How Fintech Companies Achieve Strong Return on Investment by Leveraging Data

A solid foundation of integrated data is essential for boosting return on investment in fintech marketing efforts. Credit: Getty Images

Stephen Williams, CEO of Marketing Evolution, explains why unified data represents a crucial missing element for fintech companies aiming to scale artificial intelligence initiatives while achieving measurable performance improvements.

Williams operates at the intersection of two major forces transforming the financial services sector: the rapid advancement of AI technologies and the growing complexity of data-driven marketing strategies within regulated environments.

Having risen from an engineer to CEO of Marketing Evolution, his career experience in developing modern marketing infrastructure informs his approach to helping fintech firms better understand performance metrics in an increasingly fragmented industry landscape.

At its core, Marketing Evolution focuses on assisting organizations transform raw marketing data into actionable intelligence. The company’s Marketing Performance Substrate consolidates disparate data sources into a unified framework for marketing insights. It fills in data gaps, incorporates causal analysis capabilities, and continuously updates performance data to create a reliable record system that can be utilized by both human teams and AI systems.

As AI models become increasingly standardized, Williams argues that the true differentiator for fintech companies lies not in the sophistication of those models themselves but in the strength of their underlying data infrastructure. This distinction is especially important in fintech, where strict privacy, governance, and compliance requirements place significant restrictions on how data can be utilized. Additionally, customer journeys — from account opening to credit acquisition — are typically nonlinear, involving multiple channels, stakeholders, and touchpoints over extended periods.

Stephen Williams, CEO at Marketing Evolution

By anonymizing data and reconstructing customer journeys using probabilistic methods, Marketing Evolution enables financial institutions to obtain a clearer, compliant view of how marketing efforts drive business outcomes, bringing much-needed transparency to one of the industry’s most complex measurement challenges.

Here, Williams explains why data — rather than AI models — is becoming the key factor determining the success of artificial intelligence initiatives and outlines how fintech companies can bridge the gap between strategic goals and actual execution.

If marketing progress outpaces infrastructure development, what impacts do these disparities have on CMOs in regulated financial services?

Let’s begin by examining the broader challenges facing organizations today: across all industries and roles, there is growing pressure to leverage artificial intelligence to work more efficiently, reduce costs, and discover new sources of value. However, this pressure is particularly intense for CMOs, as their responsibilities have evolved significantly over time.

Today, CMOs are held accountable not only for brand management but also for driving revenue and overall business performance. At the same time, they are expected to adopt artificial intelligence solutions quickly, demonstrate clear return on investment, and make decisions across increasingly fragmented systems and data sources.

In regulated financial services, this challenge becomes even more formidable due to long customer journeys, multiple stakeholders involved, strict compliance requirements, and significant data limitations.

Creating a unified environment that accurately reflects the business operations is what ultimately enables meaningful applications of artificial intelligence. Once such a foundation exists, organizations can layer models from companies like OpenAI, Anthropic, Google, and others on top of this consolidated data to derive value from existing information sources.

I describe it this way: every business — especially those in financial services — possesses vast amounts of valuable data. But before that data can generate value, it must be organized, refined, structured, and contextualized.

Another issue at play is the widespread belief among many organizations that the greater risk lies in falling behind in AI adoption rather than making mistakes in implementation. This mindset creates immense pressure to act swiftly, even when the underlying data infrastructure has not yet been fully established.

While artificial intelligence may provide some initial benefits, organizations will see diminishing returns if they lack the appropriate infrastructure to support its use.

A unified data environment serves as the essential foundation for meaningful applications of artificial intelligence. Credit: Getty Images

What is the “intermediary gap” costing marketers in banking and insurance attribution efforts?

The associated costs are substantial, though they can be mitigated. After all, marketers optimize only what they are able to measure. As a result, they often overinvest in areas that cannot be properly attributed — particularly early-stage brand activities and upper-funnel marketing efforts that occur well before any conversion occurs.

This challenge is especially pronounced in financial services, where conversions are frequently handled by intermediaries such as brokers, branch offices, advisors, platforms, or other downstream stakeholders. While marketing may drive demand, the actual transaction is often recorded elsewhere within the system.

Consequently, attribution practices tend to credit those closest to the transaction rather than those who originated the demand. This is a long-standing measurement issue that becomes more severe in industries where customer journeys are longer, more fragmented, and distributed across multiple stakeholders.

“Today, CMOs are accountable not just for brand, but for revenue and performance ”Stephen WilliamsCEO of Marketing Evolution

To address this issue, it is necessary to integrate all relevant information into a unified data set that accurately reflects business operations. Once these customer journeys can be reconstructed, marketers can begin understanding how brand investments, media exposure, and relationships with intermediaries influence downstream conversion outcomes.

For example, we have worked with one of the largest property insurance companies in the United States. This company has made significant investments at the national brand level and has achieved good results from those efforts. Historically, however, it struggled to link these brand initiatives to actual performance at the policy purchase stage, where local agents also spend marketing resources and ultimately close deals.

