Empowering Ad Effectiveness for Mars with Realeyes and Super Protocol

Case Study Snapshot: Realeyes & Mars

Realeyes, leveraging Super Protocol's confidential AI cloud powered by NVIDIA H100/H200 GPU TEEs and Intel TDX, unlocked 4x more sensitive training data. This boosted their AI's prediction accuracy, helping Mars achieve a 3-5% sales lift across 19 markets by optimizing ad creatives.

The Challenge: Optimizing Ad Spend in a Data-Sensitive World

As global digital ad spend approaches USD 600 billion annually, brands like Mars, Incorporated face the critical challenge of optimizing advertising campaign performance. Billions can be wasted on low-performing creatives. While modern techniques like AI-driven facial coding offer significant advantages over traditional lab testing ($25,000+ per ad), they hit a crucial bottleneck: accessing sufficient high-quality training data. Privacy regulations (GDPR, CCPA), compliance requirements, and concerns over proprietary data control often keep the most valuable datasets locked away.

Facial Coding vs. Traditional Testing:

Faster, Cheaper, Scalable: Analyzes consumer expressions via camera, ideal for large volumes of digital ads.
Eliminates Logistics: Removes costs and challenges of lab setups.
Proven Potential: Early use by Mars showed promise in predicting ad performance.
The Data Hurdle: Progress stalled due to the need for more high-quality data, hampered by privacy concerns.

Case in Hand: Realeyes & Mars Collaboration

Realeyes, a leader in AI-based attention and emotion measurement, partnered with Mars, Incorporated to identify high-impact advertising creatives.

Partnership Scope

Vast Database: Realeyes leverages data from 17 million+ creative tests (4,000 data points linking attention to sales).
Brand Reach: Collaboration spans 30+ Mars brands (including Sheba) across pet food, confectionery, and food categories.
Testing Volume: ~600 advertising materials tested annually (TV, digital, mobile, short videos).
Global Scale: Applied in 19+ countries (including China, Australia, Brazil).

Despite this scale, the core challenge remained: enhancing AI model accuracy required accessing more diverse, high-quality data from third parties, balanced against strict data confidentiality needs.

Traditional Cloud AI Training Environment: Key Constraints

Conventional cloud environments presented significant hurdles for Realeyes:

Locked Data Sources: Providers of sensitive/proprietary datasets hesitated due to inadequate guarantees of privacy and ownership.
Data Vulnerability: Standard cloud services risked exposing confidential user data (like biometrics from facial coding) to breaches or misuse.
Regulatory Complexity: Evolving laws like GDPR and CCPA complicated cross-border data sharing and added compliance burdens.

The Super Protocol Solution: Confidential AI Cloud

To overcome these barriers, Realeyes adopted Super Protocol – a "trusted-by-design" confidential AI cloud.

Unlock Confidential Data

Provides verifiable privacy assurances, encouraging data owners to share previously restricted, high-value datasets.

Protect Data-in-Use

Utilizes TEEs (NVIDIA GPU TEEs & Intel TDX), ensuring raw data remains encrypted and secure within enclaves during processing.

Ensure Self-Sovereignty

Decentralized governance and smart-contract orchestration prevent unauthorized control or alteration of data processing workflows.

"Since we started using Super Protocol, we have unlocked more high-quality data, facilitating a 319% growth of our AI training datasets. This ensures privacy and compliance while driving better ad performance globally and achieving a significant 3 – 5% sales lift for Mars, with predictions becoming more precise and actionable."

– Elnar Hajiyev, Co-founder and CTO at Realeyes.

How Realeyes Is Leveraging Super Protocol

By integrating Super Protocol's confidential computing environment powered by NVIDIA GPU TEEs and Intel TDX, Realeyes successfully reduced data-sharing barriers. With privacy, compliance, and security concerns addressed, Realeyes onboarded data from multiple third-party providers, significantly expanding the range and quality of datasets fueling its ad efficiency prediction model.

Business Impact: Measurable Results

The adoption of Super Protocol yielded significant benefits:

Predictive Precision & Sales Lift

Achieved a 3-5% forecasted incremental sales lift for Mars across 19 markets by identifying high-impact creatives more accurately, reducing reliance on expensive large-scale testing.

Enhanced User Confidence & Trust

Assurances of data confidentiality encouraged more data providers to participate in partnerships, enriching the available data pool.

Improved AI Prediction Power

A 319% increase in training data led to refined AI accuracy and deeper insights for global ad optimization.

Broader Industry Applications

Building on this success, Realeyes is extending its secure AI measurement solutions to other sensitive sectors like finance, medtech, and biometrics.

Technical Deployment Details

Infrastructure Overview

Super Protocol provides a confidential and self-sovereign AI cloud and marketplace, enabling secure collaboration on AI projects. Its decentralized infrastructure runs confidential workloads across independently operated nodes using smart-contract orchestration, ensuring privacy, ownership, control, security, scalability, and flexibility without reliance on a single cloud provider.

Hardware & Configuration for Realeyes Project:

Distributed Nodes: Workloads ran across two independent, physically distributed locations.
Location 1 (Gcore Colocation): Exxact 2U servers with NVIDIA H100 GPUs and 5th Gen Intel Xeon Scalable CPUs (operated by an independent TEE provider).
Location 2 (Engage Stack Cloud): Lenovo SR680a HGX 8-GPU system with NVIDIA H200 GPUs, 5th Gen Intel Xeon Scalable CPUs, and 4 NVSwitch modules.
Confidential Computing Mode: All systems utilized Intel TDX and NVIDIA GPU TEEs.
Operating System: Ubuntu OS.
Outcome: Trusted, verifiable AI execution across a distributed, multi-provider, confidential environment.

This case study demonstrates how Super Protocol's confidential AI cloud, combining advanced TEE hardware (NVIDIA GPUs and Intel TDX) with decentralized orchestration, can unlock critical data, enhance AI capabilities, and drive significant business outcomes in privacy-sensitive industries like advertising analytics.