Clearsale Blog Posts
Analyzing transactions looking for fraud requires skill. Should the solution analyze all e-commerce transactions, or only those that look like fraud? The former could take more time and money, but it is the most efficient manner to make sure fraudulent transactions are not approved. The second option takes less labor and can be managed by regular fraud filters, but leaves you open to orders that may appear innocent but are actually fraud. In the current digital market, most transactions are legitimate.
However, some are not and that is why we must take precautions. This means finding the right balance between speed and efficiency to protect your profit and your brand reputation.
Let's take a look at the pros and cons of each approach
The Risks of Selective Transaction Analysis
Whenever companies choose to use artificial intelligence for selective analysis of only those transactions that appear fraudulent, they are taking a big risk.
• Difficulty figuring out what is fraud and what is a legitimate order. Inexperienced merchants may mistakenly approve fraudulent transactions that appear legitimate.
• Reject too many transactions. A low rate of fraud could simply mean a high number of rejections, which could have a negative impact on your reputation.
• The changing face of fraud. Traditional strategies based on artificial intelligence may fail to detect new trends in fraud. When fraudsters evolve their tactics, the rules of artificial intelligence don't always keep up, rendering fraud control solutions inefficient.
• Playing with the system. Fraudsters are very good at slipping through fraud filters unnoticed. If they find out that a merchant analyzes only transactions over R$ 1.000,00, they will keep their orders under R$ 999,00 so they are automatically approved.
• Complex management of fraud filters. Merchants normally have a number of rules in their fraud filters to flag as many fraudulent transactions as possible. However, this can also increase exposure to risk. If merchants are not careful about the order in which the filters are used (first, second, etc.), some of these rules may cancel each other, reducing the amount of protection.
• Failure to see the entire picture. Handling individual transactions may mean failure to see the entire picture. For example, buying a notebook for delivery on a specific street may appear a legitimate transaction. However, unless the merchant is analyzing all of the transactions, it may not realize that 10 notebooks were delivered to different addresses on this same street, all within a short period of time. This may be an indication of a fraud attack.
Focusing on individual transactions, rather than the whole, may make preventing large attacks even harder.
The Advantages of Analyzing all Transactions
Manual selection of transactions for analysis is not an answer. It may be better to analyze all of the transactions using additional, complementary tools.
What are the advantages of this?
• Low cost may indicate danger. Fraudsters tend to place low-cost order to make sure a stolen credit card really works. if it does, this opens the door to larger fraudulent orders. Analyzing all transactions allows you to detect small purchases designed to test the fraud prevention system - transactions that might go unnoticed otherwise.
• Analyzing more transactions allows you to put together an overall view. The more transactions you analyze, the greater the amount of data in your transaction database that can be used in future to make better informed solutions. Consider our previous example of notebooks delivered to Rua Tabapuã. Each individual sale may appear innocent, but the pattern is only visible when all transactions are taken together.
• Doubtful transactions may be examined more closely. Some transactions may be undefined, and one cannot tell for sure if they are fraudulent or not. Fraud control systems based solely on artificial intelligence will automatically reject all these orders. If you have a system that investigates and analyzes these transactions, it will know that some of these are good orders and should be approved.
• Avoid large-scale fraudulent attacks. Professional fraudsters work in concert, orchestrating coordinated attacks that happen over a short period of time. If a single company analyzes the order it's like throwing cold water on a fire, attacks can be identified, interrupted and avoided far more quickly.
Artificial intelligence brought major progress to fraud control, however companies cannot rely solely on AI to check which orders are legitimate, and which are fraud. Instead, companies that combine artificial intelligence and a team of human analysts find that this team complements artificial intelligence, making it smarter and more efficient, so that the correct transactions are approved and increasing customer satisfaction, capturing their loyalty.
Developing a Comprehensive Fraud Control Solution
While fraudsters continuously improve their tactics, it is important that online merchants develop a comprehensive fraud prevention approach that includes:
- Advanced technology to quickly accumulate data
• Statistical intelligence to determine which data patterns are suspect and require detailed analysis
• Sophisticated human analysis to help companies develop a broader view to increase order approval.
This approach is ClearSale's distinction.
Technology, while impressive, is not enough to fight fraud. When merchants have a more inclusive view of their sales, including technology and human analysis, they can increase the accuracy of their fraud detection and be up-to-date in a constantly changing market.
ClearSale combines the analysis of large volumes of data, statistical intelligence and human intelligence to offer an accurate balance between fraud protection and greater sales.