The relationship between agencies and insurers is absolutely critical, as it plays a key role in the conversion process. What we helped this company do was reveal the impact of its national brand spending on those downstream conversions.

Now, when they invest in initiatives such as major sports sponsorships or nationwide advertising campaigns, they can track how that investment flows through the system and contributes to measurable conversions at the local agent level — whether in Memphis, Tennessee, or any other location across the country.

Why does artificial intelligence tend to amplify poor data rather than improve it within marketing systems?

This issue is not limited to marketing stacks; artificial intelligence will exacerbate poor data quality across all domains.

No matter how advanced artificial intelligence systems may appear, they ultimately rely on the quality and structure of underlying data to generate accurate insights. And marketing data, in particular, is often of very low quality.

A major reason for this is that marketing data comes from multiple different sources. There is no single source of marketing data — it originates from platforms such as Google Analytics, Meta, Nielsen, Lamar out-of-home advertising, and dozens of other providers. All of these sources produce data with distinct formats and characteristics.

Moreover, the granularity and completeness of this data vary significantly. It must also be integrated with first-party data, which again differs for each advertiser — including email lists, direct mail records, conversion data, web traffic information, and more.

As a result, when looking at marketing data as a whole, we do not see one cohesive dataset ready for use. Instead, there are numerous fragmented datasets that are often incomplete and sparse in certain areas. Some of these datasets cover national levels, others ZIP code regions, and still others DMA areas.

Historically, this has forced data engineers to spend considerable time building custom systems, writing queries, or creating Python scripts to process data on an ad hoc basis for specific analyses.

However, if that data can be structured uniformly, addressing inconsistencies in granularity, and integrated across siloed systems, the situation changes dramatically.

We can break down those structural barriers, unify disparate data sources under one framework, and present information from channels such as Amazon Ads alongside Google, Meta, CTV, linear television, and out-of-home advertising — all using consistent formats and structures. This enables meaningful comparisons rather than comparing apples to oranges or apples to watermelons.

While organizations already have systems of record for customers in Customer Data Platforms, for content in Digital Asset Management systems, and for campaign execution in Marketing Activity Platforms, there is no equivalent system for marketing performance tracking. This is precisely what Marketing Evolution has aimed to address.

The data exists, the tools are available, analysis capabilities are present, and artificial intelligence can provide accurate insights into market dynamics and business operations — rather than amplifying misleading signals.

Ultimately, we must maintain a data-centric approach when developing models, conducting analyses, and utilizing artificial intelligence. Models will continue to evolve rapidly, changing frequently over time.

This is fundamentally a problem related to data infrastructure, not model design. It comes down to establishing the right data infrastructure, feeding appropriate data into that infrastructure, and contextualizing it in alignment with how the business operates.

If the data layer is not properly implemented, CMOs will be stuck dealing with inaccurate insights for years, while CFOs face unaccountable spending patterns. It is critical to get this data layer right as swiftly as possible.

Marketing Evolution CEO Stephen Williams

With 71% of leaders believing they are AI ready and only 37% actually being so — what exactly constitutes the readiness gap?

I believe many business leaders — including myself at one point — rely on a simple criterion: if artificial intelligence tools are in use, the organization must be considered AI-ready. But this is rarely the case.

When we examine the actual conditions required to successfully deploy artificial intelligence or agentic workforces — such as data structure, cleanliness, integration, accessibility, and governance — this assumption quickly falls apart.

We conducted research to better understand this gap by surveying senior marketing executives — including CMOs, VPs, directors, and senior managers — and found that 71% of them believed they were AI-ready.

However, when we assessed whether the foundational data requirements for true AI readiness were actually met, only 37% satisfied those standards.

This represents a substantial gap. Even more striking is the fact that only 3% of surveyed leaders reported consistent improvements in artificial intelligence performance.

In reality, many organizations hit a ceiling in their artificial intelligence efforts very quickly because they lack a solid data foundation. When artificial intelligence systems are forced to reconcile fragmented systems, disconnected data sources, and inconsistent structures, the complexity of the problem increases dramatically, leading to weaker results, incorrect conclusions, and more frequent errors.

Organizations that meet the foundational data requirements for AI readiness achieve roughly twice as strong performance outcomes compared to those without such a foundation. It is essential to get the data layer right in order to extract real value from artificial intelligence — especially as these technologies become more expensive and organizations can no longer afford to experiment indefinitely at low cost.

Key figures from Marketing Evolution

• 71% of senior marketing leaders believe they are AI-ready
• Only 37% of marketing leaders meet the foundational data conditions required for AI readiness
• Just 3% of marketing leaders report seeing consistent AI performance gains

How does a marketing performance substrate address data unification needs for artificial intelligence?

