Twenty-six percent of legitimate online transactions are declined by automated fraud detection systems even when the buyer has the money and the intent to purchase. This number is a flat fact and it represents a massive failure of digital logic. The system looks at the transaction and it sees a pattern. The pattern does not fit the narrow box of the “ideal” consumer and the system kills the order.

The customer receives an email and the email says the order was canceled for their protection. This is a lie and the customer knows it is a lie. The protection is for the merchant and the algorithm is the guard at the gate. The guard is blind and he follows a manual that was written by a committee.

I spent my morning working with a student who sees the word “dog” and writes the word “bog.” I am a dyslexia intervention specialist and I look for the reason behind the error. My student is honest and he is trying and his brain simply processes the curves of the letters in a way that the standard curriculum does not allow. I understand him and I help him bridge the gap.

But the digital world has no intervention specialists. It only has “if-then” statements. If the IP address is different from the billing address, then the risk score increases. If the customer uses a new device, then the risk score increases. The honest man is a messy creature and he moves houses and he loses his phone and he buys gifts for friends in different states. He trips every wire and the system flags him as a threat.

The Polished Ghost vs. The Messy Human

The professional fraudster does not make these mistakes. He is a student of the system and he knows the manual by heart. He uses a virtual private network and he matches his location to the billing address of the stolen card. He uses a clean browser and he mimics the behavior of a robot that is pretending to be a human.

He is a polished ghost. He sails through the checkout and the system welcomes him. The system likes him because he is legible. He fits the data points and he does not trigger the alarms. The fraudster is a counterfeit of a “perfect” customer and the algorithm loves perfection.

The Vending Machine Crisis

In the late 1940s, the vending machine industry faced a crisis with the transition of metal alloys. The machines were built to accept nickels and they used a mechanical balance to check the weight and the size of the coin. A new nickel was heavy and it had sharp edges. A nickel that had been in a man’s pocket for ten years was light and the edges were smooth.

The machines began to reject the old nickels. The old nickels were real and they were authentic and they had been earned through labor. The machines rejected the truth and they accepted the “slugs.” The slugs were lead discs that were cast in a mold to match the exact weight and diameter of a brand-new nickel.

The scammers made the slugs “perfect” and the machines were happy. The honest man with the worn-out coin in his pocket stood hungry in front of the machine and he did not understand why his money was no longer good.

We have built a digital version of that vending machine. We have created a world where authenticity is a liability and the appearance of authenticity is the currency. I see this in my work and I see it when I try to buy things online. I live in a world of patterns and I know that the most human thing a person can do is break a pattern.

I recently sent a text message to a colleague. I was frustrated with a new district mandate and I used strong language. I hit send and I realized the name at the top of the screen was not my colleague. It was the mother of one of my students. The phone did not stop me. The system saw a valid contact and a valid network connection and it delivered the payload. The system was “correct” and I was a disaster. The technology followed the data and it ignored the context.

The Wall of Code

When an adult wants to purchase a specific product, they look for a store that understands the product. They might look for disposable vapes online because they want a device that is authentic. They want the MT15000 Turbo or the MO20000 PRO. These devices are physical and they have a specific weight and they have a specific flavor like Blue Razz Ice or Strawberry Kiwi.

The person who wants these things is a real person and they have a real life. They might be ordering from a hotel while on a business trip. They might be sending a multi-pack bundle to a cousin as a birthday present. These are “unusual” behaviors according to a crudely written risk algorithm. The algorithm sees the “unusual” and it labels it as “fraudulent.”

I hate the way we have outsourced our judgment to these machines. I say this even though I rely on them to organize my calendar and to track the progress of my students. I am a man of contradictions and I recognize that. I want the convenience of the digital age but I want the empathy of the old world. I want a merchant to look at my order and see that I have been a customer for three years. I want them to see that I am a real person who just happens to be in a different zip code for the weekend.

The Ghost in the Bell Curve

The fraud-detection systems are built on the idea of the “average.” But no one is average. The average is a mathematical ghost. When you build a system to catch the outliers, you catch the people who are living the most interesting lives. You catch the people who are traveling and the people who are changing and the people who are human.

The professional criminal stays in the center of the bell curve. He hides in the crowd and he wears the mask of the “average” man. He is the lead slug in the vending machine and he is the one who gets the candy.

The store that knows its customers is the only defense against this. A focused store that specializes in a single brand can afford to look at the details. They know what a real order looks like and they know what their people want. They do not need to rely on a giant, faceless algorithm that treats a purchase of a Nera 70K the same way it treats a purchase of a high-end laptop or a bag of diamonds. The scale of the “big data” systems is their weakness. They are too big to see the individual and they are too fast to hear the explanation.

Legibility and Control

I often think about the “legibility” of our lives. James C. Scott wrote about how states try to make their citizens “legible” so they can be taxed and controlled. They created last names and they created street addresses and they created standard units of measurement. The digital world is the final stage of this process.

This is a cold way to live and it is a cold way to do business. My student with dyslexia is not “legible” to the state testing system. The system says he is failing. I look at his drawings and I listen to his stories and I know he is brilliant. He sees the world in three dimensions and he understands mechanics in a way that the “top” students do not. He is an authentic human and the system flags him as a broken part. We do the same thing at the digital checkout. We reject the authentic person because their data is not “clean.”

The frustration of the decline is not just about the product. It is about the indignity of being told you are a lie. You hold the card in your hand and you know you are real. You see the screen and the screen says you are “suspicious.” It is a denial of your identity.

The actual bad actors do not feel this indignity. They do not care about being “real.” They only care about the result. They have a hundred cards and a thousand IP addresses and they will keep hitting the wall until it breaks. The honest man hits the wall once and he feels the sting and he walks away.

“We need a return to the human scale. We need systems that are built to recognize the messiness of life. Until then, we are all just coins in the machine, hoping our edges are sharp enough to pass the test.”

I will continue to teach my students that their “errors” are often just a different way of seeing. I will continue to hope that the person on the other side of the screen sees me, too. The algorithm rewards the polished ghost but the warehouse only ships to the man with the messy signature.