Marketing measurement lessons learned and what to expect in 2025

2025 will be the year marketing teams either master measurement or fall behind. With a greater focus on marketing mix modeling (MMM) and triangulation, teams are unlocking new levels of precision. Also, innovative marketing intelligence solutions are giving data-focused companies an edge. 

Among these recent shifts, there have been key lessons we all should be paying attention to. Read on for a comprehensive overview of the evolving measurement landscape and upcoming developments coming in 2025. If you're looking for a measurement blueprint for the future, this might be the most important article you read this year.

Recent advancements in measurement 

The past couple of years have been rough for measurement — 55% of US marketers believe that a poorly integrated data environment has caused a loss of revenue, and 34% of CMOs don’t trust their data. As organizations navigate these challenges, more insightful measurement approaches like marketing mix modeling (MMM) and triangulation have begun to take center stage, reshaping the way we think about data and decision-making. 

Triangulation became a buzzword

Triangulation has recently emerged as a key buzzword in the conversation around measurement. While the concept isn’t new, its widespread adoption in 2024 made it one of the defining trends.

A standout moment came when ThinkGoogle, under Ana Carreira Vidal’s leadership, released its playbook, offering a roadmap for implementing triangulation with Google’s tools and frameworks.

Why this matters

ThinkGoogle’s playbook offers practical guidance for performance marketing managers, showing how to connect systems and leverage experimentation results to fine-tune bidding strategies on Google Ads. It also introduces a fresh perspective on triangulation, framing it as the foundation of modern measurement.

We had Ana Carreira Vidal, the author of the playbook, on our Marketing Measurement Matters podcast. You can watch it here.

While triangulation became the buzzword of 2024, automation also made big waves — especially in how companies approach testing.

Testing became more automated

Testing became more automated in 2024, leading to more effective experimentation compared to 2023. As a result, companies grew bolder, diving deeper into testing with greater ambition. 

Meta played a key role in this shift, releasing updates to Robyn (its AI and machine learning-powered marketing mix modeling package) that enable two-way data integration for measurement models. 

Another relatively recent major release from Meta was Advantage+, an AI-powered tool suite that automates ad calibration based on experiment data. This suite optimizes ad strategies across various formats and surfaces, reducing manual work while helping advertisers better target their audiences.

Lessons learned in marketing measurement 

Lessons learned in measurement are as valuable as the breakthroughs themselves. Here’s what the industry learned in 2024 and how these lessons are shaping measurement today.

A new gold standard was needed for a holistic means of advertising measurement

Turning theory into practice, Meta’s Igor Skokan teamed up with academic Dr. Julian Runge to co-author a paper titled 'A New Gold Standard for Digital Ad Measurement.' The paper delves into the resurgence of marketing mix models in the digital advertising space, particularly in response to Apple’s new tracking limits for advertisers. 

This shift is crucial for marketers as traditional, deterministic methods of measuring digital ad effectiveness, like attribution, are becoming harder to implement. Companies that fail to adapt risk losing critical insights into their marketing strategies. 

The paper positions MMM as the new gold standard for measurement, especially in light of these data constraints. It also highlights how combining MMM with ad experiments and measurement triangulation can offer more reliable insights, making this fuller approach potentially the future of ad measurement in a more data-limited environment.

The article even breaks down the number of experiments to run and how:

Of course, adopting MMM isn’t without challenges. One of the biggest? Bias in observational data — which brings us to our next lesson.

We need to combat misinformation in business analytics 

Dr. Julian Runge's paper, shared on Eric Seufert’s blog, highlights a significant challenge in marketing mix modeling (MMM): the reliance on observational causal inference.

What is observational causal inference?

In marketing measurement, causal inference refers to estimating how changes in one variable, like ad spend, affect another variable, like sales. Observational causal inference specifically uses data that wasn’t gathered through a controlled experiment. In other words, instead of randomly assigning subjects to different conditions, you're working with data that just happens naturally from day-to-day business activities — like sales data over time or advertising spend across campaigns.

This approach is often the best option available, as most marketing data is observational by nature. However, it comes with challenges, particularly the potential for bias. Since the data wasn't gathered in a controlled experiment, there might be external factors affecting the outcome that we can't account for. 

For example, if we look at the relationship between ad spend and sales, there could be other influences, like seasonality or changes in the economy, that affect sales, making it harder to isolate the true impact of advertising.

