imagine you want to build a fraud index that will catch future card-not-present (CNP) fraud orders. The future population then consists of all CNP orders in your business. Now you need a section of the historical population (orders placed in the past) to build your analytical model. And in the past, you automatically declined 10% of all orders (those with a high fraud rate).
This implies that the historical population has two subgroups:
1. Automatically decided orders:
• Automatically approved orders. These orders may never lead to a chargeback (legitimate orders) or may lead to a fraudulent chargeback (true fraud).
It’s clear that merchants need a better transaction review process. When fraud rate tools or fraud filters are too sensitive, the result is a high number of false positive transactions, which leads to card declines for legitimate purchases.
In fact, automated review processes are not the best way to combat fraudsters who reject so many legitimate orders. When the review process is done automatically, sellers typically flag up to 27% of all transactions as high risk, the majority of which are automatically rejected by fraud prevention solutions on the market today.
So what happens when legitimate orders are rejected email list providers in germany fraudulent ones are approved? Here, we explore two of the main consequences.
1. You are missing out on profits
Fact: False declines cost more than fraud.
There can be up to 40 false positives for every true fraud attempt. This means that up to 95% of transactions flagged as high risk could be legitimate. These false positives result in declined cards, a huge loss of sales, frozen accounts, and an overall poor experience for your customers.
An average of 15% to 30% of consumers fail identity verification tests based on personally identifiable information (PII) and life history questions, while up to 60% of criminals and fraudsters do. Furthermore, Gartner’s state government clients report that more citizens have had their identity compromised than those who have not.
What’s worse, depending on the margins generated by a given business model, it could take a dozen or more legitimate transactions to cover the costs of one fraudulent order that was wrongly approved.
Retailers lose more money to false declines—$118 billion per year—than they do to credit card fraud, which totals about $9 billion. What’s more, 32 percent of customers who experience a false decline never shop with the same merchant again.
Contrary to what many merchants think, a “tougher is better” approach doesn’t actually deter criminals. In fact, it’s likely to just irritate your legitimate customers.
2. Your anti-fraud protection model is not efficient
Advanced analytics increase the concentration of fraud relative to legitimate transactions at high rates and decrease the concentration of fraud relative to legitimate transactions at lower rates. However, it is important to remember that legitimate transactions will be among those with a high fraud rate.
For example, if you have a customer who typically buys expensive items quickly, purchases at odd hours, and uses multiple shipping addresses, most merchants would assume this is a fraudulent transaction and flag it as having a high fraud rate. But it is possible that this customer is wealthy, a night owl, and impulsively buys gifts for family or friends – making this purchase completely legitimate.
A fraud rate is only as good as the information used to compute the rate. If the data analyzed is incorrect or does not represent the real world, it is not useful. This concept is simple, but often forgotten by merchants.
For example, in a risk analysis position, people often think of data as being useful for extrapolating the probability of future events. But this is only true if we have data showing that the events we are concerned about occur with the same frequency and with the same patterns as in the real world. You might conclude that this is not true for businesses that reject all high-risk transactions, assuming that they are all fraudulent.
Achieving an Effective Transaction Review Process
The difficulty in achieving an effective and efficient review process is not unique to small and medium-sized businesses. Enterprise-level companies face similar obstacles.
Most companies face the same dilemma in the review process: whether to over-process or under-process transactions. In fact, most high-risk orders are not reviewed, and companies alone do not have enough or specific information to approve the transaction.
When building any predictive index, the general rule is the more information, the better. However, while it is important to analyze all data sources when building and managing a fraud index, corrupt data can lead to erroneous answers in the final index.
Corrupt data can come from many sources—inconsistencies, incomplete or duplicate data, and more—and result in a messy analytical model.
How to build fraud rates correctly
The data used to build a fraud index should represent your future customers or orders that will be placed.
Let’s use this fraud index as an example:
Imagine you want to build a fraud index that will catch future card-not-present (CNP) fraud orders. The future population then consists of all CNP orders in your business. Now you need a section of the historical population (orders placed in the past) to build your analytical model. And in the past, you automatically declined 10% of all orders (those with a high fraud rate).
This implies that the historical population has two subgroups:
1. Automatically decided orders:
• Automatically approved orders. These orders may never lead to a chargeback (legitimate orders) or may lead to a fraudulent chargeback (true fraud).