Declining More Transactions Reduces Fraud - and Loses Good Customers

One of the easiest ways to reduce fraud is to decline more transactions. It is also one of the easiest ways to lose good customers. Tight thresholds let more fraud through when they are too loose, and catch legitimate buyers when they are too aggressive. The second outcome is harder to see. Individual signals - VPN, a new high-value customer, mismatched addresses - are context, not a decision. Precision comes from what the signals mean together, plus your risk appetite and the commercial impact of each decline.

One of the easiest ways to reduce fraud is to decline more transactions. It's also one of the easiest ways to lose good customers.

This is the trade-off at the centre of almost every fraud system. Set your risk thresholds too loosely and more fraud gets through. Set them too aggressively and legitimate customers get caught alongside the fraudsters.

The problem is that the second outcome is much harder to see. According to Ravelin Technology, false positives cost online merchants an estimated $443 billion every year, while more than 40% of customers will abandon their cart if their payment method is declined. 25% of customers who experience a false decline will go to a competitor.

A fraud signal is not a decision

That's why merchants need to be very careful about what they interpret as a fraud signal. Take VPN usage. A transaction coming through a VPN might look suspicious because fraudsters use them to disguise their location. But legitimate customers use VPNs every day too. If VPN equals decline, you might reduce some fraud. You'll also inevitably block good customers.

The same applies to countless signals:

• A new customer making a high-value purchase
• A different billing and delivery address
• Multiple cards being used by the same customer
• An unusual location or device
• A sudden change in purchasing behaviour

None of these signals independently tells you that someone is a fraudster. They're context.

Ask what the signals mean together

This is where an overreliance on rigid acts becomes dangerous. The better question is: what do all the signals together tell me about this transaction?

A VPN combined with a long-standing account, familiar device, and normal purchasing behaviour tells a very different story from a VPN combined with multiple new accounts, several payment cards, and unusual transaction velocity.

Fraud prevention is ultimately a precision problem. You want to catch as much fraud as possible without creating so broad a definition of suspicious that good customers continually get caught inside it.

That requires understanding your risk appetite, regularly reassessing thresholds and, most importantly, understanding the commercial impact of the decisions your fraud system makes. A lower fraud rate isn't automatically evidence of a better fraud strategy. Sometimes you've simply become better at declining customers.

Related: trusting individual indicators, how to reduce false declines in Stripe, how it works, and the FAQ.

If you want ranked rule changes from your full transaction context - not one signal - book a demo.

Originally shared on LinkedIn.