Why observational causal inference can be tricky

An example of observational causal inference outside of marketing is smoking. It's impossible to randomly assign people to “smokes” or “doesn’t smoke” for an experiment — it's both unethical and impractical. Similarly, in marketing, we can’t randomly assign customers to different ad exposures without the data to back it up. 

More often than not, you’re working with observational data generated by the business as it operates, and that’s not always ideal.

The key appeal of this type of data, however, is that it’s readily available, which makes it a useful tool. Because the data isn’t randomized, however, it’s subject to factors that may not be accounted for, making it harder to draw clear conclusions about cause and effect. This is where calibration and triangulation play a role.

Why calibration and triangulation matter

The paper by Runge emphasizes the importance of calibration and triangulation when working with observational causal inference. Marketers can address biases by using additional data sources or complementary methods. By combining different approaches — such as incorporating experimental data or using multiple analytical models — businesses can refine their estimates and reduce the risk of inaccurate conclusions.

So how do you better design your measurement strategies to overcome biases, experiment, calibrate and validate? Lesson three provides the way.

Learning agendas provide an action plan for improving effectiveness

A learning agenda isn’t just about organizing knowledge. It’s a specific tool marketers can use to identify critical gaps in understanding, test hypotheses and use data-driven insights to refine marketing strategies. Think of it as a strategy roadmap designed to answer critical questions like: “What’s the most effective way to allocate budget across channels?” or “What factors are driving conversions on our website?”

For example, imagine your data team is working with a limited budget and struggling to determine whether their spending on social media or paid search ads delivers higher returns. By creating a learning agenda, the team could break down this challenge into specific, testable questions, such as:

• “Does paid search drive more conversions in certain geographic areas?”
• “How does the performance of social media campaigns vary by demographic?”
• “What’s the relationship between brand awareness and conversion rates?”

Once those questions are identified, your team can create a structured approach to gather and analyze data through methods like MMM, A/B experiments or attribution studies. 

Why it matters

The agenda provides marketers and data analysts with a clear structure to focus their efforts and allocate resources effectively. For instance, once your data team has identified the right questions, they can use methods like MMM to triangulate insights across various marketing activities and optimize their strategies accordingly.

This is important because:

• It helps reduce fragmentation. In large organizations, marketing teams can often operate in silos. A learning agenda forces alignment across different marketing functions (e.g., media, creative, analytics), ensuring everyone is working toward the same goals.

• It guides measurement strategy. A well-designed agenda shows marketers how to create an effective measurement strategy using different tools and methods. 

• It validates insights through experiments. The agenda encourages using controlled experiments (e.g., A/B tests) to validate hypotheses and ensure that insights are reliable and not just based on assumptions or biased data. 

The MESI Model as a core framework for learning agendas

The MESI model is integral to the learning agenda because it provides a clear, actionable framework for assessing marketing effectiveness across all activities. MESI stands for Measurement, Experimentation, Synthesis and Insights. It's used throughout the learning agenda process to ensure that all insights are derived from valid, actionable data.

You can learn more about implementing learning agendas in the IPA’s report, Making Effectiveness Work.

With these lessons in mind, what should marketers prepare for next? Let’s look at the trends shaping marketing measurement in 2025

2025 projections to get excited about

2025 is bringing a wave of change, and advanced measurement adoption is at the forefront. Approximately 65% of businesses are expected to transition from intuition-based to data-driven decision-making by 2026. Let’s look at what’s expected to unfold this year to make that a reality and how these projections might impact the industry as a whole. 

Better measurement will help bridge the gap

This year, we’ll see the marketing team becoming a central driver of growth, with more people across the organization feeling the benefit of better measurement and more actionable insights. More individuals outside of the marketing and data teams will become familiar with MMM as it's integrated into broader strategy and financial planning. 

At the same time, we’ll see a rise in source-based triangulation solutions, which have been limited so far. While there’s already a range of MMM, MTA and testing solutions out there, 2025 will bring a surge in integrated solutions as more players — from tech giants like Adobe to specialized analytics firms — enter the marketing intelligence space. As competition grows, solutions will become even more accessible, leading to more sophisticated, agile marketing strategies.

A unified approach to measurement

The World Federation of Advertisers' Project Halo is paving the way for a unified approach to media measurement, starting with a single, consistent data source across both online and offline media. 

The first phase will focus on providing duplicated region frequency from one system, which is a major step forward. The progress made so far using this data in MMM for planning and optimization is already exciting, and it’s only set to expand. 

