We Built Risk Models for Transactions, Not for People

We’ve spent years getting better at understanding transactions, but customer relationships don’t fit as neatly into a risk model.

Businesses have become remarkably good at evaluating transactions.

Logins, purchases, account creations, password resets, and payments can all be analyzed through dozens or even hundreds of signals. Every interaction is evaluated to determine whether it looks familiar or unusual. Over time, we’ve built complex systems for determining whether a transaction should be trusted.

But customers don’t experience businesses transaction by transaction; they experience a relationship.

A customer doesn’t think of themselves as a login attempt on Monday, a purchase on Wednesday, and a password reset two months later. They may appear separately inside internal systems, but for the customer they’re all part of the same ongoing relationship.

Customers expect businesses to recognize them over and over again. Yet many of the systems responsible for evaluating customers still examine those moments one at a time.

We’re exceptional at studying the leaves.

Now we need to understand the tree.

We Keep Looking at the Leaves Instead of the Tree

The closer we look at an individual interaction, the more we learn about it.

We can evaluate where it came from, how it compares to previous activity, and whether it aligns with what we expected to see. Modern risk systems are remarkably good at answering those questions.

Customer relationships don’t work that way.

They unfold across months and years, accumulating context with every interaction. A purchase changes the meaning of a future purchase. A login is more meaningful when viewed alongside years of account history. A series of interactions shows something larger than any individual moment.

Yet we often try to understand that relationship one event at a time.

It’s a little like studying individual leaves and wondering why it’s so hard to understand the tree. We can examine the leaf’s shape, its color, and whether it’s healthy or damaged. But the leaf only makes complete sense when we understand what it’s attached to.

The same thing happens with customer identity.

A transaction can tell us behavior differed from expectations or a signal appeared unusual. What it can’t always tell us is how that moment fits within the broader story of the customer. Was this a meaningful deviation, or just another chapter in a long-established relationship? Is this interaction inconsistent with the identity behind it, or does it fit a pattern that has existed for years?

We have more customer data than ever before yet recognizing the customer behind those observations is still surprisingly difficult. We collect more leaves every day but still struggle to see the tree.

Customer relationships are built across thousands of moments. The more interactions we collect, the more important it is to connect them into something coherent. Otherwise, every transaction is forced to explain itself in isolation, disconnected from the history that gives it meaning.

That’s the difference between evaluating transactions and understanding identity.

Transactions help us understand an event. Identity helps us understand whether that event fits the person behind it. And as digital relationships become longer, more fragmented, and more complex, it’s not enough to know that a leaf has changed; you also need to know whether it still belongs to the same tree.

What Makes a Tree a Tree?

The interesting thing about a tree isn’t any individual leaf. It’s the fact that all of those leaves belong to the same living thing.

The same idea applies to identity, which is why AtData often talks about identity as a continuity problem rather than a transaction problem.

Most organizations already have plenty of information about the interactions taking place inside their business. The harder challenge is understanding how those interactions relate to one another and whether they belong to the same person.

This is one reason email continues to be such an important identity signal, because it often remains part of the relationship long after other signals have changed. It’s one of the threads connecting purchases, accounts, communications, subscriptions, and other interactions that might otherwise appear unrelated.

Risk models are ultimately making judgments about people even when the transaction is the thing being measured. A login can appear unusual. A purchase can fall outside an expected pattern. An account opening can raise questions. What those signals don’t always reveal is whether the activity fits the person behind it.

Context like this matters because transactions don’t exist on their own. They occur within customer relationships that may have been developing for years. Viewed in isolation, legitimate activity can look unfamiliar. Viewed alongside a longer identity history, that same activity may look entirely consistent.

That’s why transaction intelligence and identity intelligence solve different parts of the same problem. After all, the goal isn’t to stop evaluating leaves. It’s to understand them in the context of the tree.

Look for the Tree, Not Just the Leaves

The leaves change with the seasons. They appear, grow, change color, and eventually fall away. Yet no one mistakes the tree for something new each time it changes. We understand that what matters isn’t any single leaf, but the larger structure connecting them all.

Customer relationships work much the same way.

Signals change. Behaviors change. Attributes change. New interactions appear while old ones fade into history. The true challenge is recognizing what remained consistent through the change.

Looking at more leaves doesn’t always make the tree easier to understand.

See how AtData helps connect transactions, interactions, and identities into a more complete view of the customer.