Customer behavior signals that should change your decisions

Some customers never complain. They give good scores on the satisfaction survey, they do not post on social media and they never ask to speak with a manager. They simply buy a little less each month, until one day they stop showing up. By the time the leadership team reviews the quarterly results, the loss has already happened and nobody knows exactly when it began.

This is the second article in a three-part series. In the first, The mistake of turning a customer into a photograph, we saw how segments, Personas and journeys lose relevance when they are treated as permanent descriptions of a customer who keeps changing. Here we take the next step. If the photograph ages, what should we watch between one photograph and the next? And above all, when does a behavioral signal justify changing a decision?

The question is not academic. It determines whether an organization reacts while it can still recover a relationship, or finds out what happened after the customer is already gone.

What customers do today can say more than what they said six months ago

Because, in many cases, what the customer is doing today can tell us more about the experience they need than what they told us when we researched them six months ago.

Let’s consider a simple situation.

In a survey, a customer explains that she values convenience, that she tends to be loyal to brands that work for her, and that she prefers to always shop at the same place. The conversation is productive, the information is consistent, and several people with similar characteristics respond in a similar way.

A profile is created.

Six months later, that same customer continues to say she values convenience and likely still considers herself loyal. But something begins to happen in her transactions.

• First, she buys a little less.
• Then she starts switching to another supplier.
• After that, she stops buying a product category she used to purchase regularly.
• , she responds to a promotion, comes back once, and then disappears again.

If the company looks only at what the customer said, the relationship appears stable.

If it looks at her behavior, the story is different.

And neither source is necessarily wrong.

That’s one of the most interesting aspects of trying to understand customers: people can’t always explain exactly how their own decisions are changing.

A customer may still think that a brand is their first choice and, at the same time, have started buying more frequently from a competitor. They may say that price isn’t the most important factor and yet consistently respond to certain promotions. They may claim to be satisfied and quietly begin to reduce their engagement with the company.

There isn’t necessarily a contradiction.

1. What a person says usually explains how they interpret their relationship.
2. What they do reveals how that relationship is actually unfolding.

We need both.

Listening to the customer doesn’t mean simply asking them questions

In Customer Experience, we tend to associate the Voice of the Customer with surveys, interviews, feedback, or qualitative studies. These are indispensable sources because they give us access to something that a transaction alone can hardly explain: emotions, expectations, frustrations, motivations, and reasons.

But the customer also “speaks” when they don’t respond to a survey.

They speak when:

• they reduce their frequency;
• they stop using a channel;
• stops following through on a process they used to complete;
• starts buying only on sale;
• switches to a different product or category;
• calls three times about the same problem;
• stops using a feature;
• decreases the value of their relationship;
• returns after months of absence.

These are not answers to a research question.

They are behavioral clues.

And they often appear before the customer has an explicit conversation with the company about what is happening.

The ICX methodology incorporates precisely this dual perspective. On the one hand, CX Intelligence seeks to understand the customer experience, its challenges, and the necessary interventions; on the other, Commercial Intelligence analyzes behaviors, opportunities, and what might be changing in the business relationship with the customer. These are not two separate conversations: one helps us understand how the customer experiences the relationship, while the other allows us to observe how that relationship manifests itself in specific decisions.

This combination avoids a fairly common mistake: treating experience and behavior as if they belonged to different worlds.

Because a poor experience can also manifest itself in a transaction.

It’s just that it’s rarely labeled as a “bad experience.”

Sometimes the customer doesn't complain. They just move on.

Let’s imagine another scenario now.

A customer of a financial institution has been using their card regularly for two years. They have never filed a major complaint, and their previous surveys show high levels of satisfaction.

• One month, a charge appears that they don’t recognize.
• He calls the customer service center.
• The call takes longer than expected, and resolving the issue requires several steps.
• Finally, the problem is resolved.
• He doesn’t file an additional complaint. He doesn’t post anything on social media. He doesn’t request to cancel the card.

From a traditional perspective, the incident is closed.

But over the following weeks, something different happens:

Regular use → incident → decreased use → shift in shopping habits → almost no transactions.

The customer never said:
“This experience undermined my trust, and I’m going to start using another card.”

They didn’t have to.

Their behavior began to speak for them.

That kind of signal is precisely what a static view can overlook. The profile remains the same. The demographic data hasn’t changed. Perhaps the segment hasn’t either. Even the latest satisfaction survey may still show a positive result.

What changed was the trajectory of the relationship.

And to see it, you have to look at the timeline—not just the data.

