How analytics and AI are shaping marketing effectiveness measurement - today and tomorrow 

From classical models to AI-enabled decision engine: measurement is becoming continuous, granular, integrated and actionable.  

Two hands shaking in front of a beautiful sunset, with the sun setting in the background. The hands are animated, giving the image a dynamic feel.

How analytics and AI are shaping marketing effectiveness measurement - today and tomorrow  

From classical models to AI-enabled decision engine: measurement is becoming continuous, granular, integrated and actionable. 

Two hands shaking in front of a beautiful sunset, with the sun setting in the background. The hands are animated, giving the image a dynamic feel.
Sunando Das

Sunando Das
Head of Predictive Analytics
Unilever

Das leads the company’s global predictive analytics centre of excellence, where he and his team oversee the development and deployment of AI and ML measurement products across markets, as one of their focus areas. In this exclusive piece for the IPA AI Magazine, he explores the biggest shift ahead: a “relay” from yesterday’s aggregate models to today’s art of the possible – granular, campaign-level, continuous models that harness classical AI to generative and agentic AI augmentation to enable decision intelligence, and, ultimately, quantum + AI.

Three Levels of AI 

Analytics and AI are reshaping how we measure marketing effectiveness, today and tomorrow. However, AI is not a single thing: there are three levels of AI, each of which plays a role: 

  • Classical AI: machine learning (ML) models we have used for more than two decades, still evolving to address new challenges. 
  • Generative and agentic AI. 
  • Quantum + AI. 

The Challenge: Two Gaps to Close 

Marketing effectiveness measurement faces two gaps today: 

  • Outcomes vs. data: the huge gap between the outcomes we want to achieve and the data that exists to enable them. 
  • Dynamic vs. static: the gap between the dynamic nature of media and physical availability levers and the static assumptions in many models. 

What AI Makes Possible 

A vibrant blue and pink abstract background with curved lines, creating a unique and eye-catching design. The colors blend together to create a mesmerizing effect, making it a perfect wallpaper for any device.

The desire for continuous, granular measurement and optimisation across mental and physical availability levers has long been elusive. With AI, it is now possible across four dimensions. 

1. Granularity: Breadth and Depth 

  • Covering the breadth across levers such as media, price, promotion, distribution, shopper marketing and content quality. And within levers such as within media – covering all channels across traditional, digital, retail media, influencer media. 
  • Covering the depth within each lever such as within media, going beyond channels and platforms down to individual campaign level; in price, promotion and distribution, going to region × brand variant × PPG (pack price group) level. 
  • And dissecting the breadth and depth by brand variant/range, offline and online sales. 

The challenge is delivering this granularity in data-sparse markets. In many markets, data is available only at an aggregated monthly level (not at individual customer level); some markets have weekly data; and very few have daily data. With so many features relative to data points in monthly and weekly markets, a classical model cannot deliver this granularity unless we construct the model with too many starting assumptions, which defeats the purpose of modelling. 

2. Continuity: Monthly Updates in 2 Weeks 

Models need to be updated every month within two weeks; otherwise, they will not be actionable. This is not just a data refresh with an additional month’s data, which can be done in 1–2 days: coefficients must also be evaluated and updated where required, which means model evaluation and rebuild across all the features in the model. 

3. Dynamic: Time-Varying Impact 

Classical models assume that each feature’s coefficient of impact stays constant over the modelling period. In today’s dynamic environment, this assumption is artificial: the impact of media channels, platforms and campaigns cannot be the same every week or month. This calls for time-varying models, in which coefficients of impact can vary over time. 

4. Short- and Long-Term Impact 

The short- and long-term impact of advertising must be integrated, and the long-term component decomposed by its different drivers, because different channels with a long-term skew can work in different ways, which affects the optimal media mix. 

Traditional solutions cannot deliver all of this. Harnessing a mix of econometrics and AI can, through a single solution at the convergence of MMM, MTA and experimentation.  

This avoids situations where unifying different solutions that give different results takes a long time. 

 

Beyond Reporting: Predicting What to Do Next 

A purple and blue wavy background with wavy lines, creating a mesmerizing optical illusion.

The solution does more than report past performance. It predicts what to do next, and why, to drive incremental sales and media RoI, with high model accuracy and validation to in-market results. Guardrails keep scenarios realistic to business realities. 

