When faced with a measurement problem, many marketers are tempted to solve it by buying a shiny new toy. Today, that shiny toy is AI. It promises faster models, faster answers and, if you believe the hype, a fully formed growth strategy before you’ve finished your morning coffee.
But a faster answer is only useful if the question being asked is relevant to your business, the data behind it is reliable, and the organization is able to act on it. That was the clearest message from a panel at the 2026 ANA Measurement & Analytics Conference, where Gain Theory’s Karen Kaufman was joined by Forrester senior analyst Brad Haag and Bill Mackinson, who leads insights and analytics at global confectionery manufacturer Perfetti Van Melle (PVM).
Start with the Strategy
Marketers are investing heavily in automated models, self-service dashboards and AI agents to track and improve effectiveness. On the face of it, this means they have never been better equipped to get to insights quickly. Yet Karen noted this intelligence does not consistently translate into business growth at the same speed. She called it the “speed paradox”.
One reason is that efficiency continues to be prioritized over effectiveness. Measurement creates the most value when it helps the business make better strategic investments, not when it identifies the most efficient ones. Before going all in on AI, marketers need to know what they are trying to achieve from a business strategy perspective.
This is an approach Bill is focused on at PVM: “I’m not ROI driven. I’m strategy driven,” he said. “Sometimes that means you have to take a lower ROI to enhance the strategy, rather than just following the ROI.”
He cited the example of a new product launch, which will typically deliver a lower ROI than an established brand. But if that investment gets the product onto shelves, protects distribution and sets up the product for future success, it is still creating value and is worth the lower initial ROI.
“It’s the strategy that needs to drive the choice,” Bill said. “I need ROI to improve each year, but I don’t need it to lead me down a rabbit hole that is not productive for the organization.”
Get Your Data in Order
Many AI ambitions are derailed by inadequate data. The bottom line is that AI-enabled measurement cannot turn data that isn’t fit for purpose into a trusted business insight.
Karen shared a finding from Gain Theory’s latest research that highlighted this very issue: 55% of senior marketers their analysts spend at least 30% of their time reworking incomplete or incorrect data.
Brad noted that the reasons why this happens: data often sits in different places and is owned by different teams with uneven access. Given assembling a usable dataset can take weeks or months, his advice was to focus on getting the fundamentals right: data hygiene, taxonomy, standardization, access, governance and, above all, security.
I’ll admit this is not glamorous work, but neither is discovering that your multimillion-pound AI investment is being fed three versions of the same customer!
Think Workflow, Not Party Trick
When it comes to implementing AI into a marketing measurement program, marketers shouldn’t ask “Where can we use AI?” but “Where does work currently stall, and how could AI improve existing processes?”
The answer will be different for every organization: it may be validating incoming data, agentic modelling, or turning a dense analytical output into a clear recommendation that finance can understand and act on. No matter the starting point, Brad recommended aiming AI at a complete workflow that is going to help with a genuine growth opportunity.
Contact Us about AI Unlocked, our proprietary assessment tool that enables you to discover how AI-ready your measurement program is.
Keep the Human “So What?”
It’s also important to remember what AI cannot do. Understanding the commercial context of a business – which decisions are coming up, where trust is thin, which strategic bet needs protecting and what will make a recommendation land – are tasks that humans should not outsource.
At PVM, Bill said measurement outputs are aligned to the company’s three annual decision windows to minimize the gap between usable insight and the moment a choice is made. He stressed that measurement teams need to ensure they understand what drives business growth as much as everyone else in the company.
Karen added that building trust and being transparent were other important areas to focus on. “Make sure people understand what you’re doing, why you’re doing it, how it’s working,” she said. “Tailor those conversations to make sure you’re meeting people where they are.”
A Four-point AI Sense Check
To ensure you use AI to improve your measurement program in the most effective way, I recommend focusing on the following four areas.
- Start with the business objective: What growth outcome or decision are you trying to improve?
- Fix the foundations: Centralize and standardize the data; agree access, governance and security before pointing models at it.
- Choose the right first use case: Diagnose the tasks that one that fits your maturity—not the one making the most noise.
- Keep humans accountable. Use AI to accelerate the work, then apply business context, judgment and a clear recommendation.