With Project Halo set to launch in the UK and then expand to the US, the future looks bright for performance marketers, offering better insights for faster, more informed decisions

Expansion of Meta’s MMM capabilities

Meta’s Robyn, now available in Python, is taking multi-channel optimization to the next level. Soon, it will allow marketers to seamlessly analyze performance across multiple platforms, all in one place, making it easier to spot what’s working and what’s not.

But that’s not all. Meta is pushing the envelope by adding real-time incremental measurement. This means marketers will be able to tweak their strategies on the fly, based on up-to-the-minute performance data. No more waiting around for reports; adjustments will happen instantly, giving you the agility to stay ahead.

Plus, with even deeper AI integration, Robyn will soon offer automated recommendations for reallocating your budget. Say goodbye to manual adjustments. AI will take care of them, helping you focus on making smarter decisions faster.

A move from session-based to event-based measurement

With growing privacy regulations and the phasing out of third-party cookies, brands will have to prioritize first-party data collection through direct consumer interactions and owned platforms to get accurate insights. This means switching from session-based to event-based tracking is crucial for businesses to stay competitive. 

Unlike session-based tracking, which groups all user interactions into a single session, event-based tracking (in Google Analytics 4, for example) captures specific actions, like clicks, video views or purchases, across different platforms and devices. This level of granularity allows businesses to track precise user behavior, leading to better insights into what’s driving conversions.

For example, Watches of Switzerland Group (WOSG) leveraged event-based tracking to optimize campaigns during key sales events, like Christmas and Valentine’s Day. With event-based data, they could pinpoint exactly which interactions were influencing purchases, leading to more effective budget allocation and higher ROI.

Event-based tracking provides a clearer picture of the entire customer journey, improving attribution accuracy and decision-making. As cookies are phased out, this also allows businesses to gather meaningful insights in a more privacy-conscious way, ensuring they remain compliant without sacrificing data quality. 

Building on the rise of event-based tracking, marketers are also beginning to harness AI to move from reactive analysis to predictive strategy.

All-in-one AI-driven solutions

As automated attribution continues to evolve, we’re moving toward a future where these models go beyond just combining MMM, MTA and testing. One significant development will be the integration of Advanced AI models that don’t just analyze marketing effectiveness but also predict future trends based on social signals and emerging market conditions. 

For example, companies like Adobe are already researching the use of AI-driven predictive attribution models that can analyze real-time social media chatter, news sentiment and competitor activity to predict how different marketing strategies will perform in the coming weeks or months. 

These models would allow businesses to adjust their marketing strategies in anticipation of shifts in consumer behavior before they even occur.

Integration of IoT data into attribution models

As more devices become “smart” and connected, IoT data from things like wearables, smart home devices and even connected cars will start influencing attribution decisions. 

A retail company might receive data from a customer’s wearable device that tracks their fitness activity. When it syncs with an online ad campaign promoting a new sports gear line, the attribution system can instantly update the customer journey with this new data point. 

Quantum computing solving complex business problems

The introduction of quantum computing might seem far off, but it could revolutionize attribution models by enabling them to handle an even larger volume of data and provide near-instantaneous optimization recommendations. Companies like Google are already researching how quantum computing can accelerate data processing for marketing attribution. 

This technology could allow real-time, hyper-personalized attribution, where every user’s unique behaviors, preferences and purchase patterns are continuously fed into the system for immediate updates on the most effective marketing strategies.

These new advancements in AI, IoT and even quantum computing are setting the stage for an even more automated and intelligent future in attribution. Businesses will soon be able to make data-driven decisions in real time, leveraging a wealth of new, dynamic data sources to get a clearer picture of their marketing effectiveness and adjust strategies on the fly. 

The future of automated attribution is about actionable intelligence at scale, powered by technology that’s evolving faster than ever.

Triangulation and the future of measurement

As we look ahead, it’s clear that triangulation remains at the heart of everything, shaping not just measurement strategy but the future of marketing itself.

Marketing measurement is no longer just about proving ROI — it’s about driving smarter strategy across the entire business. In the future, the integration of triangulation, advanced modeling and automation will shift measurement from a siloed function into a strategic advantage.

The most successful teams will be those who treat measurement as a continuous learning process — one grounded in experimentation, calibrated with data and aligned across every organization function. Whether you’re exploring new tools, refining your learning agenda, or building your next MMM model, now’s the time to put these insights into action.

To dive deeper into what’s coming next — including how Meta is pushing the envelope on automation — check out our full conversation with Igor Skokan on the Marketing Measurement Matters podcast: