What Data-Driven Marketing Actually Requires

Data-driven marketing gets talked about as if it were a destination, a state you arrive at once you've bought the right platform, hired the right analyst, and connected your ad accounts to a dashboard.

It isn't. And the chasm between believing you're data-driven and actually operating that way is where most marketing organizations lose ground.

Thanks to our awesome podcast, “What Gets Measured,” we've spent considerable time in conversation with practitioners who work at the intersection of data, AI, and marketing strategy — researchers, operators, and executives who've each confronted this gap from a different angle. What our former guests of the show describe isn't a technology problem, but a thinking problem. One that technology can either solve or make worse, depending on how you approach it.

Here's what they've taught us about what data-driven marketing actually requires.

The Mindset Problem Comes First

Before any conversation about tools or platforms is worth having, there's a prior question: are you actually using the data you have, or are you selecting the data that confirms what you already believe?

In his episode, “The Data-Driven Mindset“ Brent Dykes, author and analytics expert, frames it as the difference between leaning into the numbers and cherry-picking them:

"I think when I say data driven is we're always looking at the numbers… are we leaning into the numbers or are we just picking and choosing when we pay attention to them?"

This is what Brent calls "selective attention," which is a trap where data becomes a tool for justifying decisions rather than informing them. Leaders use dashboards to confirm their instincts, not challenge them. The result isn't data-driven marketing, but usually, the Highest Paid Person Opinion-driven marketing with a better-looking slide deck.

The goal, Brent argues, is something harder to build but more durable: intuition that has been refined by data over time. Not replaced by it — refined.

"How can data help us to make intuitively better decisions on an ongoing basis? That intuition will then become more reliable… through data."

Former CFO and AI educator Angeline Corvaglia describes in her episode, “When Data and Decisions Meet,” what that refinement Brent mentioned actually looks like in practice — and how uncomfortable it can be. Angeline recounts an inflection point where a year and a half of data trends directly contradicted her instincts about what was happening in her market:

"This was the biggest moment where I realized the power of data — that you can’t follow your instincts always, sometimes you have to follow the numbers."

Her conclusion isn't that instinct is useless, but that data is the only mechanism that can reliably correct instinct when instinct is wrong. And the leaders who refuse that correction, who treat their read of the room as more valid than eighteen months of signal, are the ones who drive their team off a cliff they can't recognize.

Truth Before Meaning

Once the data-driven marketing mindset is right, there's still a structural problem that most organizations skip past: the validity of data itself.

Scott Taylor, a data puppeteer, keynote speaker, data management consultant, and author of Lies, Damned Lies and Data, is direct about where companies go wrong. In his episode, “Truth Before Meaning: Data, Storytelling & AI,” Scott argues that the rush to derive "meaning" (analytics and AI) often fails because organizations haven't established the "truth" (standardized data) of their own business entities. He frames the challenge for leaders this way:

"You’re trying to provide value to your relationships through your brands at some form of scale. We don’t have common definitions for our brands; we don’t have a standard structure for our relationships. At scale, you need technology... but the data is what makes it work and we’ve got to invest on that data foundation, that data structure to enable all that kind of value you want to provide"

The rush to insight — to dashboards, to AI, to attribution models — consistently outpaces the work of establishing clean, standardized, trustworthy data. Taylor calls this the "truth vs. meaning" problem. Organizations invest heavily in meaning (analytics, visualization, machine learning) while neglecting the foundation that makes meaning possible.

The result is sophisticated-looking outputs built on fractured inputs. AI that confidently surfaces patterns in data that doesn't accurately represent reality. Dashboards that look authoritative but measure the wrong things. Reports that get cited in strategy meetings but were generated from inconsistent records.

In her episode, “Data Cleaning: Fix Dirty Data and Boost ROI,” Susan Walsh, known as The Classification Guru, delivered our audience a practical framework for this: COAT. Consistent, Organized, Accurate, Trustworthy. Four properties that data needs to have before it's ready to power decisions, let alone AI systems. She calls the failure state "analytical amnesia" — where inconsistent records make it impossible to see what's actually happening:

"What missed opportunities can you not see because you don't have the visibility in your data?"

For agencies and franchise brands managing data across dozens or hundreds of accounts, this isn't abstract. Every inconsistency in how campaigns are named, categorized, or attributed across locations is a gap in what you can actually see. The aggregate view you're trying to build is only as reliable as the least reliable piece of it.

What AI Requires — and What It Can't Replace

There's a version of the AI conversation that treats data literacy as optional, as if the tools are now smart enough to compensate for organizations that haven't done the foundational work. Ben Tasker, marketing technology strategist, disagrees with that take in his episode, “Why Skills Beat Tools in AI”:

"Data is the oil to AI's engine; they go hand in hand. So if the data capability/maturity isn't that high at your organization, AI's probably going to be ranked even lower than whatever that score is. So you have to figure out the upskilling together.”