That’s why variables such as recency, frequency, spending trends, retention, reactivation, price sensitivity, affinity between categories, or time to churn are relevant within a Customer Experience Intelligence system. Not because they all need to be analyzed for every customer, but because they allow us to detect patterns that a single snapshot doesn’t reveal.

An isolated data point provides information. A sequence begins to tell a story.

That difference deserves attention.

• A $50 purchase says very little.
• A $50 purchase every week for a year tells a different story.
• Five weeks without a purchase after that usual pattern says even more.
• And a poor service rating right before those five weeks adds a piece of information that could change our interpretation of everything that came before.

We could represent it this way:

Transaction
What did they do?

+ history
Is this normal?

+ Context
What was happening?

+ perception
How did he experience it?

= meaningful signal

That’s where intelligence begins to emerge.

Not when we gather more information about the customer, but when we can connect the dots to understand what might be happening right now.

The methodology approaches this combination by using different sources: transactions, digital behavior, purchase history, surveys, open-ended comments, and analytical models. The goal is not to indiscriminately transfer all that information to the CRM, but rather to first convert it into knowledge that can be used in the customer relationship.

This matters because organizations typically already have a considerable amount of data.

The challenge isn’t always about getting more.

It’s learning to recognize when a change warrants attention.

Not every change means there's a problem

We need to be careful here, too.

If a customer buys less, it doesn't automatically mean they're dissatisfied.

They may have finished a project.

They may be traveling.

His financial situation may have changed.

They may need the product less frequently.

In B2B, a decline in usage may be due to seasonality or a change in the customer’s operations. In insurance, less interaction may be perfectly normal. In healthcare, it could even be a good sign.

That’s why observing behavior doesn’t mean interpreting every variation as a cause for alarm.

It means establishing enough context to distinguish between noise and signal.

One fewer purchase is a data point.

A sustained decrease over several periods is a pattern.

A sustained decline following several service incidents begins to form a hypothesis.

And if that same behavior appears among hundreds of customers with similar experiences, we are no longer observing just individual cases.

We are learning something about the experience.

That is precisely the leap from collecting data to generating insights.

The question shifts from “What do we know?” to “What is changing?”

This shift in the question forces us to reevaluate how we understand the customer.

• The Persona still makes sense.
• The customer journey still makes sense.
• Research still makes sense.

But now they coexist with an additional layer that allows them to be constantly confronted with reality.

• Do the customers we identified as highly service-sensitive continue to behave that way?
• Do the moments we define as critical still have the same impact?
• Is a new source of friction emerging?
• Has a segment started to behave differently?
• Does a trait we considered central to a Persona truly still explain their decisions?

The answer no longer depends solely on conducting a comprehensive study every few years. Part of it can be found in the signals customers leave behind every day.

And this takes us into even more challenging territory.

Because detecting that something has changed can be interesting for an analyst.

It can generate a dashboard.

It could even make for a great presentation to a committee.

But from the customer’s perspective, none of that matters if the company continues to do exactly the same thing.

A signal begins to have real value when it changes a decision.

Signals matter when they change a decision

A signal begins to have real value when it changes a decision.

Let's go back to the credit card customer.

The financial institution has already noticed that, following a service incident, the customer began using the product less frequently. The pattern does not appear to be random. There is a sequence clear enough to warrant attention.

So far, the organization has learned something.

But the customer hasn’t noticed any difference yet.

There may be a dashboard showing the drop in usage. An analyst may have identified the pattern. It might even come up in a monthly meeting as an interesting case.

As long as no one acts on that information, from the customer’s perspective, everything remains the same.

That’s one of the areas where many customer experience initiatives fall short: they’re good at explaining what happened, but much less effective at deciding what should happen next.

And it’s not always because of a lack of data.

Sometimes the exact opposite is true. There’s simply too much.

We know when they made a purchase, how much they spent, which channel they used, whether they called the contact center, how long the interaction lasted, what they answered in the last survey, and which products they have subscribed to.

The difficult question comes next:

What are we doing differently now that we know all that?

Detection is not the same as intervention

Let’s assume the signal arrives in time.

The institution identifies that this is a valuable customer who, until the incident, had a stable usage pattern and whose usage began to decline immediately after a bad experience.

There are several possibilities:

• You could offer them a promotion to encourage them to use the service again.
• It could call the customer to try to rebuild the relationship.
• You could provide them with access to a more direct customer service channel.
• You could acknowledge the problem they had and verify that it has truly been resolved.
• Or I could decide not to do anything yet and continue to monitor the situation.

All of these are possible decisions.

The difference is that now the action doesn’t stem simply from the segment to which the customer belongs. It stems from a combination of who they are, what happened, and what signal they’re showing right now.

That’s where personalization begins to take on a whole new meaning.

It’s not just about changing the name in an email.

Nor is it sending a different offer just because someone belongs to segment A instead of segment B.

It’s about deciding which experience makes the most sense right now.

The ICX methodology embodies precisely this logic under the concept of the Next Best Experience. Unlike the traditional Next Best Offer, the response doesn’t always have to be commercial. It can involve facilitating, anticipating, helping, recognizing, recovering, rewarding, selling, or even deciding that the best course of action is to do nothing.

That last point deserves a moment’s consideration.

Sometimes, the best experience is to not intervene

In an environment where companies can automate virtually any communication, there is a natural temptation: to turn every signal into a message.

• They bought something: message.
• They stopped buying: message.
• Visited a page: message.
• Opened the app: message.
• Didn't open it: another message.

The risk is ending up using customer intelligence to increase the volume of contact, rather than to improve the experience.

That’s why the idea proposed by the methodology is so important:

Personalization isn’t about communicating more; it’s about intervening more effectively.

A customer who has just solved a complex problem may not immediately need a cross-selling campaign.

Perhaps they simply need the company to verify that the solution worked.

A customer who has been buying the same product for years might appreciate it if the next interaction were simpler, rather than someone trying to sell them something extra.

A user who is browsing information may not need any intervention just yet.

And a customer who has shown clear signs of frustration probably shouldn’t receive the same promotional message as someone whose relationship with the brand is at its peak.

The signal, therefore, should not automatically trigger a campaign.

It should trigger a decision.

From “this customer belongs to…” to “this customer needs…”

Here, there’s an interesting shift in the way we think about the customer experience.

An approach based solely on profiles usually starts with something like this:

The customer belongs to segment X
↓
Segment X receives experience Y

It's organized. Easy to automate. And, in many cases, useful.

But when we incorporate behavior and context, the sequence becomes richer:

Customer belongs to segment X
+
has a certain relationship with the company
+
an event has just occurred
+
is exhibiting behavior that differs from its usual pattern
↓
What is the most appropriate course of action right now?

That’s when we stop using the segment as a response and start using it as part of the response.

The difference may seem subtle, but it significantly changes the quality of decisions.

Let’s consider three customers who belong to the same Persona.

All three have similar characteristics. They buy similar products. They use digital channels and have comparable economic value.

But today they are in different situations:

The person hasn't changed.

The recommended experience did.

And that is precisely why the methodology proposes translating the analysis into concrete actions through CX Activation Plays®, such as protecting a valuable relationship at risk, responding to a bad experience, facilitating a repeat purchase, or supporting a customer’s reactivation.

The true value becomes apparent at the moment of decision.

This is also a good test for any customer analytics initiative.

We might ask ourselves:

• What decision changes if this metric goes up or down?
• Who should receive that signal?
• When is there still time to act?
• What kind of intervention would make sense?
• How will we know afterward if it worked?

If we can’t answer those questions, we probably have some interesting insights, but we still don’t have a fully functional end-to-end Customer Experience Intelligence system.

Because the goal isn’t just to accumulate signals.

It’s to turn them into insights.

And that’s where another important difference comes in: intelligence doesn’t have to reside in the same place as all the data that produces it.

A company may have millions of transactions, comments, clicks, tickets, surveys, and historical records spread across different systems. Trying to bring every single one of them into the CRM doesn’t necessarily help the sales executive, the consultant, or the service agent who has to decide what to do with a customer.

What that person needs isn’t always to see twenty years of history.

They need context.

They need to know, for example:

this customer is high-value, the relationship had been stable, an incident has just occurred, their activity has decreased, and there is a risk of deterioration.

That is no longer raw data.

It is an interpretation that enables action.

And that is precisely the bridge to the next part of this conversation.

Because if we want the signals to actually change decisions, then we have to answer a very practical question:

What should the CRM know about the customer to help the organization deliver a better experience, without turning it into a data warehouse?

From behavior to context

That question is the starting point of the third article in this series, Your CRM doesn’t need more data, it needs customer context, where we look at what intelligence should reach the CRM and how to make the experience keep learning as the customer changes.

If your organization already has behavioral data but does not yet turn it into experience decisions, our Customer Experience team can help you define which signals matter and which interventions make sense at each moment of the relationship.

Request a diagnostic

Complimentary initial diagnostic