For AI in Ads to Work, IAB Says Fix the ‘Garbage In’ Problem First

Why It Matters

Improved data standards will let AI deliver accurate, cross‑channel insights, unlocking billions in reallocated spend and giving emerging media a measurable foothold.

Key Takeaways

• Industry fragmentation hampers AI-driven advertising measurement across multiple platforms
• Inconsistent taxonomies cause “garbage in, garbage out” AI results
• Project Eidos aims to create cross‑channel data standards
• $32 billion could be reallocated by improving data quality
• Standardization will cut manual effort, boost strategic insights

Summary

The Interactive Advertising Bureau (IAB) announced that its new Project Eidos will tackle the “garbage‑in” problem that is limiting AI‑driven advertising measurement. The initiative comes as the ecosystem wrestles with fragmented data sources, privacy‑driven signal loss, and an expanding mix of channels—from CTV and gaming to podcasts and retail media.

Executives highlighted that inconsistent taxonomies and proprietary methodologies force marketers to spend countless hours normalizing data, which degrades AI model performance. IAB’s recent State of Data report estimates roughly $32 billion could be redeployed—$26 billion into under‑represented channels and $9 billion from manual tasks to strategic analysis—if data quality improves.

“Garbage in, garbage out,” the speaker warned, emphasizing that AI merely amplifies existing data flaws. He also clarified that Eidos is not a new technology platform but a collaborative effort to establish common standards that enable measurement providers, brands, agencies, and publishers to speak the same language.

By standardizing cross‑channel metrics, the industry can unlock more reliable AI insights, accelerate decision‑making, and shift spend toward emerging media. The move promises to reduce operational overhead, increase ROI, and give emerging channels the analytical support needed for growth.

Original Description

Artificial intelligence is often hailed as the solution to advertising’s most persistent measurement headaches, but could itn be doing more to illuminate the problem than to solve it?

After all, AI is only as good as the data it is fed. That means the perennial challenges of fragmented platforms, data silos, and inconsistent taxonomies risk being amplified, not erased, by the emergence of new channels.

The Interactive Advertising Bureau (IAB) is stepping in with a new initiative, dubbed Project Eidos, designed to address the foundational issues that have plagued measurement for decades.

The goal is to clean up the data pipeline upstream, creating a common language for the industry before data ever reaches an AI model, said Angelina Eng, VP, measurement, addressability and data center, IAB, in this video interview with Beet.TV.

Getting to the root of the problem

“One of the things that we plan to do with Project Eidos is really get to the root of the problem,” Eng said. “Over the years, a lot of companies have created their own proprietary approaches, their methodologies, creating their own taxonomies and that has resulted in a lot of frustration.”

This frustration has a direct impact on the effectiveness of advanced modeling techniques like marketing mix modeling (MMM) and the application of AI. When data is inconsistent, AI-driven platforms cannot perform optimally, according to Eng. “When you have inconsistent data, it is really hard for the AI to actually do a great job in providing insights and providing results and outcomes because garbage in, garbage out,” she explained.

The ultimate aim of Project Eidos is to give emerging and underrepresented channels a fair shot in media plans by ensuring the data fed into models is clean. “For us, Eidos is about getting to the root of the issues and addressing those challenges that are upstream,” Eng said. The hope is to “create some standards so that we can get to understanding what is attributable from a cross-channel perspective and help feed proper data into an MMM to enable emerging channels.”

A common language for the ecosystem

Eng was clear that the IAB is not looking to compete with the existing ecosystem of measurement companies. Instead, it wants to be a facilitator that improves the inputs for everyone, from brands and agencies to the measurement platforms themselves.

“We’re not building a technology, we’re not building a platform,” she said. “We are, in fact, helping to develop and improving data quality so that those platforms have the right proper data sets to enable their systems to provide better services and outcomes.”

The financial stakes are significant. According to a recent IAB report, improving data quality and processes could unlock billions of dollars in value. The IAB State of Data 2024 report suggests better measurement could free up massive investment. “Approximately $32 billion could be reinvested into the marketplace towards underrepresented channels, emerging channels, as well as improving productivity,” Eng said.

That figure breaks down into two key areas of opportunity, according to Eng. “About $26 billion of media spend could be reinvested more properly and $9 billion could be shifted from manual operational tasks to more strategic insights,” she said. Such a shift would allow teams to focus on analysis rather than on the manual and time-consuming work of data transformation.

Old challenges surfaced by new tech

“I think AI is actually bringing to the surface what has been the challenges over the year,” she said.

Eng’s experience has shown her that the core issues are operational, stemming from inconsistent workflows in campaign activation and data collection. “There’s also different ways that companies activate campaigns and it can be different from not only from brand to brand, but from campaign to campaign,” she said.

By establishing a common framework, the IAB hopes to bring order to this operational chaos, which in turn will fuel the very technologies that are currently struggling with it. “If we can provide a playbook and a set of standards that allows them to have better control and consistency, then we’re improving the data quality,” Eng said. This would “help empower AI to provide better insights more efficiently and faster.”