The Path to Sales Lift Through Attention AI

This article is the first part of a series of posts that explore how AI-powered synthetic attention data opens up the possibility of greater creative effectiveness at scale. They are foundational to our contribution to the "WARC "Guide to Attention," which WARC members can access.

Our first article deals with how the AI (artificial intelligence) revolution has caused the amount of creative brands can produce to balloon, leading to challenges in measuring ad effectiveness.

Brands now allocate most of their advertising investments to digital and programmatic platforms, especially paid social and video, and connected TV (CTV). According to WARC, five platforms alone – Alphabet, Amazon, Meta, Alibaba and Bytedance – will drive over 50% of 2024 global ad spend. Adoption of these platforms is driving a proliferation of ads targeted to finer audience segments, and even individuals, via Dynamic Content Optimization (DCO) and Generative AI (GenAI) production tools. Consequently, advertisers’ creative production is ballooning, in some cases by orders of magnitude. 

This big opportunity introduces an equally big challenge: brand advertisers have had no way to economically or efficiently test every single ad for effectiveness in a world of fast-moving, high-volume, dynamic ad portfolios. Some advertisers have adopted video view-through rate as a KPI, though it usually lacks a strong relationship with sales. While direct-response advertisers can allocate campaign budget to validate creative effectiveness through A/B testing in-market, creative effectiveness is a profound challenge for brand-led and consumer packaged-goods companies. 

Their advertising impact is strongest when they optimize toward validated future-sales outcomes. Yet these advertisers typically don’t receive sales data until months after a campaign ends. This is profound when you consider that NCSolutions once found that creative is responsible for 49% of sales impact in advertising. The problem is getting worse, as ads proliferate without robust pre-testing.

One global CPG company pre-tested up to 3,000 ads annually, though its asset volume ballooned in the past two years to over 100,000 ads, as brand leaders adopt GenAI and modular creative production techniques. Platform media investments like Meta, YouTube, TikTok and Snap require more platform-specific creative, curated for finer audience segments. A brand manager at a large CPG recently approached Realeyes to test 180 ads built with GenAI in a single day, a 12x increase in volume versus past scenarios with traditional creative production. 

Figure1: Attentive Reach Advertising Model

AI Bridges Research with Big Data Analytics for Sales Outcomes
How can brand advertisers build ads for sales outcomes, as GenAI and other creative tools fuel ad creative volume? Fight AI with AI.

AI is proving to be a useful technology for advertising measurement and analytics. It is bridging the gap between once-in-awhile market research on humans, and the programmatic world of always-on instant scale. It is enabling innovators to model historical measurement databases of human response into instant prediction systems that perform a good job of estimating how humans will respond in the future. 

For example, Realeyes’ partner Kantar has deployed AI to its historical database of tens –of thousands of human copy tests and its norms database to quickly predict brand lift of an ad creative without engaging a human test participant. This is often called “Synthetic Data.”

Within that, the intersection of attention measurement and AI is especially powerful. According to the Advertising Research Foundation’s Advertising Model, attention is the dividing line between ad stimuli and ad effect. Moreover, multiple validations have proven that visual attention to an ad is among the strongest signals of campaign outcomes. Media platform Teads discovered that a 5% increase in attention to a creative can result in up to a 40% increase in awareness. A large consumer electronics brand found that its high-attention ad creatives delivered 55% stronger shopping conversions.


Figure 2: The ARF’s Advertising Model Concerning Attention

Source: ARF Ad Response Model 2017

Most recently, Realeyes applied AI to its database of 17 million human webcam tests of visual attention (aka “Passive Attention”) and emotions (aka “Active Attention”) to enable instant scoring of ad creatives (aka “Synthetic Attention”) and auto-generated creative recommendations that link to sales outcomes – all by ingesting the creative, without human test participants. 

Mars Inc., for example, reported a 3-5% sales lift potential in its first validation of AI-powered attention scoring and decision-making. The synthetic attention prediction model already is achieving up to 78% accuracy to the ground truth of human training data using visual and auditory inputs. Accuracy is expected to increase above 80% with inputs of demographics and platform environments to the model’s training set.

Figure 3: Relationship Between AI-Powered Attention Score and Sales Lift

Source: I-COM 2024, Mars Inc validation of ad Attention predictions to Sales Lift 

In effect, attention AI is not about attention. Rather, attention AI is the path to sales lift, at scale, potentially for any ad. By appending a performance signal to each ad, scene and individual video frame, AI enables instant analysis and real-time decision-making, providing advertisers with the data they need to make informed choices though an ad’s lifecycle. 
This applies to early conceptual stages of creative strategy, and to the final ads running on large media platforms. Synthetic attention data is assisting manual human workflow, as well as powering automated, programmatic ad creation and delivery systems.

In our next blog, we'll discuss some of our recent partnerships and further explore how Synthetic AI drives scale.

Putting Synthetic AI Attention Measurement at the Center of Everything		Three Steps to Success with Attention AI