A Marketing Performance Substrate — or a system of record specifically designed for marketing performance tracking — solves this problem because it is inherently intelligent. It can recover missing data, synthetically fill in observational gaps, and reconstruct consumer-level journeys that would not be possible if the data remained stored in static warehouses or data lakes. In this way, it extracts meaningful insights from data before it enters more complex downstream models.

Even prior to this stage, the data has been aligned with a business ontology that reflects the organization’s structure. It is mapped to appropriate schemas, taxonomies, and business contexts, ensuring that naming conventions match how the company internally refers to its creative assets, media channels, campaigns, and customers. This consistency is maintained across all underlying datasets.

By unifying both the data structure and consumer journeys, we open up numerous new opportunities for artificial intelligence applications in areas such as discovery, analysis, and decision-making.

This is not an exaggeration — once artificial intelligence operates against a truly unified and contextualized understanding of business operations, the potential becomes almost limitless.

Organizations with AI-ready data foundations achieve up to twice as strong artificial intelligence outcomes. Credit: Getty Images

How would you advise organizations shifting from backward-looking measurement to forward-looking proactive intelligence?

It is indeed a challenging transition that requires a significant shift in perspective. Until recently, marketing measurement has been retrospective — campaigns are executed, results are awaited, and then analyzed afterward. But now we have the data and tools needed to simulate potential outcomes before any action is taken.

This completely changes the role of marketing intelligence. Instead of simply measuring results after they occur, organizations can begin modeling likely outcomes in advance by using data from previous months, quarters, or years to forecast what is likely to happen next. This enables marketers to mitigate risks, make faster and more informed decisions, and defend those decisions with data-driven evidence.

“No matter how sentient AI may appear, it still fundamentally depends on the quality and structure of the underlying data ”Stephen WilliamsCEO of Marketing Evolution

It also provides much more comprehensive full-funnel visibility. To illustrate this concept, consider how human vision works. Most people assume that what we see is simply the visual cortex processing light entering our eyes. In reality, the brain constantly runs a predictive model based on past sensory experiences to anticipate what it will see.

This predictive model, shaped by a lifetime of structured sensory data, continuously predicts the world around us. Incoming visual information updates this model, but much of what we perceive is actually the brain’s prediction of what is about to occur.

This is how we manage tasks like catching a baseball or reacting quickly to a car braking suddenly in front of us — we are not simply responding in real time to raw sensory input but are acting based on predictions informed by prior data and experience.

I believe marketing is moving in a very similar direction.

Organizations with strong data foundations will be able to not only explain past performance but also predict and simulate future outcomes.

Stephen Williams, CEO of Marketing Evolution, has built his career on developing the systems that power modern marketing

How would you advise a brand that claims, “I’m working with an MMM partner. Where do I go from here?”

If you are already collaborating with an MMM partner, that is an excellent starting point because you already possess historical insights into media performance. This knowledge can be built upon to create a more predictive intelligence layer. Your existing MMM partnership can continue to support the system as you transition toward more adaptive decision-making approaches.

However, the next step requires investing in upgrading your underlying data infrastructure to accommodate more granular and integrated data sources.

It also means evolving toward measurement methodologies and models that can respond more swiftly to changing market conditions, shifting media spending patterns, and evolving consumer behavior compared to traditional MMM solutions.

Consider the impact of events like COVID-19, during which consumer behavior changed dramatically on a monthly basis. MMM systems eventually caught up to these shifts, but often after they had already occurred.

A more dynamic intelligence layer — one that operates on a weekly or even daily basis — can keep pace with business developments in real time, identify changing trends as they happen, and help organizations respond faster and with greater confidence.

This represents the broader shift occurring in the industry. Historically, the martech stack consisted of separate systems such as Customer Data Platforms, CRM systems, Digital Asset Management platforms, Marketing Activity Platforms, and measurement tools — each serving distinct operational functions. Increasingly, however, organizations need an integrated intelligence layer that connects these systems and makes the data accessible for artificial intelligence applications, simulations, optimization, and decision-making. This is the direction the industry is heading.

How should brands approach building their martech stack today?

In my view, a modern martech stack should begin with an assessment of existing data sources and systems — whether that includes Customer Data Platforms, CRM systems, DSPs, or any other relevant tools — ensuring that these systems can work together rather than operate in isolation.

Traditionally, the martech stack has been viewed as a layered structure. Yet increasingly, it resembles a spoke-and-wheel model, with a central data and intelligence layer providing decision support across the entire organization.

This central hub interacts bidirectionally with systems such as Customer Data Platforms — pulling data from them while also feeding information back to improve understanding of customer behavior and overall ecosystem dynamics.

Regarding measurement, we are now in an era where measurement processes can keep pace with the speed at which data is generated. We no longer need to rely solely on MMM solutions. While MMMs still have their place, especially for analyzing longer-term trends, marketers need rapid decision-making capabilities — sometimes even on a daily basis — which requires more agile measurement systems.

Therefore, the modern martech stack should be centered around a unified data layer that supports fast measurement, optimization, simulation, and decision-making, while still maintaining seamless integration with other systems across the organization.