Know Before You Spend 

It’s a question that’s plagued CMOs for decades: can we prove what campaign spending is actually causing? And for as long as anyone can remember, the answer has been ‘no.’ Sure, marketing can hypothesize based on broad conclusions, but these conclusions are educated guesses at best, and they’re almost always made after the fact. That is, after the campaign is over and its budget has been spent. 

“Marketing is one of the few line items in business where cost is known and the return is kind of a matter of opinion,” explained Tapan Patel, Research Director, Customer Data Platform, Intelligence & Analytics at IDC at a recent Uniphore webinar on marketing and AI. “We know who we spent money on and who has converted. What we don’t know is whether those two facts have anything to do with each other.”

According to Patel, that lack of clarity shows up in conversations marketing has with finance, sales, and revenue ops. Without a clear explanation of what worked in the last campaign (and what didn’t), marketers have no baseline for predicting what might work in the next one. “If you cannot measure incrementally after the fact, the only route to a defensible number is forecasting it before you commit.”

From a business perspective, that’s unacceptable. “The choices need to be made before the money gets invested, not after,” Patel says.

But how can marketers know where their campaign dollars will have the biggest impact?

Campaign simulation shows where spending will drive outcomes.

Until now, campaign spending decisions were largely driven by segment-based forecasts. This was out of necessity, however, not by choice. Audience segmentation (as we covered in our post on segment averaging and how the digital twin is rewriting the rules of customer marketing) enables humans to make some sense of the massive amounts of customer data living in customer data platforms and data warehouses. However, campaigns built around one or more segmented “averages” are notoriously imprecise.

“Segment averages built on demographics will not tell you what one person will do,” Patel says. “You actually need to know about them. You need to understand the context of the customer.”

Understanding that context—at the individual customer level—is impossible for humans alone. “You can review an audience, but you cannot review thousands or hundreds of thousands of individual decisions. That requires an operating system at a different level.”

Marketing AI is that operating system. It bridges the divide between customer data (which describes who a customer is) and learned context (their actions, behaviors, and intents). This system enables marketers to simulate how a campaign will perform at the individual customer level—not for a segment average—before a single dollar is spent.

Here’s how it works:

Each customer gets their own “digital twin.” 

The cornerstone of campaign simulation is the digital twin. This small language model (SLM) acts as a customer’s mirror representation, providing the foundation for personalization at scale through agentic motions. Because of their small size and narrow domain focus, SLMs enable individual-level modeling at scale. [The token cost of running a frontier large language model (LLM) at the depth required to predict individual-level behavior would be prohibitive at any scale.]

These twins replicate how real-world campaigns will perform.

Using digital twins, marketers can test different campaign tactics within a simulated environment. Teams can experiment with different audiences, messages, channels, timing, and other scenarios, and “preview” predicted outcomes (conversions, drop-off, revenue, cost, etc.) before launch. They can further use those outcomes to optimize and fine-tune their strategies.

The process creates a continually improving marketing AI flywheel.

The more the cycle runs—and the more customer feedback it gets—the sharper the customer picture becomes. The process creates a flywheel of intelligence: the simulation predicts, the campaign runs, the system measures the results against the prediction, that learning retrains the simulation models, and the system gets smarter with every cycle.

The more the system learns, the more the prediction becomes the plan.

The first simulation is an informed prediction. After five cycles, it is close to a reliable forecast. After twenty, the gap between simulated and actual closes to where the forecast is the plan, not an estimate attached to it.  

For CMOs, the benefit is clear: the ability to see outcomes—and how they respond to different variables—before a campaign runs fundamentally changes marketing decisioning. According to Joe Pulickal, Director of Product Management at Uniphore, that’s what marketing has long been missing. 

“We’ve spent years getting better data, better activation, better targeting, better measurement. And a lot of that’s going to get faster; but faster does not also mean better automatically. You need to really focus on that better decisioning.” 

Pulickal shared his advice on how CMOs should approach Marketing AI in the webinar: 

Treat Marketing AI as an operating model decision, not a software decision.

The gap between marketing and AI is organizational. Workflow redesign, governance, roles, and change management must keep pace with technology.

Move your planning from segments to people and then to agents.

Personalization and targeting are still largely aimed at segments. The problem today is not how much you personalize, but what you personalize to.

Set the decision rights before you widen autonomy.

Name the decisions, the threshold, the approver, and the audit trail. Start conservative and expand autonomy as you gain trust.

Budget jointly with the CIO.

Most agentic AI funding now sits in a central AI/IT budget. Joint martech investment with aligned roadmaps beats routing around IT.

Marketing AI makes better decisioning possible.

Uniphore’s Marketing AI solution delivers what no customer marketing tool or data platform has ever been able to achieve: an accurate cost-to-outcome campaign prediction for every customer, every variable, every scenario. While that means marketers can finally prove what campaign spending is directly responsible for, it’s much bigger than that. By testing variables in a simulated environment, teams can optimize the ROI of every campaign—before budget even enters the equation.