Marketing effectiveness is also broader than measurement and optimisation. It answers strategic marketing questions such as: 

  • How should investment be split between media and promotions? 
  • How can we avoid diluting the long-term contribution of media to sales? 
  • How can we maximise the halo between the master brand and category brands or 
  • individual brand ranges? 
  • How can ‘others say’ advertising balance short- and long-term sales? 
  • How should investment be split between types of influencers? 

It Takes a Village to Land Business Impact 

A large group of people in a variety of colors, creating a vibrant and lively atmosphere. The people are arranged in a circle, with some overlapping each other, and the colors range from bright blues and greens to warm oranges and yellows. The image is animated, giving it a lively and energetic feel.

Deploying such a solution with business impact literally takes a village. A cross-functional core team across media, media agency, insights and the analytics partner works through the details before arriving at the final changes to deploy, with buy-in from brand, finance and executive leadership teams.

Generative and Agentic AI: Accelerating Adoption  

An abstract background with colorful circles on a dark blue background. The circles are arranged in a pattern and the colors range from bright blues and greens to warm oranges and yellows.

Generative and agentic AI will accelerate the pace of deployment and the magnitude of adoption, in three ways: 

  • Discoverability and adoption: an LLM query engine will let end users query any analysis, output or recommendation through simple natural-language prompts. This shifts the consumption layer from dashboards, simulators and PowerPoints to natural-language conversation, improving user experience and adoption. 
  • Analysis and productivity gains: multi-agent systems will help move from modelled outputs to business outcomes and recommendations (with humans in the loop) faster, with substantially less human time involved. This will also enable optimisation and impact monitoring via natural-language prompts. 
  • Interoperability: integrating measurement with other data inputs and solutions, and analysing them together to diagnose the reasons for performance and recommendations. 

Human Judgement Remains Essential 

A close up of a colorful eye on a black background. The eye is vibrant and full of life, with its iris and pupil standing out against the dark background.

Generative and agentic AI improve discoverability (and hence adoption), automate analysis and recommendations (and hence productivity), and enable interoperability (and hence added insight). 

These are assistive tools rather than direct mathematical solvers. 

The core still depends on a mix of econometrics and classical AI. So, as AI gets better at recommendations, asking what role human judgement will play is probably the wrong question. Human judgement is critical in deploying classical AI models and in training agents on decision intelligence. It is also preserved through the human in the loop, who refines and finalises recommendations to account for business realities. 

The Road Ahead

A pink and yellow staircase with a blue sky in the background. The stairs are painted in a combination of pink, yellow, and orange, creating a vibrant and eye-catching scene.

The role of AI will grow further, especially in three areas: 

  • Self-directed AI models: learning from the initial models to generate updates faster, and harnessing learnings from hundreds of models across markets to build models for additional brands in the same markets. This will increase the scale of deployment with no adverse commercial impact. 
  • Integration: better integration of measurement with planning and buying decisions. 
  • Quantum + AI to tackle combinatorial explosion: as model granularity increases and more intermediary variables are introduced to capture the full-funnel impact, optimisation will face a combinatorial explosion that classical AI cannot handle on its own. This is where Quantum + AI will come into play. 

The Bottom Line 

Two hands shaking in front of a beautiful sunset, with the sun setting in the background. The hands are animated, giving the image a dynamic feel.
The future of marketing effectiveness is not human versus machine, or today’s models versus tomorrow’s. It is a relay.  

Classical AI delivers the granular, continuous and dynamic foundation now; generative and agentic AI turns it into conversation and action; Quantum + AI will unlock optimisation at a scale we cannot yet reach. Throughout, human judgement stays in the loop. The winners will be those who close the gap between data and outcomes by harnessing analytics and AI rather than being constrained by the realities of data, and who bring the whole village along to make the models more intelligent and to refine and act on what they say.    

Published by the IPA, the IPA AI Magazine helps member agencies make sense of the fast-changing AI landscape, with practical insights, expert perspectives and real-world applications of generative AI.
Managing Editor: Conor Nichols
Design: Jenna Betts

© 2026 The Institute of Practitioners in Advertising. All rights reserved. No part of this site may be reproduced without our permission.