Here’s a truth that keeps me up at night: most businesses are using machine learning to predict the past.
They feed their models purchase history. They optimize for “likelihood to click.” They serve winter coat ads in October because someone bought a winter coat last November.
That’s not prediction. That’s pattern matching. And pattern matching is a rearview mirror.
The real opportunity-the one almost nobody is talking about-is something I call Anticipatory Empathy. It’s the difference between knowing what a customer will buy and understanding why they’ll buy it, and when they’re ready to hear from you.
Let me show you what this looks like in practice.
The Fallacy of the Historical Model
Most ML strategies are built on a logical but flawed premise: “Past behavior predicts future behavior.”
Here’s why this fails.
A customer buys a premium mattress from you. Eighteen months later, your model flags them as a “high probability” sheet buyer. So you serve them sheet ads. Makes sense, right?
Except your model missed the real signal: They bought the mattress because they just bought their first house.
That house purchase is a life architecture event. It opens a window-maybe 72 hours-where this person is making dozens of high-stakes decisions. Furniture. Appliances. Services. Everything is on the table.
But here’s the problem: During that window, their cognitive load is massive. They’re overwhelmed. They’re decision-paralyzed.
Most brands see this paralysis and pull back. “They’re not clicking. They’re not ready.”
But a brand practicing Anticipatory Empathy sees something different: a customer who needs a guide, not another ad.
The ML model shouldn’t just predict “buy sheets.” It should predict “this customer is in a high-stakes decision window and needs a different kind of support.”
That’s a fundamentally different strategic question.
Three Strategic Vectors for Anticipatory Empathy
Here’s how we operationalize this thinking. These aren’t theoretical. They’re the frameworks we use daily.
1. Predicting Cognitive Load
Standard ML tracks bounce rates and cart abandonment. It asks: “Did they leave?”
We ask a different question: “How hard is their brain working right now?”
Using Natural Language Processing on customer support tickets and social DMs, we score every interaction for emotional state. Is this customer frustrated? Confused? Overwhelmed?
When someone asks the same question three times, that’s not a support issue. That’s a cognitive load signal. Their brain is full. They’re not shopping-they’re surviving.
The strategic response isn’t to serve them a retargeting ad. It’s to offer a human conversation. A phone call. A concierge service.
Here’s what’s counterintuitive: When you predict high cognitive load, the right move is often less marketing, not more. You earn trust by stepping back. And that trust becomes the foundation for every future transaction.
2. The Receptivity Score
Everyone looks at watch time on TikTok and Reels. We look at something different: parasocial proximity.
This is our ML model’s analysis of commenting patterns, direct messages, and engagement depth. We’re looking for semantic signals. Is this person saying “Cool video” or “You always know what I need”?
The difference matters. It tells us how close this person feels to our brand. And that tells us what kind of message they’re ready for.
- Low proximity: They need proof. Social proof, testimonials, case studies, comparison charts. Don’t ask for commitment. Show evidence.
- High proximity: They need stories. Founder narratives, behind-the-scenes content, long-form YouTube deep dives. They’re ready to feel something, not just evaluate something.
Most brands treat all warm audiences the same. They blast identical retargeting creative to everyone. Anticipatory Empathy means reading the room-predicting which message this specific person is ready to receive.
3. The Decision Velocity Window
This is where we see the biggest edge on Facebook and Pinterest.
We build models that predict not just “likelihood to buy” but “likelihood to buy under indifference.”
Here’s the profile: High frequency of browsing. Low frequency of conversion. High engagement with aspirational content-saves, collections, “dream boards.”
These users aren’t price-sensitive. They’re decision-paralyzed. They want to buy. They’re just stuck.
The ML prediction here isn’t “show them a discount.” The prediction is “they need a permission structure.”
Our response: Social proof carousels. Time-bound scarcity signals. “Here’s exactly what someone like you chose.”
We don’t optimize for the lowest CPA. We optimize for the highest decision velocity within this specific user cluster. Speed of conviction. Speed of action.
The Lean Startup Approach to ML
I believe in running a tight ship. So here’s a 90-day roadmap that doesn’t require a data science team or a six-month implementation.
Days 1-30: Find your friction signals
Stop looking at your CRM. Start looking at your support data and community interactions. Tag every touchpoint for emotion.
- When are customers frustrated?
- When are they confused?
- When are they delighted?
This is your fuel. Don’t build a model yet. Just understand the emotional landscape of your customer journey.
Days 30-60: Build a catalyst model
Don’t train a model to predict “churn.” Train it to predict “catalyst.”
- When do your customers experience their biggest life changes?
- New job? New baby? New home? New role?
These are the moments when everything is in flux. These are the moments when a brand with Anticipatory Empathy can step in and say: “We see you. We know what you’re going through. Here’s how we can help.”
Days 60-90: Close the creative loop
Feed your predictions into your creative strategy.
If the model predicts high frustration, run utilitarian, problem-solving creative. Show them the solution before you show them the story.
If the model predicts high aspiration, run narrative-driven creative. Let them feel the possibility before you ask for the click.
The offer doesn’t change. The creative barely changes. What changes is the order and the timing.
What This Means for Business Leaders
Machine learning isn’t going to replace strategists. It’s going to give great strategists superhuman empathy.
The AI tells you what’s happening. But you-the leader, the marketer, the human-have to decide what it means and what to do about it.
That’s the gap most brands are missing. They have the data. They have the models. But they’re asking the wrong questions.
They’re asking: “Will this customer buy?”
They should be asking: “What is this customer experiencing right now, and how can we show up for them?”
That’s Anticipatory Empathy. And it’s the only machine learning metric that matters.