Ben’s argument is that AI implementation will always be ranked lower than an organization's current data maturity. You can't run a sophisticated engine on contaminated fuel. And the solution isn't a better AI tool, but upskilling the people who generate, manage, and use the data in the first place.

"By looking at it in a skills framework,” says Ben, “you can understand where the organization needs to go."

In his episode, “The AI-Human Edge” Bill Schmarzo, the "Dean of Big Data," adds a dimension that dashboards tend to obscure: the difference between summary statistics and predictive precision. 

Most marketing analytics is built on averages — average CTR, average conversion rate, average ROAS. Averages hide variance. They smooth over the differences between individual accounts, locations, and customer segments that actually drive decision-making.

The shift he advocates is toward propensity models — approaches that predict what a specific customer in a specific situation is likely to do, rather than what customers do on average:

"Understanding propensities… allowed us to make very precise decisions based on what we thought from a probability perspective what that customer was likely to want."

But he's equally clear that data alone doesn't produce this kind of insight. You also have to understand what your customers actually value — which requires being close to the work, not just the dashboard:

"The only person in the whole equation that defines how value is created is the customer. But yet every organization says 'Well, we're going to sell value.' Well, you can't sell value if you don't understand what your customers define as valuable. You have to get in with those stakeholders to understand where and how value is defined, measured, and delivered

The Causality Gap

James Ward, causal AI researcher, identifies a failure mode that most data-driven marketing organizations don't even know they're in: mistaking correlation for causation at scale.

Most marketing measurement — attribution models, lift studies, A/B tests, media mix models — is built on correlative statistics. It identifies patterns in historical data. It finds associations. What it cannot reliably do is tell you why something worked or whether it will work again in a different context.

Ward's argument, pulled from his episode, “Rethinking Marketing Strategy with Causal AI,” is that marketers who rely entirely on correlative data are, in his words, trapped in "epistemic false certainty" — using AI to recycle received wisdom rather than identify the actual mechanisms that drive purchasing behavior:

"Study causal science if you want to be a better marketer… because there is a mathematics of causality which makes this a computable problem."

The practical implication is significant. If you're optimizing toward correlations that don't hold under changed conditions — a different season, a different competitive environment, a different audience — you're not becoming more data-driven. You're becoming more efficiently wrong.

"Every attempt to create a sale that fails is wasted energy… look at how much is wasted — it's like 99% of the effort in the ad industry."

What This Means for Agencies and Multi-Account Teams

For agencies managing dozens of clients, and for franchise brands managing performance data across hundreds of locations, the stakes of each of these problems compound.

Selective attention becomes structural when different account managers are looking at different metrics and reporting different stories to the same client. Dirty data becomes invisible when it's distributed across hundreds of accounts with no standardized naming conventions or taxonomy. Summary statistics obscure everything interesting when you're aggregating across locations that have fundamentally different customer bases, competitive sets, and seasonal patterns.

The operational question isn't just "are we data-driven?" It's: do we have the infrastructure to actually be data-driven at scale?

This is where the distinction between marketing data platforms matters, and where platform philosophy matters more than feature checklists.

NinjaCat vs Funnel vs Improvado

Competing tools with NinjaCat, like Funnel, solve the collection and normalization problem well. If your goal is moving Facebook, Google, LinkedIn, and HubSpot data into a warehouse without engineering overhead, Funnel is designed for exactly that. Improvado, another NinjaCat competitor, solves the enterprise data modeling problem — custom transformations, complex governance, warehouse integration at scale across hundreds of business units.

What neither platform is primarily designed to do, and what NinjaCat was built for, is taking unified data and putting it to work operationally — generating reports, monitoring accounts, detecting anomalies, automating workflows, and executing repeatable tasks across an unlimited number of accounts simultaneously.

That last capability is where data-driven marketing either pays off or doesn't. Clean, unified data is the prerequisite. What you do with it determines the outcome.

For agencies and franchise brands in particular, the question Bill Schmarzo asked earlier — what do your customers define as valuable? — has a specific answer. They want to know what's working, where, and why. They want that answer quickly and consistently, not just for the accounts that happen to get analyst attention this week.

Getting there requires the mindset Brent Dykes describes, the foundation Scott Taylor and Susan Walsh prescribe, the skills Ben Tasker advocates for, and the precision Schmarzo and James Ward point toward. It also requires multi-account marketing data infrastructure that can execute on all of it across every account, not just the top tier.

The Real Standard

Data-driven marketing isn't a feature you turn on. It's an operating standard that runs from how your organization thinks about numbers, through how it manages and structures data, through how it uses AI, to how it acts on what it learns — and whether it acts consistently, at scale, every time.

Most organizations are somewhere in the middle of that arc. The ones moving fastest are the ones who've stopped treating data infrastructure as a cost center and started treating it as the operating system for everything else they do. That's a different question than which platform has more data connectors.

For a deeper look at what data-driven marketing involves, and how to build toward it, see our full guide to data-driven marketing.

‍