For almost two decades, marketing measurement has been organized around an implicit question: what channel deserves credit for a sale. Google, Meta, email, retail media, each with its own argument about its contribution to the bottom line. This attribution logic ended up fragmenting organizations into silos that optimize their own metrics, with no one observing the complete system.
Fernanda Rivera, Omnichannel Measurement Strategy Director at Entropy and former leader of Google Marketing Platform for Latin America, argues that this fragmentation has a cost that is rarely mentioned in investment committees:
The business does not grow by departments, it grows as a system, and measuring it in parts produces decisions misaligned with the total growth of the company.
From sharing the cake to making it bigger
Fernanda Rivera precisely describes the structural problem behind digital attribution: each area ended up defending its own merit. Branding claims to have built the brand, performance claims to have closed the sale, retail media claims the channel where the transaction took place. They all have some truth, and they are all moving in different directions.
The alternative she proposes is a change of question. Instead of sharing credit among channels that already participated in a sale, incrementalism asks whether that investment generated customers and revenue that would not have existed without it.
It’s the difference between sharing the slices of a cake and understanding who made the cake grow.
Conventional attribution rewards volume over quality. By assigning value to the click or impression prior to conversion, the more times you impacted a person, the higher the probability that the medium appeared within the attribution model.
In contrast, new models reward quality over volume, because a medium must demonstrate that it truly succeeded in generating an increase in total sales: whether comparing a control group against an exposed one or incorporating its effect within an MMM. That is, the impact must have been significant enough to modify the actual behavior of the consumer and generate incremental sales.
Privacy did not break measurement: it exposed a prior problem
The tightening of digital privacy and the progressive end of cookies are often presented as the cause of the current measurement crisis. Fernanda offers a different reading: the promise of tracking each user throughout their purchase journey already had a side effect before privacy interrupted it; it had fragmented organizations into performance teams, branding teams, offline sales, and online sales, each with separate budgets and objectives.
Consumer behavior never respected that internal organization. The same person who sees an advertising billboard also shops online hours later. According to her, the underlying problem is not technological; it is governance: defining a common goal of incremental growth that guides investment decisions over the silos that attribution helped to build.
Marketing Science ceased to be just a marketing matter
One of Fernanda's central propositions is that the discipline integrating this data, marketing science, no longer exclusively belongs to the marketing area.
Its function is to become the common language that aligns marketing, finance, and general management around the same objective: maximizing the value created for the business.
This integration begins to rely on three methodologies that are gaining traction among business leaders: marketing mix modeling, experimentation, and causal inference, clearly distinguishing their roles.
• The Marketing Mix Modeling helps to understand what is happening in the business as a whole.
• Experimentation shows what is causing it.
• Causal inference does not appear and is not explained?? Investigating, we can leave it like this: it allows identifying and quantifying cause-and-effect relationships from observational data, when it is not possible to conduct an experiment.
Used together, they allow answering questions that have been resolved for years with assumptions: how many new sales does activating a channel generate, and how many of those users would not have arrived without that investment.
This approach also implies abandoning the idea of a single source of truth, a concept that dominated the measurement conversation for years. Fernanda suggests that the way forward requires observing attribution, correlation, and causation together, without relying on a single model to explain everything that happens in the business.
Artificial intelligence: faster access to econometrics, with new risks
Artificial intelligence is notably accelerating the adoption of measurement tools that previously required highly specialized teams. Fernanda Rivera illustrates this with a concrete example from her own professional experience:
Building a Marketing Mix Modeling used to involve sending the work to teams of mathematicians in other regions of the world, a slow and expensive process. Today, there are software solutions that drastically reduce that time.
But that same democratized access brings a risk that is not minimized: a poorly built model can cause more harm than good for an organization, as investment decisions are directly based on its results. The speed offered by artificial intelligence does not replace the technical knowledge necessary to correctly interpret what a model is showing.
This nuance connects with another risk that Fernanda identifies in organizations with access to more data than ever:
Confusing volume of information with actionable knowledge. Her recommendation is to abandon inherited metrics, even those deeply embedded in the culture of a company like the traditional funnel, and anchor measurement to what effectively appears in the company's income statement. The rest, she warns, is jargon with no direct impact on the business.
The pending conversation between CEOs, CMOs, and CFOs
For Fernanda Rivera, the most urgent gap that business leaders face today is not technological or data infrastructure; it is a shared language. As long as each department continues to optimize its own indicators, the company as a whole will never optimize its total growth. Marketing, branding, performance, retail media, and CRM should not compete for budget among themselves, but rather collaborate around a single objective of incremental growth.
The most uncomfortable diagnosis she presents is that the problem no longer resides in the lack of sophisticated dashboards. Many organizations today have AI-driven visualizations in the cloud, and yet continue to answer wrong questions about their business. Those dashboards are also often directly connected to bonuses, budgets, and strategic decisions, which multiplies the cost of measuring incorrectly.
A new governance for investment decisions
The conclusion that arises from Fernanda Rivera's argument is not a criticism of the available technology, but a warning about how it is being used. Measurement tools have never been so accessible or so fast, and yet the risk of making wrong decisions with abundant data remains intact if the benchmark metrics do not change.
For leaders looking to invest, the call is clear: stop asking who gets the credit for a sale and start asking what investment actually grew the company. Organizations that manage to establish that common language between marketing, finance, and general management will have a structural advantage over those that continue to measure their business by separate departments, in an environment where the consumer has always never bought in silos.